diff --git a/.gitignore b/.gitignore
index 831ba58..4d8ad3c 100644
--- a/.gitignore
+++ b/.gitignore
@@ -5,3 +5,8 @@
 /**/*~
 !boot_code/crt0.S
 /simplified-runtime.xml
+**.log
+**.ini
+**.wlf
+**transcript
+**.dbg
\ No newline at end of file
diff --git a/.gitmodules b/.gitmodules
new file mode 100644
index 0000000..5ade099
--- /dev/null
+++ b/.gitmodules
@@ -0,0 +1,3 @@
+[submodule "hwpe/neureka/pulp-nnx"]
+	path = hwpe/neureka/pulp-nnx
+	url = https://github.com/pulp-platform/pulp-nnx
diff --git a/astral.yaml b/astral.yaml
new file mode 100644
index 0000000..d8acd6c
--- /dev/null
+++ b/astral.yaml
@@ -0,0 +1,19 @@
+astral:
+  parMatrixMul8:
+    path: ./astral/parMatrixMul8
+    command: make clean all run
+  parMatrixMul16:
+    path: ./astral/parMatrixMul16
+    command: make clean all run
+  parMatrixMul32:
+    path: ./astral/parMatrixMul32
+    command: make clean all run
+  dmr_matmul:
+    path: ./astral/dmr_matmul
+    command: make clean all run
+  redmule:
+    path: ./astral/redmule
+    command: make clean all run
+  neureka:
+    path: ./astral/neureka
+    command: make clean all run
\ No newline at end of file
diff --git a/astral/dmr_matmul b/astral/dmr_matmul
new file mode 120000
index 0000000..c6cc223
--- /dev/null
+++ b/astral/dmr_matmul
@@ -0,0 +1 @@
+../reliability_tests/dmr_matmul
\ No newline at end of file
diff --git a/astral/ecc_test b/astral/ecc_test
new file mode 120000
index 0000000..a80e9ae
--- /dev/null
+++ b/astral/ecc_test
@@ -0,0 +1 @@
+../reliability_tests/ecc_test
\ No newline at end of file
diff --git a/astral/hello/Makefile b/astral/hello/Makefile
new file mode 100644
index 0000000..d145a0f
--- /dev/null
+++ b/astral/hello/Makefile
@@ -0,0 +1,5 @@
+PULP_APP = test
+PULP_APP_SRCS = hello.c
+PULP_CFLAGS = -O3
+
+include $(PULP_SDK_HOME)/install/rules/pulp.mk
diff --git a/astral/hello/hello.c b/astral/hello/hello.c
new file mode 100644
index 0000000..8741206
--- /dev/null
+++ b/astral/hello/hello.c
@@ -0,0 +1,24 @@
+/*
+ * Copyright (C) 2018 ETH Zurich and University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include <stdio.h>
+
+int main()
+{
+  printf("Hello !\n");
+  
+  return 0;
+}
diff --git a/astral/icache_fi_conv16 b/astral/icache_fi_conv16
new file mode 120000
index 0000000..c97475b
--- /dev/null
+++ b/astral/icache_fi_conv16
@@ -0,0 +1 @@
+../reliability_tests/icache_fi_conv16
\ No newline at end of file
diff --git a/astral/neureka b/astral/neureka
new file mode 120000
index 0000000..98cb979
--- /dev/null
+++ b/astral/neureka
@@ -0,0 +1 @@
+../hwpe/neureka/
\ No newline at end of file
diff --git a/astral/parMatrixMul16/Makefile b/astral/parMatrixMul16/Makefile
new file mode 100755
index 0000000..0fade4a
--- /dev/null
+++ b/astral/parMatrixMul16/Makefile
@@ -0,0 +1,8 @@
+PULP_APP = test
+PULP_APP_SRCS = matrixMul.c
+
+PULP_CFLAGS = -O3
+
+include $(PULP_SDK_HOME)/install/rules/pulp.mk
+
+#pulp-bench-reg --name=parMatrixMul16.cycles --module=pulp_rtl_testset --pipeline=$(PIPELINE) --artefact=pulp_rtl_testset --cmd="make run -f Makefile.sdk" --probe-regexp='matrixMul -> success, nr. of errors: 0, execution time: (\d+)' --params="platform($(platformName)),compiler($(OR1K_TOOLCHAIN_TYPE)),nbCores(4),elemSize(16)" --probe-regexp='matrixMulTranspose -> success, nr. of errors: 0, execution time: (\d+)' --params="platform($(platformName)),compiler($(OR1K_TOOLCHAIN_TYPE)),nbCores(4),elemSize(16),transposed"
diff --git a/astral/parMatrixMul16/gen_stimuli.py b/astral/parMatrixMul16/gen_stimuli.py
new file mode 100755
index 0000000..754a8a9
--- /dev/null
+++ b/astral/parMatrixMul16/gen_stimuli.py
@@ -0,0 +1,54 @@
+#!/usr/bin/env python
+
+import sys
+import random
+
+
+def write_arr(f, name, arr):
+    f.write('const short %s[] = {\n' % name)
+    for v in arr:
+        f.write('%d,\n' % (v))
+    f.write('};\n\n')
+    return
+
+################################################################################
+f = open('parMatrixMul16_stimuli.h', 'w')
+
+
+SIZE = 24 
+RANGE = int(2**7/SIZE)
+
+m_a   = []
+m_b   = []
+m_exp = []
+
+for i in range(0,SIZE):
+    for j in range(0,SIZE):
+        a = random.randint(-RANGE, RANGE-1)
+        b = random.randint(-RANGE, RANGE-1)
+
+        m_a.append(a)
+        m_b.append(b)
+
+for i in range(0,SIZE):
+    for j in range(0,SIZE):
+        r = 0
+
+        for k in range (0,SIZE):
+            r = r + m_a[i * SIZE + k] * m_b[k * SIZE + j]
+
+        m_exp.append(r)
+
+
+write_arr(f, 'm_a',   m_a)
+write_arr(f, 'm_b',   m_b)
+write_arr(f, 'm_exp', m_exp)
+
+f.write('#define SIZE %d\n' % SIZE)
+
+
+f.write('__attribute__ ((section(".heapsram"))) short g_mA[SIZE][SIZE];\n')
+f.write('__attribute__ ((section(".heapsram"))) short g_mB[SIZE][SIZE];\n')
+f.write('__attribute__ ((section(".heapsram"))) short g_mC[SIZE][SIZE];\n')
+f.write('__attribute__ ((section(".heapsram"))) short g_mB_tmp[SIZE][SIZE];\n')
+
diff --git a/astral/parMatrixMul16/matrixMul.c b/astral/parMatrixMul16/matrixMul.c
new file mode 100644
index 0000000..5cbe65f
--- /dev/null
+++ b/astral/parMatrixMul16/matrixMul.c
@@ -0,0 +1,177 @@
+/*
+ * Copyright (C) 2018 ETH Zurich and University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+/* 
+ * Mantainer: Luca Valente, luca.valente2@unibo.it
+ */
+
+#include "pulp.h"
+
+#include "parMatrixMul16_stimuli.h"
+
+void check_matrix_mul(testresult_t *result, void (*start)(), void (*stop)());
+void check_matrix_mul_transpose(testresult_t *result, void (*start)(), void (*stop)());
+
+testcase_t testcases[] = {
+  { .name = "matrixMul",          .test = check_matrix_mul           },
+  { .name = "matrixMulTranspose", .test = check_matrix_mul_transpose },
+  {0, 0}
+};
+
+unsigned int num_cores;
+
+int main()
+{
+  if (rt_cluster_id() != 0)
+    return bench_cluster_forward(0);
+
+  num_cores = get_core_num();
+
+  if(rt_core_id() < num_cores) {
+    run_suite(testcases);
+  }
+
+  synch_barrier();
+
+  return 0;
+}
+
+void matrix_init();
+unsigned int matrix_check();
+
+void check_matrix_mul(testresult_t *result, void (*start)(), void (*stop)()) {
+  int core_id;
+  unsigned int i, j, k;
+  unsigned int chunk;
+  unsigned int lb, ub;
+
+  core_id = get_core_id();
+
+  // number of rows each core has to multiply
+  chunk = SIZE / num_cores;
+  // lower bound
+  lb = core_id * chunk;
+  // upper bound
+  ub = lb + chunk;
+
+  if(core_id == 0) {
+    matrix_init();
+  }
+
+  if(num_cores != 1) synch_barrier();
+
+  // start benchmark
+  start();
+
+  for(i = lb; i < ub; i++) {
+    for(j = 0; j < SIZE; j++) {
+      g_mC[i][j] = 0;
+  
+      for(k = 0; k < SIZE; k++) {
+        g_mC[i][j] += g_mA[i][k] * g_mB[k][j];
+      }
+    }
+  }
+
+  if(num_cores != 1) synch_barrier();
+
+  stop();
+
+  if(core_id == 0) {
+    result->errors = matrix_check();
+  }
+}
+
+void check_matrix_mul_transpose(testresult_t *result, void (*start)(), void (*stop)()) {
+  int core_id;
+  unsigned int i, j, k;
+  unsigned int chunk;
+  unsigned int lb, ub;
+
+  core_id = get_core_id();
+
+  // number of rows each core has to multiply
+  chunk = SIZE / num_cores;
+  // lower bound
+  lb = core_id * chunk;
+  // upper bound
+  ub = lb + chunk;
+
+  if(core_id == 0) {
+    matrix_init();
+  }
+
+  if(num_cores != 1) synch_barrier();
+
+  // start benchmark
+  start();
+
+  // transpose array before using it
+  for(i = lb; i < ub; i++) {
+    for(j = 0; j < SIZE; j++) {
+      g_mB_tmp[i][j] = g_mB[j][i];
+    }
+  }
+
+  if(num_cores != 1) synch_barrier();
+
+  for(i = lb; i < ub; i++) {
+    for(j = 0; j < SIZE; j++) {
+      g_mC[i][j] = 0;
+
+      for(k = 0; k < SIZE; k++) {
+        g_mC[i][j] += g_mA[i][k] * g_mB_tmp[j][k];
+      }
+    }
+  }
+
+  if(num_cores != 1) synch_barrier();
+
+  stop();
+
+  if(core_id == 0) {
+    result->errors = matrix_check();
+  }
+}
+
+void matrix_init() {
+  unsigned int i, j;
+
+  // init, copy to TCDM
+  for(i = 0; i < SIZE; i++) {
+    for(j = 0; j < SIZE; j++) {
+      g_mA[i][j] = m_a[i * SIZE + j];
+      g_mB[i][j] = m_b[i * SIZE + j];
+      g_mC[i][j] = 0;
+    }
+  }
+}
+
+unsigned int matrix_check() {
+  unsigned int errors = 0;
+  unsigned int i, j;
+  // check
+  for(i = 0; i < SIZE; i++) {
+    for(j = 0; j < SIZE; j++) {
+      if(g_mC[i][j] != m_exp[i * SIZE + j]) {
+        printf("At index %d, %d\n", i, j, 0, 0);
+        errors++;
+      }
+    }
+  }
+
+  return errors;
+}
diff --git a/astral/parMatrixMul16/parMatrixMul16_stimuli.h b/astral/parMatrixMul16/parMatrixMul16_stimuli.h
new file mode 100644
index 0000000..2797565
--- /dev/null
+++ b/astral/parMatrixMul16/parMatrixMul16_stimuli.h
@@ -0,0 +1,1742 @@
+const short m_a[] = {
+3,
+2,
+-4,
+0,
+0,
+0,
+-5,
+-3,
+2,
+4,
+0,
+2,
+2,
+-3,
+2,
+-1,
+-2,
+-1,
+1,
+4,
+-3,
+-3,
+-1,
+0,
+4,
+-2,
+-1,
+-4,
+-1,
+4,
+1,
+3,
+-2,
+-2,
+-2,
+-1,
+3,
+-5,
+2,
+0,
+1,
+-1,
+-4,
+4,
+0,
+-2,
+4,
+-5,
+3,
+-3,
+-4,
+3,
+-2,
+2,
+-2,
+3,
+3,
+4,
+0,
+-4,
+4,
+0,
+3,
+-5,
+1,
+0,
+0,
+2,
+-3,
+-2,
+-3,
+4,
+-3,
+1,
+-5,
+-5,
+2,
+-3,
+-3,
+-2,
+-4,
+0,
+4,
+1,
+1,
+0,
+4,
+0,
+-3,
+-3,
+-1,
+-2,
+-5,
+4,
+-2,
+1,
+3,
+-1,
+-1,
+-4,
+4,
+-1,
+3,
+-3,
+2,
+2,
+3,
+1,
+2,
+1,
+2,
+4,
+-4,
+-3,
+-2,
+-3,
+4,
+0,
+1,
+0,
+-3,
+0,
+-2,
+-3,
+1,
+3,
+4,
+-1,
+4,
+4,
+3,
+-4,
+-5,
+3,
+2,
+4,
+-4,
+4,
+1,
+-2,
+2,
+-1,
+4,
+2,
+2,
+-5,
+0,
+-2,
+-4,
+2,
+-3,
+-1,
+4,
+0,
+4,
+4,
+1,
+4,
+4,
+-3,
+3,
+4,
+-2,
+4,
+-4,
+2,
+-2,
+-3,
+3,
+1,
+3,
+3,
+0,
+-4,
+1,
+-4,
+4,
+2,
+-2,
+-5,
+-4,
+-4,
+2,
+-1,
+-2,
+0,
+4,
+-3,
+4,
+-5,
+3,
+-1,
+-5,
+-4,
+-2,
+-5,
+4,
+-2,
+3,
+-3,
+4,
+3,
+-4,
+4,
+3,
+-3,
+4,
+-1,
+3,
+-4,
+-5,
+1,
+-3,
+-4,
+2,
+3,
+2,
+-2,
+4,
+3,
+2,
+-4,
+0,
+-2,
+-5,
+-3,
+3,
+1,
+3,
+-5,
+3,
+1,
+-5,
+1,
+-2,
+-4,
+1,
+-5,
+4,
+1,
+0,
+0,
+-4,
+-2,
+2,
+0,
+-3,
+-4,
+-1,
+-5,
+-2,
+-4,
+-2,
+3,
+4,
+-5,
+0,
+2,
+2,
+4,
+0,
+4,
+-5,
+-3,
+-5,
+-5,
+0,
+0,
+-4,
+3,
+1,
+1,
+-5,
+2,
+-5,
+2,
+-4,
+3,
+4,
+-3,
+-4,
+4,
+-2,
+-4,
+4,
+-3,
+-5,
+-4,
+2,
+-2,
+3,
+-2,
+1,
+4,
+1,
+4,
+-2,
+1,
+-4,
+4,
+2,
+-1,
+1,
+1,
+-2,
+-2,
+0,
+-5,
+2,
+4,
+-4,
+1,
+-5,
+0,
+4,
+3,
+-4,
+1,
+-3,
+-4,
+2,
+-3,
+-3,
+1,
+2,
+4,
+-3,
+-2,
+3,
+-3,
+-5,
+4,
+-5,
+-5,
+4,
+4,
+3,
+-1,
+-3,
+-4,
+-1,
+1,
+0,
+1,
+4,
+-1,
+0,
+1,
+-4,
+0,
+-4,
+-3,
+0,
+-4,
+2,
+-2,
+-2,
+-1,
+-2,
+0,
+-5,
+-1,
+-3,
+3,
+4,
+-4,
+4,
+4,
+0,
+-4,
+-3,
+-3,
+4,
+0,
+2,
+3,
+-1,
+2,
+0,
+-4,
+3,
+-1,
+-3,
+-4,
+4,
+0,
+4,
+3,
+0,
+-3,
+-5,
+4,
+-3,
+3,
+4,
+3,
+-1,
+3,
+-3,
+1,
+3,
+4,
+3,
+-1,
+3,
+2,
+4,
+4,
+-2,
+-1,
+2,
+-4,
+-4,
+-1,
+0,
+-1,
+4,
+-4,
+4,
+2,
+0,
+0,
+1,
+-1,
+-5,
+1,
+-5,
+2,
+-2,
+2,
+1,
+3,
+1,
+-3,
+-4,
+-2,
+0,
+4,
+-2,
+0,
+-1,
+-1,
+-3,
+-2,
+4,
+3,
+4,
+2,
+3,
+4,
+-4,
+-2,
+-1,
+-2,
+0,
+-1,
+-1,
+-4,
+-2,
+2,
+-5,
+4,
+-5,
+0,
+4,
+4,
+-2,
+-4,
+1,
+-5,
+-5,
+4,
+-3,
+0,
+-1,
+-5,
+4,
+-4,
+0,
+4,
+4,
+-4,
+-2,
+-4,
+1,
+-4,
+-5,
+4,
+-5,
+-3,
+2,
+-2,
+-1,
+-3,
+0,
+-3,
+-2,
+4,
+-5,
+-5,
+3,
+-5,
+-1,
+3,
+2,
+3,
+0,
+-2,
+-3,
+-4,
+3,
+0,
+3,
+3,
+-1,
+-4,
+0,
+3,
+-5,
+-3,
+4,
+-3,
+2,
+-1,
+0,
+2,
+-1,
+3,
+-1,
+-5,
+3,
+4,
+3,
+3,
+4,
+3,
+1,
+-4,
+1,
+-2,
+4,
+-3,
+2,
+0,
+-2,
+-5,
+-5,
+-2,
+-2,
+0,
+4,
+0,
+-4,
+-1,
+0,
+-5,
+-3,
+4,
+0,
+1,
+-3,
+3,
+-2,
+-2,
+2,
+-1,
+-1,
+-3,
+-1,
+-2,
+-5,
+-5,
+-1,
+-1,
+-4,
+-4,
+};
+
+const short m_b[] = {
+-2,
+3,
+2,
+-5,
+-3,
+2,
+0,
+-5,
+1,
+-3,
+-2,
+-3,
+-3,
+3,
+-4,
+-2,
+-4,
+1,
+1,
+1,
+4,
+4,
+3,
+3,
+-4,
+-3,
+-2,
+-2,
+-3,
+-2,
+-1,
+-5,
+-4,
+1,
+-4,
+-3,
+3,
+-5,
+-5,
+-4,
+-2,
+-2,
+1,
+3,
+3,
+3,
+4,
+0,
+1,
+0,
+4,
+3,
+-3,
+3,
+-1,
+0,
+2,
+-1,
+-1,
+0,
+0,
+3,
+-4,
+4,
+3,
+3,
+-1,
+-4,
+-1,
+-4,
+-2,
+-4,
+-4,
+3,
+0,
+3,
+-4,
+2,
+0,
+0,
+-4,
+-5,
+-4,
+4,
+0,
+-4,
+-2,
+-4,
+3,
+4,
+-4,
+-3,
+-2,
+0,
+-2,
+2,
+-5,
+2,
+1,
+2,
+0,
+0,
+-2,
+-2,
+-1,
+1,
+-1,
+3,
+3,
+-1,
+3,
+-1,
+-2,
+1,
+-3,
+3,
+3,
+3,
+2,
+0,
+1,
+-2,
+1,
+4,
+3,
+-2,
+-5,
+0,
+2,
+0,
+1,
+-3,
+-1,
+3,
+-3,
+-2,
+3,
+4,
+-2,
+-1,
+-5,
+1,
+-1,
+2,
+-3,
+-4,
+0,
+-1,
+0,
+2,
+3,
+4,
+-4,
+-1,
+-1,
+2,
+-1,
+-5,
+1,
+-3,
+0,
+-4,
+0,
+-1,
+2,
+-5,
+-1,
+1,
+-2,
+-5,
+-3,
+-3,
+2,
+2,
+-4,
+2,
+-5,
+-4,
+4,
+4,
+-1,
+1,
+2,
+-1,
+-3,
+0,
+2,
+4,
+-3,
+-2,
+-1,
+4,
+-5,
+4,
+0,
+-1,
+-1,
+-4,
+-3,
+-3,
+-1,
+-2,
+-1,
+-2,
+-5,
+2,
+2,
+-3,
+1,
+-1,
+-2,
+0,
+0,
+1,
+0,
+4,
+-4,
+-4,
+-4,
+-5,
+4,
+-4,
+4,
+-3,
+4,
+-5,
+-2,
+3,
+-2,
+-4,
+1,
+2,
+-5,
+-2,
+-5,
+-3,
+2,
+-1,
+3,
+-4,
+-4,
+4,
+-3,
+-5,
+0,
+-3,
+2,
+2,
+3,
+-3,
+-3,
+-2,
+-5,
+4,
+1,
+3,
+-1,
+-4,
+3,
+1,
+-2,
+1,
+3,
+4,
+1,
+-2,
+0,
+0,
+-3,
+-1,
+1,
+-1,
+0,
+-5,
+-2,
+-1,
+1,
+-2,
+4,
+-3,
+2,
+1,
+0,
+0,
+3,
+-4,
+2,
+1,
+0,
+3,
+-5,
+-1,
+4,
+3,
+4,
+-2,
+-1,
+-3,
+-4,
+3,
+3,
+-4,
+4,
+-5,
+1,
+-1,
+0,
+3,
+4,
+3,
+-3,
+-5,
+-3,
+-2,
+-4,
+1,
+-2,
+-3,
+-2,
+1,
+2,
+4,
+4,
+0,
+2,
+4,
+0,
+-1,
+1,
+-4,
+-1,
+-3,
+-2,
+-5,
+3,
+-3,
+-1,
+0,
+-3,
+-3,
+-4,
+3,
+3,
+-5,
+-2,
+-1,
+3,
+1,
+1,
+1,
+3,
+1,
+1,
+-5,
+-2,
+2,
+4,
+3,
+-3,
+-5,
+3,
+-4,
+0,
+2,
+-4,
+-5,
+-3,
+-3,
+-3,
+2,
+-1,
+-3,
+1,
+-3,
+-1,
+-5,
+2,
+1,
+1,
+-3,
+2,
+4,
+1,
+-5,
+3,
+0,
+-5,
+-4,
+2,
+-5,
+3,
+-1,
+0,
+2,
+-3,
+-1,
+-2,
+-5,
+-4,
+0,
+-5,
+-3,
+3,
+1,
+0,
+-1,
+-4,
+1,
+-4,
+-5,
+-1,
+3,
+3,
+-3,
+-1,
+4,
+-1,
+-5,
+-4,
+-1,
+0,
+1,
+1,
+4,
+-3,
+-2,
+4,
+1,
+-4,
+-1,
+-3,
+2,
+-2,
+4,
+2,
+-1,
+2,
+-4,
+-3,
+4,
+-1,
+1,
+-4,
+0,
+1,
+-4,
+4,
+-4,
+-5,
+-5,
+-3,
+2,
+2,
+-1,
+0,
+-2,
+-5,
+1,
+-1,
+-5,
+2,
+-4,
+-5,
+-3,
+1,
+-4,
+0,
+1,
+-5,
+0,
+-2,
+1,
+-4,
+-5,
+3,
+0,
+1,
+2,
+4,
+-1,
+-4,
+0,
+-4,
+0,
+-5,
+3,
+-5,
+-5,
+2,
+2,
+3,
+-3,
+4,
+4,
+2,
+3,
+-2,
+4,
+3,
+4,
+1,
+1,
+0,
+2,
+2,
+4,
+-1,
+-5,
+-5,
+3,
+-5,
+-2,
+-4,
+-4,
+0,
+4,
+3,
+0,
+4,
+-3,
+4,
+0,
+-1,
+3,
+2,
+-4,
+2,
+-1,
+-3,
+-5,
+-3,
+3,
+-4,
+1,
+-3,
+3,
+-4,
+0,
+3,
+-2,
+-1,
+2,
+-4,
+2,
+4,
+0,
+2,
+-2,
+-5,
+0,
+-1,
+-5,
+4,
+-4,
+3,
+-1,
+4,
+1,
+4,
+0,
+-1,
+1,
+4,
+3,
+3,
+1,
+2,
+-3,
+-2,
+-5,
+-2,
+1,
+-5,
+-4,
+-5,
+-2,
+};
+
+const short m_exp[] = {
+-32,
+48,
+2,
+-56,
+46,
+-52,
+39,
+-77,
+41,
+-50,
+-50,
+-14,
+-45,
+-31,
+3,
+-22,
+-52,
+-31,
+-32,
+8,
+61,
+97,
+49,
+-12,
+12,
+-25,
+-26,
+-15,
+15,
+45,
+-29,
+-27,
+-7,
+-23,
+30,
+-36,
+-56,
+27,
+-59,
+35,
+6,
+44,
+72,
+23,
+17,
+86,
+-8,
+5,
+-25,
+41,
+-30,
+-78,
+120,
+-1,
+69,
+-20,
+24,
+-66,
+6,
+68,
+-59,
+2,
+27,
+-21,
+-71,
+-61,
+-49,
+21,
+-24,
+27,
+-47,
+29,
+71,
+37,
+-9,
+-58,
+13,
+13,
+73,
+-9,
+7,
+28,
+23,
+35,
+46,
+-8,
+86,
+-7,
+25,
+-65,
+-18,
+43,
+58,
+22,
+44,
+-51,
+-33,
+10,
+-18,
+-17,
+-62,
+24,
+18,
+-68,
+23,
+25,
+23,
+-32,
+23,
+37,
+55,
+-16,
+24,
+4,
+-16,
+9,
+77,
+44,
+48,
+-36,
+-81,
+-100,
+-18,
+36,
+-3,
+-72,
+-43,
+38,
+-2,
+34,
+53,
+-42,
+-18,
+56,
+19,
+34,
+3,
+-58,
+-16,
+-34,
+-71,
+-12,
+12,
+-22,
+-9,
+57,
+-46,
+-70,
+25,
+-62,
+10,
+-5,
+68,
+-86,
+-15,
+-19,
+-100,
+78,
+23,
+47,
+5,
+-64,
+31,
+-16,
+-9,
+-22,
+11,
+30,
+-74,
+31,
+50,
+38,
+-59,
+24,
+12,
+-18,
+4,
+33,
+-24,
+-44,
+-19,
+2,
+-47,
+48,
+-30,
+28,
+-38,
+-29,
+20,
+52,
+-21,
+-5,
+49,
+-41,
+-17,
+-2,
+68,
+-2,
+122,
+-29,
+13,
+23,
+46,
+40,
+-12,
+-58,
+59,
+2,
+15,
+-49,
+-38,
+-15,
+46,
+-1,
+-54,
+-73,
+10,
+-1,
+43,
+80,
+-55,
+119,
+82,
+22,
+-12,
+-14,
+-12,
+-3,
+36,
+27,
+-3,
+50,
+59,
+61,
+13,
+-14,
+5,
+51,
+-54,
+-45,
+-9,
+109,
+10,
+-28,
+1,
+0,
+-50,
+-23,
+27,
+49,
+16,
+3,
+23,
+43,
+31,
+29,
+-27,
+-45,
+9,
+61,
+42,
+34,
+30,
+26,
+38,
+-84,
+5,
+98,
+13,
+55,
+15,
+56,
+63,
+22,
+80,
+10,
+52,
+52,
+49,
+101,
+37,
+71,
+-20,
+-38,
+-36,
+-72,
+-40,
+12,
+77,
+-19,
+32,
+9,
+-15,
+85,
+-13,
+-33,
+-8,
+-20,
+33,
+36,
+76,
+6,
+67,
+-51,
+36,
+76,
+-48,
+30,
+42,
+-45,
+-22,
+-9,
+65,
+-29,
+18,
+89,
+34,
+-25,
+41,
+30,
+26,
+49,
+34,
+-6,
+-28,
+10,
+-106,
+6,
+86,
+5,
+-3,
+-105,
+-103,
+-45,
+-55,
+-60,
+29,
+41,
+54,
+-29,
+36,
+-30,
+-8,
+25,
+36,
+40,
+22,
+-35,
+-50,
+33,
+3,
+-19,
+-30,
+3,
+-11,
+9,
+-4,
+-18,
+39,
+92,
+-25,
+0,
+-40,
+64,
+-18,
+58,
+-2,
+27,
+-89,
+32,
+-2,
+56,
+79,
+-67,
+87,
+-41,
+20,
+29,
+4,
+51,
+45,
+26,
+-54,
+8,
+-37,
+-37,
+-39,
+-35,
+-36,
+-23,
+-30,
+11,
+1,
+-63,
+-21,
+5,
+13,
+35,
+-15,
+27,
+-59,
+-10,
+29,
+4,
+-58,
+-56,
+3,
+-10,
+27,
+0,
+29,
+-4,
+-17,
+28,
+2,
+17,
+9,
+-27,
+53,
+-19,
+-36,
+104,
+25,
+38,
+84,
+-29,
+42,
+-10,
+-47,
+-36,
+1,
+10,
+43,
+-46,
+-57,
+38,
+44,
+-9,
+26,
+-15,
+10,
+31,
+34,
+-18,
+17,
+24,
+-3,
+5,
+11,
+4,
+23,
+5,
+-19,
+51,
+-59,
+-29,
+62,
+-27,
+35,
+81,
+65,
+-10,
+-1,
+-1,
+45,
+20,
+26,
+-34,
+-7,
+35,
+-7,
+2,
+19,
+115,
+-32,
+-40,
+-92,
+47,
+-65,
+23,
+53,
+-31,
+22,
+20,
+45,
+91,
+64,
+29,
+49,
+30,
+74,
+-15,
+22,
+120,
+16,
+122,
+-51,
+51,
+78,
+60,
+-53,
+40,
+-49,
+73,
+-2,
+14,
+-17,
+-86,
+35,
+-25,
+36,
+2,
+5,
+38,
+-9,
+16,
+-12,
+-41,
+-24,
+-31,
+-72,
+-68,
+-9,
+5,
+11,
+-63,
+15,
+12,
+21,
+22,
+19,
+62,
+-28,
+122,
+12,
+28,
+-59,
+-19,
+49,
+57,
+-12,
+29,
+-1,
+-23,
+126,
+75,
+-29,
+-21,
+-37,
+8,
+-70,
+-8,
+91,
+-34,
+-10,
+31,
+78,
+44,
+59,
+23,
+66,
+42,
+38,
+6,
+19,
+10,
+1,
+57,
+38,
+71,
+-10,
+-12,
+-7,
+-13,
+-15,
+6,
+};
+
+#define SIZE 24
+__attribute__ ((section(".heapsram"))) short g_mA[SIZE][SIZE];
+__attribute__ ((section(".heapsram"))) short g_mB[SIZE][SIZE];
+__attribute__ ((section(".heapsram"))) short g_mC[SIZE][SIZE];
+__attribute__ ((section(".heapsram"))) short g_mB_tmp[SIZE][SIZE];
diff --git a/astral/parMatrixMul32/Makefile b/astral/parMatrixMul32/Makefile
new file mode 100755
index 0000000..bd55e15
--- /dev/null
+++ b/astral/parMatrixMul32/Makefile
@@ -0,0 +1,8 @@
+PULP_APP = test
+PULP_APP_SRCS = matrixMul.c
+
+PULP_CFLAGS = -O3
+
+include $(PULP_SDK_HOME)/install/rules/pulp.mk
+
+#pulp-bench-reg --name=parMatrixMul32.cycles --module=pulp_rtl_testset --pipeline=$(PIPELINE) --artefact=pulp_rtl_testset --cmd="make run -f Makefile.sdk" --probe-regexp='matrixMul -> success, nr. of errors: 0, execution time: (\d+)' --params="platform($(platformName)),compiler($(OR1K_TOOLCHAIN_TYPE)),nbCores(4),elemSize(32)" --probe-regexp='matrixMulTranspose -> success, nr. of errors: 0, execution time: (\d+)' --params="platform($(platformName)),compiler($(OR1K_TOOLCHAIN_TYPE)),nbCores(4),elemSize(32),transposed"
diff --git a/astral/parMatrixMul32/gen_stimuli.py b/astral/parMatrixMul32/gen_stimuli.py
new file mode 100755
index 0000000..32926e9
--- /dev/null
+++ b/astral/parMatrixMul32/gen_stimuli.py
@@ -0,0 +1,54 @@
+#!/usr/bin/env python
+
+import sys
+import random
+
+
+def write_arr(f, name, arr):
+    f.write('const int %s[] = {\n' % name)
+    for v in arr:
+        f.write('%d,\n' % (v))
+    f.write('};\n\n')
+    return
+
+################################################################################
+f = open('parMatrixMul32_stimuli.h', 'w')
+
+
+SIZE = 24
+RANGE = int(2**15/SIZE)
+
+m_a   = []
+m_b   = []
+m_exp = []
+
+for i in range(0,SIZE):
+    for j in range(0,SIZE):
+        a = random.randint(-RANGE, RANGE-1)
+        b = random.randint(-RANGE, RANGE-1)
+
+        m_a.append(a)
+        m_b.append(b)
+
+for i in range(0,SIZE):
+    for j in range(0,SIZE):
+        r = 0
+
+        for k in range (0,SIZE):
+            r = r + m_a[i * SIZE + k] * m_b[k * SIZE + j]
+
+        m_exp.append(r)
+
+
+write_arr(f, 'm_a',   m_a)
+write_arr(f, 'm_b',   m_b)
+write_arr(f, 'm_exp', m_exp)
+
+f.write('#define SIZE %d\n' % SIZE)
+
+
+f.write('__attribute__ ((section(".heapsram"))) int g_mA[SIZE][SIZE];\n')
+f.write('__attribute__ ((section(".heapsram"))) int g_mB[SIZE][SIZE];\n')
+f.write('__attribute__ ((section(".heapsram"))) int g_mC[SIZE][SIZE];\n')
+f.write('__attribute__ ((section(".heapsram"))) int g_mB_tmp[SIZE][SIZE];\n')
+
diff --git a/astral/parMatrixMul32/matrixMul.c b/astral/parMatrixMul32/matrixMul.c
new file mode 100644
index 0000000..990d411
--- /dev/null
+++ b/astral/parMatrixMul32/matrixMul.c
@@ -0,0 +1,177 @@
+/*
+ * Copyright (C) 2018 ETH Zurich and University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+/* 
+ * Mantainer: Luca Valente, luca.valente2@unibo.it
+ */
+
+#include "pulp.h"
+
+#include "parMatrixMul32_stimuli.h"
+
+void check_matrix_mul(testresult_t *result, void (*start)(), void (*stop)());
+void check_matrix_mul_transpose(testresult_t *result, void (*start)(), void (*stop)());
+
+testcase_t testcases[] = {
+  { .name = "matrixMul",          .test = check_matrix_mul           },
+  { .name = "matrixMulTranspose", .test = check_matrix_mul_transpose },
+  {0, 0}
+};
+
+unsigned int num_cores;
+
+int main()
+{
+  if (rt_cluster_id() != 0)
+    return bench_cluster_forward(0);
+
+  num_cores = get_core_num();
+
+  if(rt_core_id() < num_cores) {
+    run_suite(testcases);
+  }
+
+  synch_barrier();
+
+  return 0;
+}
+
+void matrix_init();
+unsigned int matrix_check();
+
+void check_matrix_mul(testresult_t *result, void (*start)(), void (*stop)()) {
+  int core_id;
+  unsigned int i, j, k;
+  unsigned int chunk;
+  unsigned int lb, ub;
+
+  core_id = get_core_id();
+
+  // number of rows each core has to multiply
+  chunk = SIZE / num_cores;
+  // lower bound
+  lb = core_id * chunk;
+  // upper bound
+  ub = lb + chunk;
+
+  if(core_id == 0) {
+    matrix_init();
+  }
+
+  if(num_cores != 1) synch_barrier();
+
+  // start benchmark
+  start();
+
+  for(i = lb; i < ub; i++) {
+    for(j = 0; j < SIZE; j++) {
+      g_mC[i][j] = 0;
+  
+      for(k = 0; k < SIZE; k++) {
+        g_mC[i][j] += g_mA[i][k] * g_mB[k][j];
+      }
+    }
+  }
+
+  if(num_cores != 1) synch_barrier();
+
+  stop();
+
+  if(core_id == 0) {
+    result->errors = matrix_check();
+  }
+}
+
+void check_matrix_mul_transpose(testresult_t *result, void (*start)(), void (*stop)()) {
+  int core_id;
+  unsigned int i, j, k;
+  unsigned int chunk;
+  unsigned int lb, ub;
+
+  core_id = get_core_id();
+
+  // number of rows each core has to multiply
+  chunk = SIZE / num_cores;
+  // lower bound
+  lb = core_id * chunk;
+  // upper bound
+  ub = lb + chunk;
+
+  if(core_id == 0) {
+    matrix_init();
+  }
+
+  if(num_cores != 1) synch_barrier();
+
+  // start benchmark
+  start();
+
+  // transpose array before using it
+  for(i = lb; i < ub; i++) {
+    for(j = 0; j < SIZE; j++) {
+      g_mB_tmp[i][j] = g_mB[j][i];
+    }
+  }
+
+  if(num_cores != 1) synch_barrier();
+
+  for(i = lb; i < ub; i++) {
+    for(j = 0; j < SIZE; j++) {
+      g_mC[i][j] = 0;
+
+      for(k = 0; k < SIZE; k++) {
+        g_mC[i][j] += g_mA[i][k] * g_mB_tmp[j][k];
+      }
+    }
+  }
+
+  if(num_cores != 1) synch_barrier();
+
+  stop();
+
+  if(core_id == 0) {
+    result->errors = matrix_check();
+  }
+}
+
+void matrix_init() {
+  unsigned int i, j;
+
+  // init, copy to TCDM
+  for(i = 0; i < SIZE; i++) {
+    for(j = 0; j < SIZE; j++) {
+      g_mA[i][j] = m_a[i * SIZE + j];
+      g_mB[i][j] = m_b[i * SIZE + j];
+      g_mC[i][j] = 0;
+    }
+  }
+}
+
+unsigned int matrix_check() {
+  unsigned int errors = 0;
+  unsigned int i, j;
+  // check
+  for(i = 0; i < SIZE; i++) {
+    for(j = 0; j < SIZE; j++) {
+      if(g_mC[i][j] != m_exp[i * SIZE + j]) {
+        printf("At index %d, %d\n", i, j, 0, 0);
+        errors++;
+      }
+    }
+  }
+
+  return errors;
+}
diff --git a/astral/parMatrixMul32/parMatrixMul32_stimuli.h b/astral/parMatrixMul32/parMatrixMul32_stimuli.h
new file mode 100644
index 0000000..d5c1af0
--- /dev/null
+++ b/astral/parMatrixMul32/parMatrixMul32_stimuli.h
@@ -0,0 +1,1742 @@
+const int m_a[] = {
+-1344,
+-778,
+-350,
+1240,
+950,
+940,
+1262,
+285,
+738,
+-37,
+257,
+704,
+87,
+-1135,
+1175,
+960,
+984,
+611,
+-1170,
+1243,
+-729,
+-1235,
+-1229,
+8,
+52,
+1156,
+-804,
+-129,
+-112,
+-307,
+1045,
+-954,
+-944,
+477,
+1104,
+-600,
+-31,
+1364,
+-950,
+-153,
+967,
+-446,
+235,
+-197,
+763,
+-660,
+-1289,
+116,
+-458,
+994,
+1086,
+156,
+-549,
+102,
+-532,
+943,
+299,
+340,
+516,
+1117,
+515,
+1335,
+708,
+-306,
+-589,
+517,
+655,
+-279,
+-595,
+-255,
+-1104,
+-16,
+507,
+199,
+-297,
+-905,
+-1084,
+982,
+20,
+-1364,
+407,
+333,
+-798,
+-711,
+1301,
+175,
+490,
+-165,
+-1097,
+1251,
+1203,
+-884,
+419,
+-1262,
+-950,
+-1200,
+759,
+-205,
+-1365,
+-870,
+109,
+-1135,
+36,
+796,
+-1233,
+-1117,
+-826,
+241,
+588,
+-513,
+-353,
+791,
+1071,
+-1073,
+220,
+-1094,
+1096,
+-723,
+280,
+-505,
+151,
+399,
+319,
+-1120,
+-213,
+-966,
+679,
+497,
+-290,
+-300,
+-290,
+-599,
+243,
+-752,
+604,
+1196,
+-715,
+-177,
+-329,
+1337,
+596,
+1027,
+509,
+-301,
+-1211,
+-936,
+-281,
+446,
+-356,
+841,
+-1123,
+-1343,
+-140,
+-1300,
+-828,
+-237,
+1206,
+1274,
+-1092,
+-922,
+913,
+1201,
+-422,
+782,
+-230,
+633,
+-1055,
+-1160,
+329,
+1255,
+1038,
+770,
+123,
+934,
+670,
+-492,
+-24,
+-779,
+-1129,
+47,
+555,
+1214,
+-232,
+-716,
+-322,
+-126,
+178,
+827,
+710,
+-1057,
+-313,
+996,
+1149,
+-532,
+570,
+1171,
+899,
+-313,
+-790,
+1071,
+154,
+-303,
+492,
+-918,
+-1139,
+292,
+129,
+1347,
+-309,
+751,
+1262,
+142,
+-1062,
+-1305,
+250,
+657,
+238,
+-141,
+1308,
+-37,
+-514,
+-591,
+-611,
+852,
+-653,
+-640,
+91,
+254,
+-1145,
+-1263,
+-838,
+-10,
+266,
+-444,
+1129,
+762,
+-713,
+-326,
+-88,
+1063,
+-442,
+-177,
+365,
+-740,
+-1219,
+1085,
+783,
+-725,
+-1112,
+426,
+660,
+6,
+-440,
+513,
+687,
+1078,
+212,
+-434,
+-953,
+1337,
+160,
+622,
+-950,
+-943,
+288,
+-136,
+-1103,
+-223,
+1271,
+211,
+251,
+-271,
+-26,
+704,
+1177,
+544,
+699,
+-885,
+-864,
+-1280,
+877,
+-461,
+995,
+-623,
+-121,
+-146,
+-484,
+-225,
+-978,
+163,
+-278,
+-502,
+-505,
+-567,
+-771,
+1279,
+699,
+-1337,
+544,
+1145,
+1271,
+640,
+277,
+-164,
+458,
+-1280,
+-602,
+-2,
+1136,
+1203,
+-699,
+-195,
+659,
+-472,
+1230,
+1151,
+-97,
+-77,
+-772,
+-381,
+-295,
+636,
+-1341,
+-445,
+-806,
+531,
+-1186,
+-1313,
+-274,
+835,
+-446,
+558,
+-1307,
+-235,
+43,
+-254,
+-109,
+911,
+-1189,
+559,
+-854,
+-218,
+149,
+580,
+1158,
+-14,
+181,
+-1120,
+-947,
+-542,
+1142,
+631,
+-893,
+-614,
+-257,
+-365,
+-951,
+1,
+-762,
+268,
+382,
+-131,
+808,
+-234,
+839,
+346,
+-733,
+1251,
+496,
+-566,
+-751,
+581,
+-1292,
+1068,
+-932,
+-855,
+1336,
+-280,
+523,
+1294,
+-1251,
+1284,
+-1276,
+87,
+1264,
+-274,
+-922,
+-289,
+-458,
+-117,
+196,
+-79,
+-707,
+1233,
+-385,
+-620,
+-617,
+703,
+-995,
+-374,
+660,
+145,
+821,
+1289,
+582,
+-201,
+447,
+116,
+759,
+-615,
+834,
+268,
+-1114,
+-1016,
+-227,
+-589,
+-910,
+-244,
+-660,
+764,
+219,
+1165,
+506,
+-673,
+799,
+-1355,
+-872,
+491,
+689,
+176,
+-285,
+1151,
+1080,
+-319,
+286,
+833,
+217,
+-621,
+478,
+539,
+-109,
+-1273,
+-564,
+-240,
+504,
+518,
+256,
+-124,
+74,
+949,
+-912,
+-1341,
+965,
+-774,
+634,
+1009,
+1304,
+200,
+-1041,
+-1262,
+-865,
+-1065,
+-635,
+-357,
+-928,
+806,
+1148,
+-411,
+56,
+686,
+-644,
+1241,
+-430,
+297,
+127,
+457,
+-1313,
+741,
+861,
+220,
+-540,
+772,
+265,
+1066,
+679,
+177,
+-734,
+29,
+-149,
+181,
+-1042,
+-1139,
+271,
+-326,
+-29,
+1298,
+643,
+-890,
+-136,
+-1015,
+-565,
+-964,
+894,
+-312,
+698,
+159,
+-222,
+-1322,
+578,
+945,
+1124,
+1278,
+54,
+-389,
+1101,
+362,
+-543,
+380,
+959,
+-399,
+-1105,
+1308,
+338,
+-198,
+-1111,
+-278,
+-752,
+668,
+1156,
+-1226,
+579,
+184,
+-1084,
+-917,
+-498,
+-466,
+316,
+-788,
+-718,
+468,
+367,
+-1333,
+-1146,
+828,
+1329,
+311,
+-1346,
+54,
+-976,
+854,
+-658,
+-198,
+-979,
+156,
+385,
+-659,
+1326,
+1351,
+-1173,
+-648,
+720,
+-40,
+313,
+729,
+-416,
+351,
+452,
+-413,
+-4,
+-1113,
+-612,
+-28,
+-721,
+400,
+1072,
+-1010,
+};
+
+const int m_b[] = {
+-1316,
+319,
+963,
+-608,
+519,
+-783,
+-676,
+181,
+172,
+203,
+-1351,
+-935,
+-12,
+758,
+-746,
+1226,
+127,
+-1346,
+1251,
+-377,
+889,
+-23,
+-417,
+-122,
+680,
+1363,
+729,
+-907,
+-1263,
+-431,
+363,
+1355,
+-566,
+-517,
+-1186,
+1318,
+-1104,
+-1245,
+950,
+687,
+252,
+-270,
+1081,
+-1290,
+656,
+8,
+60,
+1171,
+915,
+-500,
+678,
+-953,
+307,
+-35,
+-1334,
+-888,
+598,
+1160,
+722,
+850,
+-268,
+988,
+635,
+-340,
+252,
+1208,
+420,
+82,
+1283,
+-319,
+-666,
+172,
+583,
+174,
+471,
+-1063,
+452,
+-191,
+-1188,
+116,
+-927,
+1086,
+119,
+-245,
+-717,
+-657,
+417,
+319,
+1133,
+1338,
+141,
+-546,
+567,
+-1089,
+-191,
+-1138,
+-201,
+-1286,
+-820,
+-1356,
+1177,
+-317,
+191,
+67,
+164,
+-306,
+-1015,
+1147,
+-482,
+1229,
+-259,
+-207,
+1309,
+847,
+-399,
+-1005,
+-995,
+140,
+-567,
+-1220,
+-427,
+180,
+-571,
+997,
+-783,
+-316,
+-1360,
+736,
+75,
+-1251,
+-307,
+-902,
+1181,
+1057,
+-141,
+-1098,
+776,
+1096,
+-923,
+914,
+1049,
+-28,
+-742,
+-804,
+-467,
+567,
+329,
+-309,
+-161,
+-157,
+-430,
+-639,
+1138,
+-165,
+292,
+-20,
+777,
+-715,
+60,
+-1359,
+35,
+307,
+-1092,
+271,
+548,
+822,
+-50,
+-475,
+-103,
+784,
+537,
+152,
+-517,
+-1097,
+117,
+-619,
+538,
+941,
+172,
+-223,
+1161,
+-1004,
+-1145,
+-455,
+255,
+363,
+859,
+403,
+-861,
+-657,
+-537,
+-1084,
+-1042,
+541,
+1283,
+-356,
+1298,
+-1254,
+-303,
+203,
+104,
+1123,
+-72,
+-171,
+-1122,
+-533,
+440,
+275,
+613,
+-846,
+-189,
+884,
+704,
+-570,
+-440,
+-1157,
+-200,
+-80,
+616,
+799,
+757,
+-264,
+-1256,
+-690,
+152,
+184,
+-810,
+-221,
+-821,
+-243,
+508,
+-709,
+574,
+-693,
+315,
+-952,
+952,
+697,
+875,
+-480,
+-691,
+422,
+-413,
+-1199,
+441,
+-751,
+821,
+1303,
+-410,
+416,
+566,
+-131,
+-551,
+46,
+978,
+-228,
+1117,
+-251,
+-537,
+874,
+-882,
+260,
+-213,
+248,
+-1296,
+1343,
+-626,
+-812,
+629,
+-601,
+-378,
+-1314,
+-889,
+774,
+-307,
+692,
+-1125,
+-692,
+923,
+947,
+1158,
+-939,
+1284,
+35,
+1299,
+369,
+-8,
+43,
+768,
+524,
+137,
+659,
+285,
+-1315,
+-457,
+871,
+-768,
+1107,
+-695,
+488,
+-527,
+-161,
+414,
+-526,
+-1164,
+1059,
+-1108,
+560,
+-622,
+898,
+-50,
+-286,
+-170,
+513,
+952,
+433,
+237,
+584,
+-665,
+-960,
+585,
+-434,
+1223,
+-130,
+1035,
+430,
+202,
+1312,
+1152,
+1059,
+-1082,
+-1295,
+805,
+-18,
+613,
+-94,
+557,
+548,
+1354,
+116,
+289,
+-1358,
+-1234,
+1237,
+451,
+820,
+-102,
+974,
+832,
+-1019,
+914,
+-512,
+-267,
+1329,
+-910,
+-1341,
+862,
+-381,
+-23,
+-658,
+40,
+-71,
+-782,
+1240,
+-956,
+1241,
+-291,
+-884,
+1250,
+699,
+834,
+190,
+960,
+-1260,
+177,
+464,
+155,
+-1105,
+768,
+424,
+621,
+740,
+-1357,
+1186,
+-594,
+1329,
+829,
+126,
+1101,
+1146,
+-95,
+605,
+-673,
+1334,
+440,
+-10,
+12,
+-745,
+20,
+19,
+-793,
+999,
+1083,
+487,
+-657,
+-356,
+654,
+-326,
+-250,
+-718,
+-947,
+-235,
+558,
+974,
+-981,
+637,
+-861,
+-768,
+1045,
+-583,
+-910,
+128,
+734,
+896,
+-1156,
+223,
+284,
+272,
+634,
+-473,
+363,
+359,
+-1185,
+14,
+-33,
+-1122,
+140,
+900,
+439,
+-944,
+-770,
+663,
+865,
+1056,
+-238,
+86,
+-1294,
+-44,
+-603,
+602,
+20,
+397,
+-423,
+-703,
+-209,
+-906,
+-1236,
+945,
+-737,
+578,
+904,
+645,
+1225,
+-877,
+-425,
+-493,
+-1326,
+424,
+965,
+1300,
+-1210,
+823,
+1345,
+626,
+-427,
+592,
+-869,
+-1055,
+-938,
+-427,
+1066,
+472,
+-1055,
+48,
+-1200,
+-349,
+313,
+-1227,
+-228,
+783,
+839,
+187,
+1021,
+-1355,
+1284,
+68,
+-1321,
+-997,
+1286,
+-887,
+772,
+-156,
+-105,
+1329,
+1141,
+-377,
+-881,
+-341,
+1316,
+-391,
+-1249,
+-205,
+53,
+-266,
+-540,
+-289,
+-1011,
+602,
+-1032,
+-1097,
+-202,
+-467,
+-1047,
+-867,
+-340,
+-109,
+-496,
+967,
+1147,
+108,
+384,
+-12,
+1216,
+137,
+1318,
+151,
+219,
+-543,
+391,
+668,
+-1348,
+-1244,
+-810,
+-676,
+321,
+-1258,
+1343,
+1214,
+791,
+35,
+1219,
+1278,
+1037,
+-1282,
+661,
+585,
+921,
+-880,
+-989,
+-1192,
+-207,
+273,
+-382,
+690,
+165,
+271,
+-212,
+739,
+-343,
+-42,
+226,
+40,
+859,
+-153,
+622,
+-1059,
+};
+
+const int m_exp[] = {
+-4942391,
+2289133,
+-1363225,
+1978230,
+1580032,
+-625813,
+-3230128,
+2236653,
+19494,
+3242695,
+-1080745,
+-34154,
+4086860,
+-1370876,
+3997221,
+-1812380,
+4705498,
+7690207,
+-4068140,
+3595067,
+-1103308,
+-939857,
+-4249710,
+-8650816,
+-2013119,
+2933624,
+235162,
+-453807,
+-4447391,
+3527041,
+2046492,
+411956,
+-994117,
+-1411344,
+1333704,
+-519761,
+3026373,
+-564969,
+3749147,
+2447173,
+-557628,
+1138674,
+-1426096,
+-4033488,
+-1829685,
+2815607,
+2382958,
+1714081,
+-1470484,
+3379876,
+3660759,
+-2439960,
+-1180478,
+-3300785,
+-5104533,
+-309753,
+-1667400,
+3258850,
+1805449,
+2481948,
+-944985,
+-363123,
+4227063,
+3022289,
+2763211,
+5114077,
+-1534394,
+-2957168,
+3401637,
+-1195822,
+-747480,
+-2915318,
+-2505013,
+-174927,
+276733,
+2899369,
+-6702856,
+923396,
+-2741169,
+4270685,
+-1020657,
+-2562887,
+2074098,
+-2382784,
+1366504,
+691209,
+4127820,
+400356,
+-7415505,
+823772,
+-3848400,
+-158560,
+3759990,
+2298445,
+323394,
+873625,
+474364,
+2617120,
+-1382444,
+1735284,
+-5799715,
+1915577,
+7016057,
+-1212904,
+919286,
+2949768,
+1228832,
+572192,
+4145710,
+-6809520,
+-2199597,
+677764,
+-7169579,
+-4904277,
+6902014,
+231123,
+-4797299,
+3093608,
+989455,
+4324476,
+3121268,
+810907,
+-2457323,
+2299211,
+-1625774,
+-141013,
+3343022,
+-2044657,
+4089375,
+-291323,
+-1950307,
+2480885,
+2846731,
+-2139146,
+-2718414,
+-1997531,
+-2399245,
+-4060224,
+423228,
+-205276,
+-1602384,
+910872,
+-3535114,
+6008729,
+-559984,
+-301205,
+-5407307,
+2981269,
+1079061,
+6602535,
+-857708,
+2756391,
+-5304566,
+-3769267,
+2620777,
+-4409088,
+502077,
+7568647,
+-37918,
+2315061,
+-2540065,
+8296540,
+-7465282,
+1553910,
+-4736227,
+-2139045,
+95614,
+342546,
+1309722,
+1777391,
+2981296,
+-736899,
+-2572111,
+-972463,
+-2793724,
+-2893912,
+-1230264,
+-871649,
+-1439985,
+3074445,
+3339004,
+2423842,
+1751086,
+4064832,
+1550243,
+6225792,
+-503750,
+-2567772,
+5671219,
+-2054796,
+-551487,
+-2787790,
+-3835027,
+-4272806,
+2725813,
+-2982521,
+1803437,
+3024675,
+-201092,
+-1626608,
+1548043,
+2303810,
+3032912,
+398283,
+1704371,
+1860306,
+-4101665,
+5187913,
+4233418,
+2054883,
+3603470,
+1935132,
+-17548,
+-4362444,
+-2806918,
+-5651039,
+-1853372,
+-1707208,
+-153048,
+-2791834,
+402265,
+2815962,
+3391662,
+-4833520,
+-1190520,
+2302448,
+80738,
+2089586,
+174096,
+2837490,
+-5514606,
+2138871,
+336249,
+-378675,
+5833977,
+4367000,
+-2445147,
+-3652299,
+-1794451,
+-1471577,
+-1263012,
+3719274,
+-3404819,
+5765304,
+-4256415,
+3558206,
+-1884441,
+-475244,
+3659623,
+-2914867,
+689238,
+-2576754,
+7739914,
+1823902,
+2077002,
+-2365242,
+2023481,
+1663749,
+-4973435,
+-694558,
+1118078,
+2260786,
+-3256285,
+4596746,
+-5421599,
+594942,
+-1730692,
+-4626077,
+3077882,
+-2232009,
+2672161,
+3135747,
+-4602601,
+859784,
+-3530668,
+21600,
+-4690786,
+2023164,
+-496745,
+-2728919,
+281474,
+-108745,
+-809613,
+1445687,
+-5781458,
+2097169,
+1594266,
+-4504019,
+2460482,
+6259537,
+-700848,
+413263,
+-1212884,
+-5695130,
+2094147,
+-750529,
+-1379008,
+6029072,
+722889,
+1719449,
+1682336,
+-4219755,
+1971162,
+66575,
+-1195119,
+141466,
+-1083536,
+-718558,
+-4039954,
+-168429,
+-2026861,
+2025800,
+-761083,
+-4194692,
+2013337,
+-1156936,
+3823019,
+4081732,
+-3145845,
+-1733615,
+-1371947,
+-3811245,
+-1584663,
+-3547009,
+-3267886,
+8255291,
+-3232160,
+3404636,
+3248369,
+3233853,
+671601,
+-1009897,
+1821121,
+-3517645,
+2005444,
+-2768741,
+115998,
+533867,
+4717709,
+1315923,
+-3510545,
+-3539595,
+-538461,
+4529529,
+2792584,
+-107486,
+-1840413,
+-1474849,
+1579605,
+-4197602,
+-34825,
+-462678,
+1294881,
+-1730927,
+-2549709,
+-1531672,
+-271859,
+-1181904,
+-1680154,
+-2321723,
+-6641222,
+1127764,
+893535,
+-2804646,
+5653509,
+2657606,
+-1751466,
+-4669812,
+-827592,
+-126901,
+-2599752,
+-845148,
+1390838,
+8975481,
+-7663778,
+3572438,
+5920790,
+5233883,
+-613590,
+-881500,
+-3974422,
+-5523348,
+-3243204,
+-6405765,
+-4376438,
+1352634,
+-105650,
+2650174,
+1442151,
+5088231,
+2974595,
+-4501663,
+-841006,
+-3101819,
+-1265401,
+-2756903,
+2579743,
+2045040,
+-5328835,
+2801176,
+-386694,
+-3068782,
+3147225,
+-248211,
+-662659,
+-1112717,
+2733193,
+336344,
+3107302,
+2244003,
+4285762,
+1998904,
+1888720,
+-1174981,
+-2567532,
+-5588952,
+-101948,
+4004848,
+-610048,
+793760,
+3345423,
+716318,
+1033698,
+4011882,
+-965219,
+1258434,
+1579522,
+-4249500,
+3233648,
+-424838,
+2640541,
+1020028,
+4933599,
+-1964947,
+3237309,
+-1251962,
+-437406,
+-2749192,
+-2943112,
+-117113,
+778507,
+2757711,
+3478291,
+-661571,
+1077087,
+-3821174,
+2731860,
+3035264,
+-4424379,
+295413,
+3873542,
+-1272809,
+4145370,
+-363272,
+2240544,
+88954,
+-2016552,
+-862779,
+-844808,
+3142493,
+2019692,
+3648148,
+3857820,
+593190,
+1285134,
+-4257140,
+-1476035,
+-1951773,
+-2334649,
+1355368,
+-4390456,
+3666652,
+562848,
+-8226958,
+1134896,
+1136697,
+-2132899,
+3300228,
+1855661,
+6476864,
+5097743,
+-1373818,
+3287769,
+1709294,
+-2926119,
+2463141,
+400199,
+3051372,
+1815531,
+1746372,
+398117,
+2333959,
+-708565,
+-4241370,
+51697,
+-1626217,
+26865,
+2248300,
+3357859,
+-325912,
+194201,
+612298,
+388227,
+256018,
+-5630155,
+-1085451,
+653494,
+-1966315,
+-273079,
+-4296295,
+-2813232,
+2079672,
+2378463,
+-3869089,
+-438799,
+-725265,
+3152791,
+3461913,
+-777750,
+47521,
+2588203,
+1888001,
+-4445421,
+654349,
+811737,
+418334,
+-1854075,
+-5194402,
+-1571674,
+-622026,
+-1091628,
+1787463,
+3439585,
+2923276,
+-1997884,
+-193963,
+-731696,
+3686658,
+-1311796,
+-5219031,
+-2906251,
+2140229,
+-1846978,
+2541247,
+-3677377,
+-3935140,
+3605308,
+4807232,
+-1633864,
+344286,
+-2051894,
+2498349,
+-3085,
+379207,
+-701595,
+-1080351,
+3161365,
+-1606976,
+1640595,
+3757649,
+798095,
+-3167055,
+-2288739,
+2301831,
+-3324819,
+3219538,
+516049,
+-3153835,
+7342606,
+1098913,
+-2522436,
+376783,
+47367,
+530901,
+-395499,
+304200,
+};
+
+#define SIZE 24
+__attribute__ ((section(".heapsram"))) int g_mA[SIZE][SIZE];
+__attribute__ ((section(".heapsram"))) int g_mB[SIZE][SIZE];
+__attribute__ ((section(".heapsram"))) int g_mC[SIZE][SIZE];
+__attribute__ ((section(".heapsram"))) int g_mB_tmp[SIZE][SIZE];
diff --git a/astral/parMatrixMul8/Makefile b/astral/parMatrixMul8/Makefile
new file mode 100755
index 0000000..7c755b8
--- /dev/null
+++ b/astral/parMatrixMul8/Makefile
@@ -0,0 +1,8 @@
+PULP_APP = test
+PULP_APP_SRCS = matrixMul.c
+
+PULP_CFLAGS = -O3
+
+include $(PULP_SDK_HOME)/install/rules/pulp.mk
+
+#pulp-bench-reg --name=parMatrixMul8.cycles --module=pulp_rtl_testset --pipeline=$(PIPELINE) --artefact=pulp_rtl_testset --cmd="make run -f Makefile.sdk" --probe-regexp='matrixMul -> success, nr. of errors: 0, execution time: (\d+)' --params="platform($(platformName)),compiler($(OR1K_TOOLCHAIN_TYPE)),nbCores(4),elemSize(8)" --probe-regexp='matrixMulTranspose -> success, nr. of errors: 0, execution time: (\d+)' --params="platform($(platformName)),compiler($(OR1K_TOOLCHAIN_TYPE)),nbCores(4),elemSize(8),transposed"
diff --git a/astral/parMatrixMul8/gen_stimuli.py b/astral/parMatrixMul8/gen_stimuli.py
new file mode 100755
index 0000000..153d5c3
--- /dev/null
+++ b/astral/parMatrixMul8/gen_stimuli.py
@@ -0,0 +1,54 @@
+#!/usr/bin/env python
+
+import sys
+import random
+
+
+def write_arr(f, name, arr):
+    f.write('const char %s[] = {\n' % name)
+    for v in arr:
+        f.write('%d,\n' % (v))
+    f.write('};\n\n')
+    return
+
+################################################################################
+f = open('parMatrixMul8_stimuli.h', 'w')
+
+
+SIZE = 24 
+RANGE = 4
+
+m_a   = []
+m_b   = []
+m_exp = []
+
+for i in range(0,SIZE):
+    for j in range(0,SIZE):
+        a = random.randint(-RANGE, RANGE-1)
+        b = random.randint(-RANGE, RANGE-1)
+
+        m_a.append(a)
+        m_b.append(b)
+
+for i in range(0,SIZE):
+    for j in range(0,SIZE):
+        r = 0
+
+        for k in range (0,SIZE):
+            r = r + m_a[i * SIZE + k] * m_b[k * SIZE + j]
+
+        m_exp.append(r)
+
+
+write_arr(f, 'm_a',   m_a)
+write_arr(f, 'm_b',   m_b)
+write_arr(f, 'm_exp', m_exp)
+
+f.write('#define SIZE %d\n' % SIZE)
+
+
+f.write('__attribute__ ((section(".heapsram"))) char g_mA[SIZE][SIZE];\n')
+f.write('__attribute__ ((section(".heapsram"))) char g_mB[SIZE][SIZE];\n')
+f.write('__attribute__ ((section(".heapsram"))) char g_mC[SIZE][SIZE];\n')
+f.write('__attribute__ ((section(".heapsram"))) char g_mB_tmp[SIZE][SIZE];\n')
+
diff --git a/astral/parMatrixMul8/matrixMul.c b/astral/parMatrixMul8/matrixMul.c
new file mode 100644
index 0000000..357fdf0
--- /dev/null
+++ b/astral/parMatrixMul8/matrixMul.c
@@ -0,0 +1,177 @@
+/*
+ * Copyright (C) 2018 ETH Zurich and University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+/* 
+ * Mantainer: Luca Valente, luca.valente2@unibo.it
+ */
+
+#include "pulp.h"
+
+#include "parMatrixMul8_stimuli.h"
+
+void check_matrix_mul(testresult_t *result, void (*start)(), void (*stop)());
+void check_matrix_mul_transpose(testresult_t *result, void (*start)(), void (*stop)());
+
+testcase_t testcases[] = {
+  { .name = "matrixMul",          .test = check_matrix_mul           },
+  { .name = "matrixMulTranspose", .test = check_matrix_mul_transpose },
+  {0, 0}
+};
+
+unsigned int num_cores;
+
+int main()
+{
+  if (rt_cluster_id() != 0)
+    return bench_cluster_forward(0);
+
+  num_cores = get_core_num();
+
+  if(rt_core_id() < num_cores) {
+    run_suite(testcases);
+  }
+
+  synch_barrier();
+
+  return 0;
+}
+
+void matrix_init();
+unsigned int matrix_check();
+
+void check_matrix_mul(testresult_t *result, void (*start)(), void (*stop)()) {
+  int core_id;
+  unsigned int i, j, k;
+  unsigned int chunk;
+  unsigned int lb, ub;
+
+  core_id = get_core_id();
+
+  // number of rows each core has to multiply
+  chunk = SIZE / num_cores;
+  // lower bound
+  lb = core_id * chunk;
+  // upper bound
+  ub = lb + chunk;
+
+  if(core_id == 0) {
+    matrix_init();
+  }
+
+  if(num_cores != 1) synch_barrier();
+
+  // start benchmark
+  start();
+
+  for(i = lb; i < ub; i++) {
+    for(j = 0; j < SIZE; j++) {
+      g_mC[i][j] = 0;
+  
+      for(k = 0; k < SIZE; k++) {
+        g_mC[i][j] += g_mA[i][k] * g_mB[k][j];
+      }
+    }
+  }
+
+  if(num_cores != 1) synch_barrier();
+
+  stop();
+
+  if(core_id == 0) {
+    result->errors = matrix_check();
+  }
+}
+
+void check_matrix_mul_transpose(testresult_t *result, void (*start)(), void (*stop)()) {
+  int core_id;
+  unsigned int i, j, k;
+  unsigned int chunk;
+  unsigned int lb, ub;
+
+  core_id = get_core_id();
+
+  // number of rows each core has to multiply
+  chunk = SIZE / num_cores;
+  // lower bound
+  lb = core_id * chunk;
+  // upper bound
+  ub = lb + chunk;
+
+  if(core_id == 0) {
+    matrix_init();
+  }
+
+  if(num_cores != 1) synch_barrier();
+
+  // start benchmark
+  start();
+
+  // transpose array before using it
+  for(i = lb; i < ub; i++) {
+    for(j = 0; j < SIZE; j++) {
+      g_mB_tmp[i][j] = g_mB[j][i];
+    }
+  }
+
+  if(num_cores != 1) synch_barrier();
+
+  for(i = lb; i < ub; i++) {
+    for(j = 0; j < SIZE; j++) {
+      g_mC[i][j] = 0;
+
+      for(k = 0; k < SIZE; k++) {
+        g_mC[i][j] += g_mA[i][k] * g_mB_tmp[j][k];
+      }
+    }
+  }
+
+  if(num_cores != 1) synch_barrier();
+
+  stop();
+
+  if(core_id == 0) {
+    result->errors = matrix_check();
+  }
+}
+
+void matrix_init() {
+  unsigned int i, j;
+
+  // init, copy to TCDM
+  for(i = 0; i < SIZE; i++) {
+    for(j = 0; j < SIZE; j++) {
+      g_mA[i][j] = m_a[i * SIZE + j];
+      g_mB[i][j] = m_b[i * SIZE + j];
+      g_mC[i][j] = 0;
+    }
+  }
+}
+
+unsigned int matrix_check() {
+  unsigned int errors = 0;
+  unsigned int i, j;
+  // check
+  for(i = 0; i < SIZE; i++) {
+    for(j = 0; j < SIZE; j++) {
+      if(g_mC[i][j] != m_exp[i * SIZE + j]) {
+        printf("At index %d, %d\n", i, j, 0, 0);
+        errors++;
+      }
+    }
+  }
+
+  return errors;
+}
diff --git a/astral/parMatrixMul8/parMatrixMul8_stimuli.h b/astral/parMatrixMul8/parMatrixMul8_stimuli.h
new file mode 100644
index 0000000..78d026c
--- /dev/null
+++ b/astral/parMatrixMul8/parMatrixMul8_stimuli.h
@@ -0,0 +1,1742 @@
+const char m_a[] = {
+-1,
+0,
+0,
+0,
+-4,
+-4,
+-4,
+1,
+-1,
+-2,
+-3,
+3,
+2,
+-4,
+1,
+-2,
+-3,
+0,
+-3,
+-4,
+-3,
+-1,
+-3,
+-3,
+2,
+-2,
+2,
+3,
+0,
+2,
+0,
+-1,
+-3,
+-2,
+-1,
+-3,
+1,
+-1,
+-1,
+0,
+-1,
+-2,
+-1,
+-4,
+-2,
+-3,
+-1,
+-2,
+1,
+1,
+2,
+-1,
+0,
+2,
+1,
+-3,
+2,
+-3,
+0,
+0,
+0,
+2,
+-2,
+3,
+-4,
+2,
+-2,
+-3,
+-2,
+-3,
+3,
+-3,
+0,
+1,
+-1,
+-1,
+1,
+-2,
+-1,
+-4,
+-2,
+-1,
+-4,
+-4,
+-2,
+-1,
+-2,
+-3,
+-4,
+-4,
+1,
+0,
+-2,
+1,
+-1,
+1,
+2,
+-3,
+-4,
+0,
+2,
+-3,
+-3,
+0,
+0,
+-2,
+-3,
+-1,
+-2,
+-1,
+-4,
+-1,
+-4,
+-3,
+-3,
+0,
+1,
+-3,
+-4,
+1,
+2,
+0,
+2,
+-1,
+-4,
+-1,
+-4,
+2,
+2,
+-1,
+3,
+0,
+0,
+-1,
+1,
+3,
+0,
+2,
+0,
+3,
+2,
+2,
+2,
+-3,
+2,
+2,
+-4,
+0,
+-3,
+-1,
+-1,
+-3,
+-4,
+3,
+2,
+0,
+2,
+1,
+0,
+-2,
+-1,
+3,
+3,
+0,
+1,
+-3,
+3,
+0,
+-3,
+2,
+-3,
+1,
+2,
+0,
+3,
+-1,
+0,
+-2,
+-4,
+0,
+-3,
+0,
+3,
+0,
+-1,
+-3,
+0,
+-2,
+-4,
+-4,
+-1,
+1,
+2,
+-4,
+1,
+-4,
+-4,
+-3,
+-3,
+3,
+-1,
+-2,
+-4,
+2,
+1,
+2,
+-1,
+1,
+3,
+0,
+2,
+-4,
+-3,
+1,
+-2,
+0,
+1,
+1,
+3,
+-2,
+-4,
+2,
+-1,
+-1,
+1,
+2,
+-2,
+1,
+1,
+2,
+0,
+3,
+-4,
+2,
+-1,
+-1,
+-3,
+-3,
+-3,
+-3,
+2,
+-3,
+1,
+0,
+3,
+-3,
+3,
+3,
+-4,
+1,
+-4,
+1,
+2,
+-3,
+1,
+-4,
+-1,
+-4,
+-3,
+-2,
+2,
+3,
+0,
+-4,
+1,
+3,
+0,
+3,
+-1,
+-4,
+3,
+-2,
+-2,
+2,
+-4,
+-1,
+-1,
+3,
+-2,
+-4,
+-1,
+-1,
+2,
+-1,
+2,
+0,
+-2,
+-4,
+0,
+2,
+-4,
+0,
+1,
+-4,
+2,
+0,
+-1,
+-1,
+-3,
+2,
+3,
+2,
+-2,
+1,
+-4,
+-3,
+1,
+2,
+-1,
+-4,
+3,
+-1,
+2,
+-2,
+-3,
+0,
+-1,
+3,
+-4,
+-3,
+-3,
+-3,
+0,
+0,
+3,
+1,
+-2,
+-2,
+-4,
+2,
+2,
+0,
+3,
+-2,
+1,
+3,
+-1,
+3,
+-4,
+-2,
+0,
+-4,
+0,
+-1,
+-3,
+-2,
+-4,
+-4,
+-2,
+-4,
+-3,
+-3,
+0,
+2,
+-4,
+1,
+0,
+-3,
+0,
+-4,
+-4,
+3,
+3,
+-2,
+1,
+3,
+3,
+1,
+-3,
+-4,
+-4,
+1,
+1,
+2,
+-4,
+0,
+3,
+-1,
+3,
+1,
+-4,
+-2,
+1,
+-1,
+3,
+3,
+-1,
+-1,
+0,
+-2,
+2,
+-2,
+-2,
+1,
+3,
+1,
+3,
+-1,
+2,
+-1,
+0,
+1,
+2,
+-4,
+-3,
+1,
+-3,
+-1,
+-4,
+-1,
+-4,
+-4,
+-3,
+-3,
+3,
+0,
+3,
+0,
+-2,
+3,
+-3,
+3,
+-1,
+-1,
+2,
+-2,
+1,
+-4,
+0,
+-3,
+3,
+0,
+-2,
+-1,
+3,
+2,
+-2,
+3,
+-3,
+0,
+-2,
+1,
+0,
+-1,
+1,
+3,
+-4,
+3,
+0,
+3,
+-4,
+3,
+-1,
+-1,
+1,
+3,
+1,
+0,
+-4,
+3,
+-3,
+-4,
+-1,
+2,
+-3,
+-4,
+-2,
+-3,
+2,
+3,
+-2,
+-2,
+-4,
+-1,
+1,
+-4,
+-2,
+2,
+-3,
+1,
+-3,
+-2,
+2,
+3,
+-2,
+0,
+1,
+-1,
+2,
+-2,
+1,
+-1,
+-4,
+-3,
+-4,
+-3,
+-3,
+-3,
+-4,
+3,
+1,
+1,
+3,
+-4,
+0,
+-4,
+2,
+2,
+3,
+2,
+-4,
+1,
+-1,
+-2,
+-2,
+2,
+-2,
+2,
+-3,
+2,
+0,
+0,
+0,
+-4,
+-2,
+-2,
+2,
+1,
+0,
+2,
+3,
+2,
+3,
+-3,
+0,
+-3,
+-2,
+2,
+-1,
+2,
+-3,
+1,
+0,
+1,
+3,
+-4,
+0,
+-3,
+-1,
+1,
+-2,
+-2,
+0,
+0,
+1,
+-3,
+-2,
+0,
+3,
+2,
+0,
+3,
+-4,
+2,
+1,
+3,
+-2,
+-2,
+-3,
+2,
+2,
+2,
+-1,
+3,
+2,
+-3,
+};
+
+const char m_b[] = {
+3,
+-2,
+-3,
+-4,
+-3,
+2,
+-4,
+-1,
+-4,
+-4,
+-1,
+-3,
+3,
+0,
+2,
+1,
+2,
+-2,
+0,
+3,
+-4,
+-2,
+2,
+-1,
+1,
+-4,
+-4,
+-1,
+1,
+-3,
+-2,
+1,
+-3,
+1,
+-2,
+2,
+1,
+-3,
+-1,
+-4,
+1,
+-1,
+-3,
+3,
+0,
+0,
+-2,
+0,
+-4,
+-3,
+0,
+-3,
+0,
+3,
+-1,
+-4,
+-2,
+2,
+-4,
+-1,
+-3,
+1,
+-4,
+-4,
+0,
+2,
+3,
+2,
+0,
+-3,
+-1,
+-1,
+2,
+2,
+-2,
+-2,
+-2,
+-4,
+3,
+-3,
+2,
+1,
+-1,
+1,
+-3,
+-1,
+2,
+-1,
+1,
+-1,
+0,
+-4,
+-2,
+3,
+0,
+-4,
+-2,
+3,
+-3,
+1,
+-1,
+-1,
+-2,
+1,
+-2,
+-4,
+0,
+0,
+0,
+0,
+-4,
+-3,
+3,
+-2,
+3,
+3,
+2,
+3,
+-4,
+1,
+-2,
+-3,
+-3,
+2,
+3,
+-4,
+-1,
+3,
+0,
+3,
+2,
+1,
+-1,
+-1,
+-4,
+-1,
+3,
+2,
+-1,
+3,
+0,
+3,
+0,
+1,
+-4,
+0,
+3,
+2,
+-4,
+-2,
+-2,
+-4,
+-3,
+-2,
+1,
+0,
+-2,
+1,
+3,
+-2,
+-1,
+-3,
+0,
+3,
+0,
+0,
+0,
+2,
+3,
+-4,
+-4,
+2,
+2,
+3,
+1,
+0,
+0,
+-2,
+1,
+-2,
+-4,
+-3,
+0,
+-2,
+2,
+3,
+2,
+3,
+3,
+1,
+1,
+1,
+3,
+1,
+3,
+2,
+-3,
+1,
+-1,
+-2,
+-2,
+-2,
+-1,
+2,
+2,
+0,
+-3,
+-1,
+2,
+-4,
+-2,
+2,
+-3,
+-2,
+-4,
+3,
+-1,
+3,
+2,
+3,
+-1,
+-1,
+-1,
+-2,
+-2,
+2,
+2,
+-1,
+-3,
+-1,
+3,
+-3,
+3,
+3,
+2,
+-3,
+2,
+-3,
+-3,
+-2,
+3,
+-1,
+0,
+-2,
+-1,
+-1,
+2,
+-1,
+-2,
+1,
+-4,
+1,
+3,
+2,
+1,
+2,
+-1,
+-1,
+-1,
+-4,
+0,
+3,
+1,
+-2,
+0,
+2,
+-2,
+-2,
+3,
+-1,
+-1,
+0,
+3,
+3,
+-1,
+0,
+0,
+-4,
+1,
+-4,
+-4,
+0,
+2,
+3,
+-3,
+-2,
+-3,
+-3,
+-2,
+0,
+-4,
+-1,
+0,
+-2,
+1,
+-1,
+-4,
+1,
+2,
+0,
+-2,
+0,
+2,
+2,
+2,
+3,
+-3,
+1,
+0,
+-2,
+2,
+3,
+-1,
+-2,
+1,
+-3,
+-1,
+2,
+2,
+1,
+3,
+-2,
+0,
+-2,
+-4,
+1,
+1,
+1,
+-2,
+-1,
+2,
+0,
+1,
+-1,
+-3,
+-1,
+1,
+-1,
+-1,
+-3,
+-4,
+-2,
+-2,
+-1,
+0,
+3,
+3,
+0,
+-2,
+-2,
+-2,
+-3,
+2,
+2,
+1,
+3,
+0,
+3,
+0,
+-1,
+-1,
+3,
+3,
+-4,
+1,
+1,
+-2,
+-4,
+-4,
+3,
+1,
+0,
+-1,
+-4,
+-2,
+2,
+0,
+1,
+-1,
+0,
+-3,
+-2,
+-1,
+1,
+-3,
+-2,
+2,
+-4,
+-3,
+-3,
+0,
+0,
+-3,
+-3,
+2,
+0,
+1,
+-2,
+-3,
+-1,
+3,
+-1,
+3,
+-3,
+-4,
+-4,
+0,
+-4,
+-1,
+2,
+1,
+0,
+0,
+2,
+2,
+3,
+2,
+-1,
+0,
+-3,
+-3,
+3,
+0,
+-4,
+0,
+-2,
+-2,
+-1,
+1,
+3,
+3,
+1,
+-3,
+1,
+-2,
+-1,
+2,
+0,
+0,
+2,
+-1,
+-4,
+-1,
+-3,
+0,
+2,
+-4,
+-1,
+0,
+-2,
+-1,
+2,
+-2,
+-2,
+3,
+3,
+0,
+3,
+1,
+0,
+2,
+3,
+2,
+-2,
+-4,
+-1,
+3,
+3,
+-1,
+2,
+0,
+-2,
+2,
+-4,
+-3,
+-3,
+-3,
+-3,
+-3,
+0,
+3,
+3,
+3,
+0,
+-3,
+-1,
+2,
+3,
+-3,
+-3,
+-4,
+-1,
+3,
+0,
+-4,
+-2,
+-3,
+1,
+3,
+3,
+-2,
+2,
+-4,
+-2,
+1,
+-3,
+-3,
+-2,
+3,
+-3,
+0,
+-4,
+0,
+2,
+-4,
+0,
+1,
+-1,
+3,
+3,
+3,
+1,
+3,
+-4,
+-4,
+-1,
+-3,
+-3,
+0,
+1,
+-1,
+-3,
+-4,
+2,
+3,
+0,
+-4,
+-2,
+-3,
+0,
+3,
+2,
+0,
+-2,
+-4,
+-3,
+-3,
+-3,
+-2,
+2,
+-4,
+-4,
+2,
+3,
+-3,
+2,
+-2,
+-2,
+-1,
+-2,
+-2,
+-1,
+3,
+-4,
+-4,
+3,
+-3,
+-3,
+-1,
+-1,
+2,
+3,
+-3,
+-1,
+-3,
+-4,
+};
+
+const char m_exp[] = {
+27,
+-26,
+-19,
+13,
+47,
+17,
+-5,
+14,
+88,
+66,
+2,
+14,
+6,
+-39,
+56,
+35,
+-19,
+61,
+-13,
+4,
+46,
+1,
+70,
+27,
+-22,
+-28,
+-13,
+-5,
+-4,
+-6,
+-10,
+5,
+26,
+14,
+3,
+-1,
+-3,
+0,
+10,
+43,
+4,
+32,
+-3,
+6,
+36,
+26,
+66,
+20,
+-20,
+-10,
+21,
+-17,
+-5,
+36,
+-9,
+11,
+10,
+17,
+-32,
+12,
+20,
+9,
+-1,
+15,
+-42,
+-21,
+-51,
+24,
+20,
+26,
+10,
+42,
+-4,
+2,
+-11,
+41,
+22,
+4,
+-17,
+10,
+28,
+-10,
+2,
+10,
+25,
+32,
+18,
+43,
+-3,
+-9,
+-9,
+-4,
+25,
+-5,
+9,
+24,
+54,
+43,
+14,
+59,
+18,
+36,
+-7,
+30,
+50,
+-30,
+21,
+3,
+27,
+7,
+21,
+55,
+-17,
+-2,
+9,
+17,
+31,
+8,
+12,
+29,
+43,
+-17,
+14,
+-55,
+13,
+50,
+24,
+-13,
+-33,
+15,
+-45,
+9,
+36,
+-9,
+-6,
+14,
+-1,
+2,
+-27,
+-37,
+-40,
+-14,
+2,
+-10,
+6,
+27,
+-8,
+-13,
+-17,
+-4,
+28,
+51,
+-5,
+-14,
+2,
+-23,
+17,
+35,
+41,
+-3,
+-37,
+-12,
+-49,
+-31,
+7,
+15,
+36,
+-15,
+-21,
+6,
+2,
+36,
+-12,
+-27,
+-34,
+9,
+44,
+8,
+57,
+20,
+-22,
+-22,
+-11,
+-15,
+-8,
+-15,
+-9,
+49,
+37,
+26,
+6,
+30,
+-17,
+-43,
+-3,
+-31,
+36,
+49,
+-15,
+8,
+51,
+-8,
+17,
+-51,
+-16,
+-21,
+12,
+49,
+-11,
+41,
+36,
+15,
+-5,
+-69,
+60,
+14,
+-24,
+-21,
+23,
+3,
+41,
+41,
+-27,
+0,
+33,
+63,
+18,
+1,
+-36,
+-20,
+18,
+7,
+-1,
+37,
+-33,
+30,
+36,
+-29,
+38,
+57,
+-8,
+-21,
+-26,
+15,
+-10,
+-4,
+-17,
+-19,
+8,
+-9,
+-13,
+0,
+10,
+-30,
+59,
+12,
+11,
+51,
+60,
+19,
+35,
+6,
+3,
+-5,
+-19,
+-10,
+20,
+6,
+-12,
+-5,
+5,
+11,
+33,
+-33,
+-27,
+-2,
+-2,
+3,
+53,
+-1,
+-15,
+-10,
+13,
+-5,
+24,
+-17,
+10,
+8,
+-3,
+41,
+-9,
+9,
+0,
+30,
+7,
+7,
+-10,
+-28,
+21,
+13,
+4,
+-3,
+44,
+9,
+-49,
+-31,
+-51,
+23,
+16,
+-13,
+-12,
+36,
+23,
+34,
+-39,
+8,
+20,
+2,
+32,
+73,
+24,
+-16,
+0,
+-30,
+17,
+52,
+-10,
+28,
+-43,
+-1,
+-34,
+16,
+-4,
+39,
+31,
+5,
+5,
+3,
+36,
+48,
+5,
+28,
+12,
+57,
+25,
+-28,
+14,
+34,
+49,
+11,
+41,
+20,
+-12,
+-18,
+-43,
+-29,
+58,
+26,
+44,
+36,
+-100,
+-46,
+-24,
+0,
+54,
+4,
+-2,
+21,
+5,
+-12,
+22,
+9,
+-35,
+-5,
+-36,
+-22,
+-10,
+-11,
+-15,
+34,
+20,
+-33,
+17,
+39,
+-13,
+-32,
+-36,
+-22,
+-3,
+-38,
+-39,
+-6,
+0,
+15,
+-6,
+-15,
+9,
+-10,
+45,
+16,
+17,
+13,
+-8,
+15,
+11,
+-4,
+59,
+18,
+-9,
+12,
+69,
+27,
+-30,
+16,
+7,
+30,
+41,
+-30,
+9,
+-4,
+23,
+-5,
+8,
+3,
+-75,
+-25,
+5,
+14,
+12,
+-21,
+-21,
+6,
+-13,
+15,
+8,
+-6,
+30,
+24,
+41,
+-12,
+5,
+39,
+18,
+-4,
+-60,
+19,
+-1,
+-3,
+-7,
+16,
+-11,
+-51,
+2,
+-10,
+-17,
+-32,
+39,
+-35,
+-46,
+10,
+27,
+1,
+21,
+13,
+78,
+-12,
+-2,
+-28,
+-20,
+11,
+75,
+85,
+-18,
+21,
+-7,
+-9,
+33,
+19,
+71,
+23,
+25,
+2,
+47,
+32,
+8,
+24,
+46,
+-22,
+23,
+18,
+-58,
+66,
+52,
+22,
+32,
+25,
+-34,
+-26,
+-21,
+-69,
+6,
+27,
+-28,
+-13,
+-15,
+10,
+-20,
+-16,
+20,
+25,
+7,
+22,
+52,
+-4,
+-8,
+-28,
+-18,
+-6,
+-36,
+6,
+-26,
+12,
+32,
+8,
+11,
+16,
+0,
+-18,
+-33,
+-10,
+11,
+-40,
+-6,
+-29,
+-34,
+-30,
+15,
+15,
+-27,
+9,
+8,
+-43,
+0,
+-5,
+4,
+12,
+60,
+-11,
+-4,
+-32,
+-25,
+-38,
+-37,
+-17,
+-1,
+20,
+37,
+43,
+7,
+-2,
+12,
+-7,
+-7,
+-14,
+-30,
+9,
+50,
+-18,
+-9,
+-1,
+-9,
+6,
+16,
+-38,
+-13,
+30,
+};
+
+#define SIZE 24
+__attribute__ ((section(".heapsram"))) char g_mA[SIZE][SIZE];
+__attribute__ ((section(".heapsram"))) char g_mB[SIZE][SIZE];
+__attribute__ ((section(".heapsram"))) char g_mC[SIZE][SIZE];
+__attribute__ ((section(".heapsram"))) char g_mB_tmp[SIZE][SIZE];
diff --git a/astral/redmule b/astral/redmule
new file mode 120000
index 0000000..bd0da93
--- /dev/null
+++ b/astral/redmule
@@ -0,0 +1 @@
+../hwpe/redmule/
\ No newline at end of file
diff --git a/astral/redmule_256iter b/astral/redmule_256iter
new file mode 120000
index 0000000..e81b225
--- /dev/null
+++ b/astral/redmule_256iter
@@ -0,0 +1 @@
+../hwpe/redmule_256iter/
\ No newline at end of file
diff --git a/astral/redmule_softclear b/astral/redmule_softclear
new file mode 120000
index 0000000..2904f52
--- /dev/null
+++ b/astral/redmule_softclear
@@ -0,0 +1 @@
+../hwpe/redmule_softclear/
\ No newline at end of file
diff --git a/astral/softex b/astral/softex
new file mode 120000
index 0000000..e1a6d4b
--- /dev/null
+++ b/astral/softex
@@ -0,0 +1 @@
+../hwpe/softex/
\ No newline at end of file
diff --git a/carfield.yaml b/carfield.yaml
new file mode 100644
index 0000000..751550e
--- /dev/null
+++ b/carfield.yaml
@@ -0,0 +1,19 @@
+carfield:
+  parMatrixMul8:
+    path: ./carfield/parMatrixMul8
+    command: make clean all run
+  parMatrixMul16:
+    path: ./carfield/parMatrixMul16
+    command: make clean all run
+  parMatrixMul32:
+    path: ./carfield/parMatrixMul32
+    command: make clean all run
+  dmr_matmul:
+    path: ./carfield/dmr_matmul
+    command: make clean all run
+  redmule:
+    path: ./carfield/redmule
+    command: make clean all run
+  neureka:
+    path: ./carfield/neureka
+    command: make clean all run
\ No newline at end of file
diff --git a/carfield/dmr_matmul/Makefile b/carfield/dmr_matmul/Makefile
new file mode 100644
index 0000000..66ed60e
--- /dev/null
+++ b/carfield/dmr_matmul/Makefile
@@ -0,0 +1,12 @@
+PULP_APP = test
+PULP_APP_SRCS = dmr_matmul.c
+
+PULP_CFLAGS = -O3
+PULP_LDFLAGS = -lm
+
+ifeq ($(fault_inject),1)
+	export FAULT_INJECTION=1
+	export FAULT_INJECTION_SCRIPT=$(CURDIR)/pulp_inject_fault.tcl
+endif
+
+include $(PULP_SDK_HOME)/install/rules/pulp.mk
diff --git a/carfield/dmr_matmul/dmr_matmul.c b/carfield/dmr_matmul/dmr_matmul.c
new file mode 100644
index 0000000..953dfdf
--- /dev/null
+++ b/carfield/dmr_matmul/dmr_matmul.c
@@ -0,0 +1,100 @@
+/*
+* @Author: Michael Rogenmoser
+* @Date:   2023-02-17 18:00:21
+* @Last Modified by:   Michael Rogenmoser
+* @Last Modified time: 2023-02-17 18:15:33
+*/
+#include <pulp.h>
+#include <stdio.h>
+#include "matmul.h"
+
+#define N_ITERS 1
+#define max(x,y) (x > y ? x : y)
+#define min(x,y) (x < y ? x : y)
+
+__attribute__ ((section(".heapsram"))) int A[SIZE][SIZE];
+__attribute__ ((section(".heapsram"))) int B[SIZE][SIZE];
+__attribute__ ((section(".heapsram"))) int C[SIZE][SIZE];
+
+void initialize_mat();
+
+void initialize_mat() {
+  int i,j;
+
+  for (i=0;i<SIZE;i++) {
+    for (j=0;j<SIZE;j++) {
+      A[i][j] = A_init[i][j];
+      B[i][j] = B_init[i][j];
+    }
+  }
+
+}
+
+void matrix_multiplication(testresult_t *result, void (*start)(), void (*stop)());
+
+testcase_t testcases[] = {
+  { .name = "Matrix Multiplication", .test = matrix_multiplication },
+  {0, 0}
+};
+
+int main() {
+  if (rt_cluster_id() != 0)
+    return bench_cluster_forward(0);
+
+  hmr_self_enable_dmr();
+  hmr_set_dmr_config_all(0    , // Core ID
+                         true , // Rapid recovery enabled
+                         true , // Setback enabled
+                         false); // Synch req
+  printf("Available Config: %x\n", hmr_get_available_config(rt_cluster_id()));
+  printf("after setup: %x\n", hmr_get_active_cores(rt_cluster_id()));
+
+  hmr_setup_barrier(hmr_get_active_cores(0));
+
+  int nbErrors = run_suite(testcases);
+
+  synch_barrier();
+
+  return nbErrors != 0;
+}
+
+void matrix_multiplication(testresult_t *result, void (*start)(), void (*stop)()) {
+  int coreid = rt_core_id();
+  int numcores = 6;
+  int *CHKSUM_RESULT;
+  short int i, iter, j, k;
+  int lb, ub, chunk;
+
+  if (coreid == 0){
+    printf("Start ParMatrixMul\n",0,0,0,0);
+    // initialize matrix A and B
+    initialize_mat();
+  }
+  //number of rows each core has to multiply
+  chunk = SIZE / numcores;
+  //lower bound
+  lb = coreid * chunk +  min(coreid, SIZE % numcores);
+  //upper bound
+  ub = (coreid + 1) * chunk + min(coreid + 1, SIZE % numcores);
+  if (coreid == numcores-1) {ub = SIZE;}
+
+  synch_barrier();
+
+  /********************* Benchmark Execution *********************/
+  if (coreid<numcores) {
+    start();
+    for (iter = 0; iter < N_ITERS; iter++) {
+      for (i = lb; i < ub; i++) {
+        for (k = 0; k < SIZE; k++) {
+          C[i][k] = 0;
+          for (j = 0; j < SIZE; j++)
+            C[i][k] += A[i][j] * B[j][k];
+        }
+      }
+    }
+  }
+  synch_barrier();
+  stop();
+
+  /********************* Benchmark Execution *********************/
+}
diff --git a/carfield/dmr_matmul/matmul.h b/carfield/dmr_matmul/matmul.h
new file mode 100644
index 0000000..eaed788
--- /dev/null
+++ b/carfield/dmr_matmul/matmul.h
@@ -0,0 +1,56 @@
+#define SIZE           24
+#define MAT_SIZE       (SIZE*SIZE*sizeof(int))
+
+int A_init[SIZE][SIZE]  = {
+    {181,242,121,183,206,92,200,157,59,246,13,24,108,168,125,179,134,202,74,79,86,66,63,29},
+    {178,198,233,162,184,134,150,155,153,199,114,94,81,0,92,164,206,186,166,191,173,25,82,38},
+    {216,38,65,42,106,189,132,201,214,41,24,231,249,78,84,109,168,159,127,107,217,254,197,217},
+    {64,22,112,106,14,207,197,73,149,38,106,250,17,254,196,121,218,22,28,237,31,87,82,132},
+    {165,67,227,22,65,61,140,27,231,194,168,145,33,44,175,127,31,56,194,227,186,13,252,27},
+    {160,183,169,113,226,50,116,138,23,147,247,20,73,174,185,178,196,200,232,178,234,117,9,251},
+    {36,192,256,31,150,165,159,79,217,121,13,123,77,233,87,219,20,225,3,27,22,69,216,92},
+    {151,236,72,44,178,105,211,237,216,24,107,168,194,178,99,168,246,250,120,212,18,39,194,112},
+    {230,114,234,101,27,26,158,119,246,3,119,80,227,214,96,87,35,179,195,156,71,84,53,89},
+    {188,225,242,243,130,165,133,225,222,186,139,152,171,23,156,42,214,11,46,45,206,78,34,196},
+    {232,161,201,187,247,80,215,11,9,164,220,201,118,161,98,1,155,14,60,7,82,118,211,230},
+    {4,91,20,188,232,229,234,82,240,175,105,12,160,204,177,201,129,241,83,235,125,216,91,29},
+    {120,97,74,42,130,29,15,160,252,82,140,48,226,25,154,19,207,190,197,216,177,8,14,111},
+    {175,204,50,69,35,251,68,208,171,18,135,186,200,139,182,170,7,144,83,83,79,15,192,165},
+    {236,198,116,130,248,198,107,218,107,105,178,209,163,22,218,163,144,65,55,48,102,217,57,158},
+    {192,201,215,171,104,183,55,72,75,214,72,181,85,212,173,88,183,144,221,90,117,129,201,103},
+    {32,75,38,70,53,76,66,27,221,10,17,122,234,218,56,39,63,112,194,124,178,81,244,38},
+    {76,225,145,248,141,149,27,180,84,164,149,19,101,179,147,10,93,14,219,241,77,241,136,156},
+    {120,146,26,53,175,54,79,69,113,29,204,90,203,227,1,180,227,100,44,206,163,17,215,25},
+    {134,167,185,215,181,100,189,203,220,158,112,2,124,100,183,251,110,159,21,118,77,12,249,5},
+    {161,110,245,113,227,139,107,236,250,244,231,63,191,75,102,110,18,210,26,133,53,66,50,140},
+    {151,76,114,163,117,76,250,117,154,97,73,235,147,9,176,91,246,256,72,13,228,172,109,156},
+    {105,13,88,253,167,189,23,42,137,249,15,115,118,234,126,129,167,54,211,229,243,249,143,118},
+    {191,245,66,166,36,208,209,127,193,246,68,224,31,65,56,167,241,144,111,30,54,115,34,56}
+};
+
+int B_init[SIZE][SIZE]  = {
+    {0,0,0,1,0,0,1,1,0,0,0,0,1,0,0,1,0,0,0,1,0,0,1,1},
+    {0,1,0,1,0,0,1,1,0,1,0,1,0,1,0,0,1,1,0,1,1,0,0,1},
+    {1,1,0,1,1,1,1,0,1,0,0,0,0,0,0,0,0,0,0,1,0,0,1,1},
+    {0,0,0,1,1,0,0,0,1,0,1,1,0,1,1,1,0,1,0,0,1,1,1,1},
+    {1,1,0,0,1,1,0,0,1,1,0,1,0,1,1,1,1,1,1,1,1,1,0,0},
+    {1,0,1,0,1,1,0,1,0,0,1,1,0,1,1,0,0,0,0,0,1,1,1,0},
+    {1,1,1,0,1,1,1,0,1,1,0,0,1,0,1,1,1,1,0,0,0,1,0,0},
+    {1,1,1,1,1,1,0,1,1,0,1,1,0,0,0,0,1,0,1,0,1,0,0,0},
+    {1,1,1,1,1,1,0,1,0,1,0,0,0,0,0,1,0,1,0,0,1,0,0,1},
+    {1,1,0,1,0,0,0,0,1,1,1,0,0,1,0,0,1,1,0,1,1,0,0,0},
+    {1,1,0,1,1,0,1,1,0,1,1,0,1,0,1,1,1,1,0,0,1,1,0,0},
+    {0,0,0,0,1,1,0,1,0,0,1,1,0,0,0,1,0,0,1,0,0,0,1,1},
+    {0,0,1,0,1,1,0,1,0,0,1,1,1,1,1,0,1,0,1,0,0,0,1,1},
+    {1,0,0,0,0,1,1,0,0,1,0,1,1,1,0,0,1,0,1,1,1,1,0,0},
+    {1,1,1,1,1,1,0,0,1,1,1,0,0,1,1,1,0,1,0,1,0,0,0,0},
+    {1,1,0,1,1,0,1,1,0,0,0,1,0,1,0,1,0,0,1,1,0,1,1,0},
+    {0,1,1,1,0,0,1,0,1,1,0,0,0,1,1,0,1,1,1,0,0,0,0,1},
+    {1,1,0,1,1,0,0,0,1,0,0,0,1,1,1,1,0,0,0,0,1,0,0,1},
+    {0,0,0,0,1,0,0,1,0,0,1,1,0,0,1,1,0,0,0,0,1,0,0,1},
+    {0,0,1,1,1,0,0,1,0,0,0,1,1,0,1,0,1,1,1,1,0,1,1,0},
+    {1,0,1,0,1,0,1,0,0,1,1,1,1,0,1,1,0,1,0,1,0,0,0,1},
+    {0,0,1,0,0,0,0,0,1,1,1,1,1,1,0,0,1,1,1,0,0,1,1,0},
+    {1,0,1,1,0,1,1,0,0,0,0,1,1,1,0,1,1,0,1,0,0,1,1,1},
+    {0,0,0,1,0,1,0,0,0,0,1,1,1,1,0,1,0,1,0,0,1,0,0,1}
+};
\ No newline at end of file
diff --git a/carfield/dmr_matmul/pulp_inject_fault.tcl b/carfield/dmr_matmul/pulp_inject_fault.tcl
new file mode 100644
index 0000000..857d373
--- /dev/null
+++ b/carfield/dmr_matmul/pulp_inject_fault.tcl
@@ -0,0 +1,51 @@
+# Copyright 2023 ETH Zurich and University of Bologna.
+# Solderpad Hardware License, Version 0.51, see LICENSE for details.
+# SPDX-License-Identifier: SHL-0.51
+#
+# Author: Michael Rogenmoser (michaero@iis.ee.ethz.ch)
+
+transcript quietly
+if {! [info exists ::env(VSIM_PATH)]} {error "Define VSIM_PATH"}
+set utils_base_path  [file join $::env(VSIM_PATH) scripts fault_injection_utils]
+set script_base_path [file join $::env(VSIM_PATH) fault_injection_sim scripts]
+
+set verbosity            2
+set log_injections       1
+# Easy way to generate a variable seed
+# set seed                 [clock seconds]
+# Default value
+set seed                 12345
+set print_statistics     1
+
+set inject_start_time 250856000000ps
+set inject_stop_time 413000000000ps
+set injection_clock "pulp_cluster_tb/cluster_i/clk_i"
+set injection_clock_trigger 0
+set fault_period 250
+set rand_initial_injection_phase 0
+# max_num set to 0 means until stop_time
+set max_num_fault_inject 0
+set signal_fault_duration 20ns
+set register_fault_duration 0ns
+
+set allow_multi_bit_upset 1
+set use_bitwidth_as_weight 0
+set check_core_output_modification 0
+set check_core_next_state_modification 0
+set reg_to_sig_ratio 1
+
+source [file join $utils_base_path pulp_extract_nets.tcl]
+
+set inject_signals_netlist []
+set inject_register_netlist []
+set output_netlist []
+set next_state_netlist []
+set assertion_disable_list []
+
+for {set idx 0} {$idx < 12} {incr idx} {
+    set inject_signals_netlist [list {*}$inject_signals_netlist {*}[get_all_core_nets $idx]]
+    set output_netlist [list {*}$output_netlist {*}[get_core_output_nets $idx]]
+}
+
+source [file join $script_base_path inject_fault.tcl]
+
diff --git a/carfield/ecc_test/Makefile b/carfield/ecc_test/Makefile
new file mode 100755
index 0000000..d19f286
--- /dev/null
+++ b/carfield/ecc_test/Makefile
@@ -0,0 +1,15 @@
+PULP_APP = test
+PULP_APP_SRCS = ecc_test.c
+
+PULP_CFLAGS = -O3
+
+ifeq ($(fault_inject),1)
+	export FAULT_INJECTION=1
+	export FAULT_INJECTION_SCRIPT=$(CURDIR)/pulp_inject_fault.tcl
+endif
+
+ifeq ($(multi_bit_upset),1)
+	export MULTI_BIT_UPSET=1
+endif
+
+include $(PULP_SDK_HOME)/install/rules/pulp.mk
diff --git a/carfield/ecc_test/ecc_test.c b/carfield/ecc_test/ecc_test.c
new file mode 100644
index 0000000..99ee1cf
--- /dev/null
+++ b/carfield/ecc_test/ecc_test.c
@@ -0,0 +1,69 @@
+/*
+ * Copyright (C) 2018 ETH Zurich and University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include <pulp.h>
+#include <stdint.h>
+#include <stdio.h>
+#include <stdlib.h>
+
+#define SIZE 0x400
+#define NUM_BANKS 16
+#define SCRUBBER_INTERVAL 2
+
+int main() {
+  // Collecting info about the core ID and the running cluster ID
+  unsigned int core_id = get_core_id();
+  unsigned int cluster_id = rt_cluster_id();
+
+  if (rt_cluster_id() != 0) return bench_cluster_forward(0);
+
+  if (core_id != 0) synch_barrier();
+
+  unsigned int *test_array = pi_l1_malloc(cluster_id, SIZE);
+
+  // Initializing the memory
+  for (int i = 0; i < SIZE; i++) {
+    pulp_write32(&test_array[i], i);
+  }
+
+  // Initialize the scrubbing interval for all memory banks
+  for (int i = 0; i < NUM_BANKS; i++)
+    tcdm_scrubber_set_interval(cluster_id, i, SCRUBBER_INTERVAL);
+
+  // Initialize the error-tracking variables
+  bool  mismatch = 0;
+  unsigned int error = 0;
+  for (int i = 0; i < SIZE; i++) {
+    mismatch = (pulp_read32(&test_array[i]) != i);
+    if (mismatch) {
+      error ++;
+      printf("Expected 0x%x, got 0x%x\n", i, pulp_read32(&test_array[i]));
+    }
+  }
+
+  unsigned int mismatch_cnt = 0;
+  unsigned int fix_cnt = 0;
+  unsigned int uncorrectable_cnt = 0;
+  for (int i = 0; i < 16; i++) {
+    mismatch_cnt += tcdm_scrubber_get_mismatch_count(cluster_id, i);
+    fix_cnt += tcdm_scrubber_get_fix_count(cluster_id, i);
+    uncorrectable_cnt += tcdm_scrubber_get_uncorrectable_count(cluster_id, i);
+  }
+
+  printf("mismatch_cnt: %d, fix_cnt: %d, uncorrectable_cnt: %d\n", mismatch_cnt, fix_cnt, uncorrectable_cnt);
+
+  return (error != 0) && (uncorrectable_cnt == 0);
+}
diff --git a/carfield/ecc_test/pulp_inject_fault.tcl b/carfield/ecc_test/pulp_inject_fault.tcl
new file mode 100644
index 0000000..45cef47
--- /dev/null
+++ b/carfield/ecc_test/pulp_inject_fault.tcl
@@ -0,0 +1,53 @@
+# Copyright 2023 ETH Zurich and University of Bologna.
+# Solderpad Hardware License, Version 0.51, see LICENSE for details.
+# SPDX-License-Identifier: SHL-0.51
+#
+# Author: Michael Rogenmoser (michaero@iis.ee.ethz.ch)
+
+transcript quietly
+if {! [info exists ::env(VSIM_PATH)]} {error "Define VSIM_PATH"}
+set utils_base_path  [file join $::env(VSIM_PATH) scripts fault_injection_utils]
+set script_base_path [file join $::env(VSIM_PATH) fault_injection_sim scripts]
+
+set verbosity            2
+set log_injections       1
+# Easy way to generate a variable seed
+# set seed                 [clock seconds]
+# Default value
+set seed                 12345
+set print_statistics     1
+
+set inject_start_time 110584000000ps
+set inject_stop_time 203880000000ps
+set injection_clock "pulp_cluster_tb/cluster_i/clk_i"
+set injection_clock_trigger 0
+set fault_period 100
+set rand_initial_injection_phase 0
+# max_num set to 0 means until stop_time
+set max_num_fault_inject 0
+set signal_fault_duration 20ns
+set register_fault_duration 0ns
+
+set allow_multi_bit_upset $::env(MULTI_BIT_UPSET)
+set use_bitwidth_as_weight 0
+set check_core_output_modification 0
+set check_core_next_state_modification 0
+set reg_to_sig_ratio 1
+
+source [file join $utils_base_path pulp_extract_nets.tcl]
+
+set inject_signals_netlist []
+set inject_register_netlist []
+set output_netlist []
+set next_state_netlist []
+set assertion_disable_list []
+
+# for {set idx 0} {$idx < 12} {incr idx} {
+#     set inject_signals_netlist [list {*}$inject_signals_netlist {*}[get_all_core_nets $idx]]
+#     set output_netlist [list {*}$output_netlist {*}[get_core_output_nets $idx]]
+# }
+
+set inject_register_netlist [list {*}$inject_register_netlist {*}[get_memory_slice {0 16} {385 449}]]
+
+source [file join $script_base_path inject_fault.tcl]
+
diff --git a/carfield/hello/Makefile b/carfield/hello/Makefile
new file mode 100644
index 0000000..d145a0f
--- /dev/null
+++ b/carfield/hello/Makefile
@@ -0,0 +1,5 @@
+PULP_APP = test
+PULP_APP_SRCS = hello.c
+PULP_CFLAGS = -O3
+
+include $(PULP_SDK_HOME)/install/rules/pulp.mk
diff --git a/carfield/hello/hello.c b/carfield/hello/hello.c
new file mode 100644
index 0000000..8741206
--- /dev/null
+++ b/carfield/hello/hello.c
@@ -0,0 +1,24 @@
+/*
+ * Copyright (C) 2018 ETH Zurich and University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include <stdio.h>
+
+int main()
+{
+  printf("Hello !\n");
+  
+  return 0;
+}
diff --git a/carfield/neureka b/carfield/neureka
new file mode 120000
index 0000000..98cb979
--- /dev/null
+++ b/carfield/neureka
@@ -0,0 +1 @@
+../hwpe/neureka/
\ No newline at end of file
diff --git a/carfield/parMatrixMul16/Makefile b/carfield/parMatrixMul16/Makefile
new file mode 100755
index 0000000..0fade4a
--- /dev/null
+++ b/carfield/parMatrixMul16/Makefile
@@ -0,0 +1,8 @@
+PULP_APP = test
+PULP_APP_SRCS = matrixMul.c
+
+PULP_CFLAGS = -O3
+
+include $(PULP_SDK_HOME)/install/rules/pulp.mk
+
+#pulp-bench-reg --name=parMatrixMul16.cycles --module=pulp_rtl_testset --pipeline=$(PIPELINE) --artefact=pulp_rtl_testset --cmd="make run -f Makefile.sdk" --probe-regexp='matrixMul -> success, nr. of errors: 0, execution time: (\d+)' --params="platform($(platformName)),compiler($(OR1K_TOOLCHAIN_TYPE)),nbCores(4),elemSize(16)" --probe-regexp='matrixMulTranspose -> success, nr. of errors: 0, execution time: (\d+)' --params="platform($(platformName)),compiler($(OR1K_TOOLCHAIN_TYPE)),nbCores(4),elemSize(16),transposed"
diff --git a/carfield/parMatrixMul16/gen_stimuli.py b/carfield/parMatrixMul16/gen_stimuli.py
new file mode 100755
index 0000000..754a8a9
--- /dev/null
+++ b/carfield/parMatrixMul16/gen_stimuli.py
@@ -0,0 +1,54 @@
+#!/usr/bin/env python
+
+import sys
+import random
+
+
+def write_arr(f, name, arr):
+    f.write('const short %s[] = {\n' % name)
+    for v in arr:
+        f.write('%d,\n' % (v))
+    f.write('};\n\n')
+    return
+
+################################################################################
+f = open('parMatrixMul16_stimuli.h', 'w')
+
+
+SIZE = 24 
+RANGE = int(2**7/SIZE)
+
+m_a   = []
+m_b   = []
+m_exp = []
+
+for i in range(0,SIZE):
+    for j in range(0,SIZE):
+        a = random.randint(-RANGE, RANGE-1)
+        b = random.randint(-RANGE, RANGE-1)
+
+        m_a.append(a)
+        m_b.append(b)
+
+for i in range(0,SIZE):
+    for j in range(0,SIZE):
+        r = 0
+
+        for k in range (0,SIZE):
+            r = r + m_a[i * SIZE + k] * m_b[k * SIZE + j]
+
+        m_exp.append(r)
+
+
+write_arr(f, 'm_a',   m_a)
+write_arr(f, 'm_b',   m_b)
+write_arr(f, 'm_exp', m_exp)
+
+f.write('#define SIZE %d\n' % SIZE)
+
+
+f.write('__attribute__ ((section(".heapsram"))) short g_mA[SIZE][SIZE];\n')
+f.write('__attribute__ ((section(".heapsram"))) short g_mB[SIZE][SIZE];\n')
+f.write('__attribute__ ((section(".heapsram"))) short g_mC[SIZE][SIZE];\n')
+f.write('__attribute__ ((section(".heapsram"))) short g_mB_tmp[SIZE][SIZE];\n')
+
diff --git a/carfield/parMatrixMul16/matrixMul.c b/carfield/parMatrixMul16/matrixMul.c
new file mode 100644
index 0000000..5cbe65f
--- /dev/null
+++ b/carfield/parMatrixMul16/matrixMul.c
@@ -0,0 +1,177 @@
+/*
+ * Copyright (C) 2018 ETH Zurich and University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+/* 
+ * Mantainer: Luca Valente, luca.valente2@unibo.it
+ */
+
+#include "pulp.h"
+
+#include "parMatrixMul16_stimuli.h"
+
+void check_matrix_mul(testresult_t *result, void (*start)(), void (*stop)());
+void check_matrix_mul_transpose(testresult_t *result, void (*start)(), void (*stop)());
+
+testcase_t testcases[] = {
+  { .name = "matrixMul",          .test = check_matrix_mul           },
+  { .name = "matrixMulTranspose", .test = check_matrix_mul_transpose },
+  {0, 0}
+};
+
+unsigned int num_cores;
+
+int main()
+{
+  if (rt_cluster_id() != 0)
+    return bench_cluster_forward(0);
+
+  num_cores = get_core_num();
+
+  if(rt_core_id() < num_cores) {
+    run_suite(testcases);
+  }
+
+  synch_barrier();
+
+  return 0;
+}
+
+void matrix_init();
+unsigned int matrix_check();
+
+void check_matrix_mul(testresult_t *result, void (*start)(), void (*stop)()) {
+  int core_id;
+  unsigned int i, j, k;
+  unsigned int chunk;
+  unsigned int lb, ub;
+
+  core_id = get_core_id();
+
+  // number of rows each core has to multiply
+  chunk = SIZE / num_cores;
+  // lower bound
+  lb = core_id * chunk;
+  // upper bound
+  ub = lb + chunk;
+
+  if(core_id == 0) {
+    matrix_init();
+  }
+
+  if(num_cores != 1) synch_barrier();
+
+  // start benchmark
+  start();
+
+  for(i = lb; i < ub; i++) {
+    for(j = 0; j < SIZE; j++) {
+      g_mC[i][j] = 0;
+  
+      for(k = 0; k < SIZE; k++) {
+        g_mC[i][j] += g_mA[i][k] * g_mB[k][j];
+      }
+    }
+  }
+
+  if(num_cores != 1) synch_barrier();
+
+  stop();
+
+  if(core_id == 0) {
+    result->errors = matrix_check();
+  }
+}
+
+void check_matrix_mul_transpose(testresult_t *result, void (*start)(), void (*stop)()) {
+  int core_id;
+  unsigned int i, j, k;
+  unsigned int chunk;
+  unsigned int lb, ub;
+
+  core_id = get_core_id();
+
+  // number of rows each core has to multiply
+  chunk = SIZE / num_cores;
+  // lower bound
+  lb = core_id * chunk;
+  // upper bound
+  ub = lb + chunk;
+
+  if(core_id == 0) {
+    matrix_init();
+  }
+
+  if(num_cores != 1) synch_barrier();
+
+  // start benchmark
+  start();
+
+  // transpose array before using it
+  for(i = lb; i < ub; i++) {
+    for(j = 0; j < SIZE; j++) {
+      g_mB_tmp[i][j] = g_mB[j][i];
+    }
+  }
+
+  if(num_cores != 1) synch_barrier();
+
+  for(i = lb; i < ub; i++) {
+    for(j = 0; j < SIZE; j++) {
+      g_mC[i][j] = 0;
+
+      for(k = 0; k < SIZE; k++) {
+        g_mC[i][j] += g_mA[i][k] * g_mB_tmp[j][k];
+      }
+    }
+  }
+
+  if(num_cores != 1) synch_barrier();
+
+  stop();
+
+  if(core_id == 0) {
+    result->errors = matrix_check();
+  }
+}
+
+void matrix_init() {
+  unsigned int i, j;
+
+  // init, copy to TCDM
+  for(i = 0; i < SIZE; i++) {
+    for(j = 0; j < SIZE; j++) {
+      g_mA[i][j] = m_a[i * SIZE + j];
+      g_mB[i][j] = m_b[i * SIZE + j];
+      g_mC[i][j] = 0;
+    }
+  }
+}
+
+unsigned int matrix_check() {
+  unsigned int errors = 0;
+  unsigned int i, j;
+  // check
+  for(i = 0; i < SIZE; i++) {
+    for(j = 0; j < SIZE; j++) {
+      if(g_mC[i][j] != m_exp[i * SIZE + j]) {
+        printf("At index %d, %d\n", i, j, 0, 0);
+        errors++;
+      }
+    }
+  }
+
+  return errors;
+}
diff --git a/carfield/parMatrixMul16/parMatrixMul16_stimuli.h b/carfield/parMatrixMul16/parMatrixMul16_stimuli.h
new file mode 100644
index 0000000..2797565
--- /dev/null
+++ b/carfield/parMatrixMul16/parMatrixMul16_stimuli.h
@@ -0,0 +1,1742 @@
+const short m_a[] = {
+3,
+2,
+-4,
+0,
+0,
+0,
+-5,
+-3,
+2,
+4,
+0,
+2,
+2,
+-3,
+2,
+-1,
+-2,
+-1,
+1,
+4,
+-3,
+-3,
+-1,
+0,
+4,
+-2,
+-1,
+-4,
+-1,
+4,
+1,
+3,
+-2,
+-2,
+-2,
+-1,
+3,
+-5,
+2,
+0,
+1,
+-1,
+-4,
+4,
+0,
+-2,
+4,
+-5,
+3,
+-3,
+-4,
+3,
+-2,
+2,
+-2,
+3,
+3,
+4,
+0,
+-4,
+4,
+0,
+3,
+-5,
+1,
+0,
+0,
+2,
+-3,
+-2,
+-3,
+4,
+-3,
+1,
+-5,
+-5,
+2,
+-3,
+-3,
+-2,
+-4,
+0,
+4,
+1,
+1,
+0,
+4,
+0,
+-3,
+-3,
+-1,
+-2,
+-5,
+4,
+-2,
+1,
+3,
+-1,
+-1,
+-4,
+4,
+-1,
+3,
+-3,
+2,
+2,
+3,
+1,
+2,
+1,
+2,
+4,
+-4,
+-3,
+-2,
+-3,
+4,
+0,
+1,
+0,
+-3,
+0,
+-2,
+-3,
+1,
+3,
+4,
+-1,
+4,
+4,
+3,
+-4,
+-5,
+3,
+2,
+4,
+-4,
+4,
+1,
+-2,
+2,
+-1,
+4,
+2,
+2,
+-5,
+0,
+-2,
+-4,
+2,
+-3,
+-1,
+4,
+0,
+4,
+4,
+1,
+4,
+4,
+-3,
+3,
+4,
+-2,
+4,
+-4,
+2,
+-2,
+-3,
+3,
+1,
+3,
+3,
+0,
+-4,
+1,
+-4,
+4,
+2,
+-2,
+-5,
+-4,
+-4,
+2,
+-1,
+-2,
+0,
+4,
+-3,
+4,
+-5,
+3,
+-1,
+-5,
+-4,
+-2,
+-5,
+4,
+-2,
+3,
+-3,
+4,
+3,
+-4,
+4,
+3,
+-3,
+4,
+-1,
+3,
+-4,
+-5,
+1,
+-3,
+-4,
+2,
+3,
+2,
+-2,
+4,
+3,
+2,
+-4,
+0,
+-2,
+-5,
+-3,
+3,
+1,
+3,
+-5,
+3,
+1,
+-5,
+1,
+-2,
+-4,
+1,
+-5,
+4,
+1,
+0,
+0,
+-4,
+-2,
+2,
+0,
+-3,
+-4,
+-1,
+-5,
+-2,
+-4,
+-2,
+3,
+4,
+-5,
+0,
+2,
+2,
+4,
+0,
+4,
+-5,
+-3,
+-5,
+-5,
+0,
+0,
+-4,
+3,
+1,
+1,
+-5,
+2,
+-5,
+2,
+-4,
+3,
+4,
+-3,
+-4,
+4,
+-2,
+-4,
+4,
+-3,
+-5,
+-4,
+2,
+-2,
+3,
+-2,
+1,
+4,
+1,
+4,
+-2,
+1,
+-4,
+4,
+2,
+-1,
+1,
+1,
+-2,
+-2,
+0,
+-5,
+2,
+4,
+-4,
+1,
+-5,
+0,
+4,
+3,
+-4,
+1,
+-3,
+-4,
+2,
+-3,
+-3,
+1,
+2,
+4,
+-3,
+-2,
+3,
+-3,
+-5,
+4,
+-5,
+-5,
+4,
+4,
+3,
+-1,
+-3,
+-4,
+-1,
+1,
+0,
+1,
+4,
+-1,
+0,
+1,
+-4,
+0,
+-4,
+-3,
+0,
+-4,
+2,
+-2,
+-2,
+-1,
+-2,
+0,
+-5,
+-1,
+-3,
+3,
+4,
+-4,
+4,
+4,
+0,
+-4,
+-3,
+-3,
+4,
+0,
+2,
+3,
+-1,
+2,
+0,
+-4,
+3,
+-1,
+-3,
+-4,
+4,
+0,
+4,
+3,
+0,
+-3,
+-5,
+4,
+-3,
+3,
+4,
+3,
+-1,
+3,
+-3,
+1,
+3,
+4,
+3,
+-1,
+3,
+2,
+4,
+4,
+-2,
+-1,
+2,
+-4,
+-4,
+-1,
+0,
+-1,
+4,
+-4,
+4,
+2,
+0,
+0,
+1,
+-1,
+-5,
+1,
+-5,
+2,
+-2,
+2,
+1,
+3,
+1,
+-3,
+-4,
+-2,
+0,
+4,
+-2,
+0,
+-1,
+-1,
+-3,
+-2,
+4,
+3,
+4,
+2,
+3,
+4,
+-4,
+-2,
+-1,
+-2,
+0,
+-1,
+-1,
+-4,
+-2,
+2,
+-5,
+4,
+-5,
+0,
+4,
+4,
+-2,
+-4,
+1,
+-5,
+-5,
+4,
+-3,
+0,
+-1,
+-5,
+4,
+-4,
+0,
+4,
+4,
+-4,
+-2,
+-4,
+1,
+-4,
+-5,
+4,
+-5,
+-3,
+2,
+-2,
+-1,
+-3,
+0,
+-3,
+-2,
+4,
+-5,
+-5,
+3,
+-5,
+-1,
+3,
+2,
+3,
+0,
+-2,
+-3,
+-4,
+3,
+0,
+3,
+3,
+-1,
+-4,
+0,
+3,
+-5,
+-3,
+4,
+-3,
+2,
+-1,
+0,
+2,
+-1,
+3,
+-1,
+-5,
+3,
+4,
+3,
+3,
+4,
+3,
+1,
+-4,
+1,
+-2,
+4,
+-3,
+2,
+0,
+-2,
+-5,
+-5,
+-2,
+-2,
+0,
+4,
+0,
+-4,
+-1,
+0,
+-5,
+-3,
+4,
+0,
+1,
+-3,
+3,
+-2,
+-2,
+2,
+-1,
+-1,
+-3,
+-1,
+-2,
+-5,
+-5,
+-1,
+-1,
+-4,
+-4,
+};
+
+const short m_b[] = {
+-2,
+3,
+2,
+-5,
+-3,
+2,
+0,
+-5,
+1,
+-3,
+-2,
+-3,
+-3,
+3,
+-4,
+-2,
+-4,
+1,
+1,
+1,
+4,
+4,
+3,
+3,
+-4,
+-3,
+-2,
+-2,
+-3,
+-2,
+-1,
+-5,
+-4,
+1,
+-4,
+-3,
+3,
+-5,
+-5,
+-4,
+-2,
+-2,
+1,
+3,
+3,
+3,
+4,
+0,
+1,
+0,
+4,
+3,
+-3,
+3,
+-1,
+0,
+2,
+-1,
+-1,
+0,
+0,
+3,
+-4,
+4,
+3,
+3,
+-1,
+-4,
+-1,
+-4,
+-2,
+-4,
+-4,
+3,
+0,
+3,
+-4,
+2,
+0,
+0,
+-4,
+-5,
+-4,
+4,
+0,
+-4,
+-2,
+-4,
+3,
+4,
+-4,
+-3,
+-2,
+0,
+-2,
+2,
+-5,
+2,
+1,
+2,
+0,
+0,
+-2,
+-2,
+-1,
+1,
+-1,
+3,
+3,
+-1,
+3,
+-1,
+-2,
+1,
+-3,
+3,
+3,
+3,
+2,
+0,
+1,
+-2,
+1,
+4,
+3,
+-2,
+-5,
+0,
+2,
+0,
+1,
+-3,
+-1,
+3,
+-3,
+-2,
+3,
+4,
+-2,
+-1,
+-5,
+1,
+-1,
+2,
+-3,
+-4,
+0,
+-1,
+0,
+2,
+3,
+4,
+-4,
+-1,
+-1,
+2,
+-1,
+-5,
+1,
+-3,
+0,
+-4,
+0,
+-1,
+2,
+-5,
+-1,
+1,
+-2,
+-5,
+-3,
+-3,
+2,
+2,
+-4,
+2,
+-5,
+-4,
+4,
+4,
+-1,
+1,
+2,
+-1,
+-3,
+0,
+2,
+4,
+-3,
+-2,
+-1,
+4,
+-5,
+4,
+0,
+-1,
+-1,
+-4,
+-3,
+-3,
+-1,
+-2,
+-1,
+-2,
+-5,
+2,
+2,
+-3,
+1,
+-1,
+-2,
+0,
+0,
+1,
+0,
+4,
+-4,
+-4,
+-4,
+-5,
+4,
+-4,
+4,
+-3,
+4,
+-5,
+-2,
+3,
+-2,
+-4,
+1,
+2,
+-5,
+-2,
+-5,
+-3,
+2,
+-1,
+3,
+-4,
+-4,
+4,
+-3,
+-5,
+0,
+-3,
+2,
+2,
+3,
+-3,
+-3,
+-2,
+-5,
+4,
+1,
+3,
+-1,
+-4,
+3,
+1,
+-2,
+1,
+3,
+4,
+1,
+-2,
+0,
+0,
+-3,
+-1,
+1,
+-1,
+0,
+-5,
+-2,
+-1,
+1,
+-2,
+4,
+-3,
+2,
+1,
+0,
+0,
+3,
+-4,
+2,
+1,
+0,
+3,
+-5,
+-1,
+4,
+3,
+4,
+-2,
+-1,
+-3,
+-4,
+3,
+3,
+-4,
+4,
+-5,
+1,
+-1,
+0,
+3,
+4,
+3,
+-3,
+-5,
+-3,
+-2,
+-4,
+1,
+-2,
+-3,
+-2,
+1,
+2,
+4,
+4,
+0,
+2,
+4,
+0,
+-1,
+1,
+-4,
+-1,
+-3,
+-2,
+-5,
+3,
+-3,
+-1,
+0,
+-3,
+-3,
+-4,
+3,
+3,
+-5,
+-2,
+-1,
+3,
+1,
+1,
+1,
+3,
+1,
+1,
+-5,
+-2,
+2,
+4,
+3,
+-3,
+-5,
+3,
+-4,
+0,
+2,
+-4,
+-5,
+-3,
+-3,
+-3,
+2,
+-1,
+-3,
+1,
+-3,
+-1,
+-5,
+2,
+1,
+1,
+-3,
+2,
+4,
+1,
+-5,
+3,
+0,
+-5,
+-4,
+2,
+-5,
+3,
+-1,
+0,
+2,
+-3,
+-1,
+-2,
+-5,
+-4,
+0,
+-5,
+-3,
+3,
+1,
+0,
+-1,
+-4,
+1,
+-4,
+-5,
+-1,
+3,
+3,
+-3,
+-1,
+4,
+-1,
+-5,
+-4,
+-1,
+0,
+1,
+1,
+4,
+-3,
+-2,
+4,
+1,
+-4,
+-1,
+-3,
+2,
+-2,
+4,
+2,
+-1,
+2,
+-4,
+-3,
+4,
+-1,
+1,
+-4,
+0,
+1,
+-4,
+4,
+-4,
+-5,
+-5,
+-3,
+2,
+2,
+-1,
+0,
+-2,
+-5,
+1,
+-1,
+-5,
+2,
+-4,
+-5,
+-3,
+1,
+-4,
+0,
+1,
+-5,
+0,
+-2,
+1,
+-4,
+-5,
+3,
+0,
+1,
+2,
+4,
+-1,
+-4,
+0,
+-4,
+0,
+-5,
+3,
+-5,
+-5,
+2,
+2,
+3,
+-3,
+4,
+4,
+2,
+3,
+-2,
+4,
+3,
+4,
+1,
+1,
+0,
+2,
+2,
+4,
+-1,
+-5,
+-5,
+3,
+-5,
+-2,
+-4,
+-4,
+0,
+4,
+3,
+0,
+4,
+-3,
+4,
+0,
+-1,
+3,
+2,
+-4,
+2,
+-1,
+-3,
+-5,
+-3,
+3,
+-4,
+1,
+-3,
+3,
+-4,
+0,
+3,
+-2,
+-1,
+2,
+-4,
+2,
+4,
+0,
+2,
+-2,
+-5,
+0,
+-1,
+-5,
+4,
+-4,
+3,
+-1,
+4,
+1,
+4,
+0,
+-1,
+1,
+4,
+3,
+3,
+1,
+2,
+-3,
+-2,
+-5,
+-2,
+1,
+-5,
+-4,
+-5,
+-2,
+};
+
+const short m_exp[] = {
+-32,
+48,
+2,
+-56,
+46,
+-52,
+39,
+-77,
+41,
+-50,
+-50,
+-14,
+-45,
+-31,
+3,
+-22,
+-52,
+-31,
+-32,
+8,
+61,
+97,
+49,
+-12,
+12,
+-25,
+-26,
+-15,
+15,
+45,
+-29,
+-27,
+-7,
+-23,
+30,
+-36,
+-56,
+27,
+-59,
+35,
+6,
+44,
+72,
+23,
+17,
+86,
+-8,
+5,
+-25,
+41,
+-30,
+-78,
+120,
+-1,
+69,
+-20,
+24,
+-66,
+6,
+68,
+-59,
+2,
+27,
+-21,
+-71,
+-61,
+-49,
+21,
+-24,
+27,
+-47,
+29,
+71,
+37,
+-9,
+-58,
+13,
+13,
+73,
+-9,
+7,
+28,
+23,
+35,
+46,
+-8,
+86,
+-7,
+25,
+-65,
+-18,
+43,
+58,
+22,
+44,
+-51,
+-33,
+10,
+-18,
+-17,
+-62,
+24,
+18,
+-68,
+23,
+25,
+23,
+-32,
+23,
+37,
+55,
+-16,
+24,
+4,
+-16,
+9,
+77,
+44,
+48,
+-36,
+-81,
+-100,
+-18,
+36,
+-3,
+-72,
+-43,
+38,
+-2,
+34,
+53,
+-42,
+-18,
+56,
+19,
+34,
+3,
+-58,
+-16,
+-34,
+-71,
+-12,
+12,
+-22,
+-9,
+57,
+-46,
+-70,
+25,
+-62,
+10,
+-5,
+68,
+-86,
+-15,
+-19,
+-100,
+78,
+23,
+47,
+5,
+-64,
+31,
+-16,
+-9,
+-22,
+11,
+30,
+-74,
+31,
+50,
+38,
+-59,
+24,
+12,
+-18,
+4,
+33,
+-24,
+-44,
+-19,
+2,
+-47,
+48,
+-30,
+28,
+-38,
+-29,
+20,
+52,
+-21,
+-5,
+49,
+-41,
+-17,
+-2,
+68,
+-2,
+122,
+-29,
+13,
+23,
+46,
+40,
+-12,
+-58,
+59,
+2,
+15,
+-49,
+-38,
+-15,
+46,
+-1,
+-54,
+-73,
+10,
+-1,
+43,
+80,
+-55,
+119,
+82,
+22,
+-12,
+-14,
+-12,
+-3,
+36,
+27,
+-3,
+50,
+59,
+61,
+13,
+-14,
+5,
+51,
+-54,
+-45,
+-9,
+109,
+10,
+-28,
+1,
+0,
+-50,
+-23,
+27,
+49,
+16,
+3,
+23,
+43,
+31,
+29,
+-27,
+-45,
+9,
+61,
+42,
+34,
+30,
+26,
+38,
+-84,
+5,
+98,
+13,
+55,
+15,
+56,
+63,
+22,
+80,
+10,
+52,
+52,
+49,
+101,
+37,
+71,
+-20,
+-38,
+-36,
+-72,
+-40,
+12,
+77,
+-19,
+32,
+9,
+-15,
+85,
+-13,
+-33,
+-8,
+-20,
+33,
+36,
+76,
+6,
+67,
+-51,
+36,
+76,
+-48,
+30,
+42,
+-45,
+-22,
+-9,
+65,
+-29,
+18,
+89,
+34,
+-25,
+41,
+30,
+26,
+49,
+34,
+-6,
+-28,
+10,
+-106,
+6,
+86,
+5,
+-3,
+-105,
+-103,
+-45,
+-55,
+-60,
+29,
+41,
+54,
+-29,
+36,
+-30,
+-8,
+25,
+36,
+40,
+22,
+-35,
+-50,
+33,
+3,
+-19,
+-30,
+3,
+-11,
+9,
+-4,
+-18,
+39,
+92,
+-25,
+0,
+-40,
+64,
+-18,
+58,
+-2,
+27,
+-89,
+32,
+-2,
+56,
+79,
+-67,
+87,
+-41,
+20,
+29,
+4,
+51,
+45,
+26,
+-54,
+8,
+-37,
+-37,
+-39,
+-35,
+-36,
+-23,
+-30,
+11,
+1,
+-63,
+-21,
+5,
+13,
+35,
+-15,
+27,
+-59,
+-10,
+29,
+4,
+-58,
+-56,
+3,
+-10,
+27,
+0,
+29,
+-4,
+-17,
+28,
+2,
+17,
+9,
+-27,
+53,
+-19,
+-36,
+104,
+25,
+38,
+84,
+-29,
+42,
+-10,
+-47,
+-36,
+1,
+10,
+43,
+-46,
+-57,
+38,
+44,
+-9,
+26,
+-15,
+10,
+31,
+34,
+-18,
+17,
+24,
+-3,
+5,
+11,
+4,
+23,
+5,
+-19,
+51,
+-59,
+-29,
+62,
+-27,
+35,
+81,
+65,
+-10,
+-1,
+-1,
+45,
+20,
+26,
+-34,
+-7,
+35,
+-7,
+2,
+19,
+115,
+-32,
+-40,
+-92,
+47,
+-65,
+23,
+53,
+-31,
+22,
+20,
+45,
+91,
+64,
+29,
+49,
+30,
+74,
+-15,
+22,
+120,
+16,
+122,
+-51,
+51,
+78,
+60,
+-53,
+40,
+-49,
+73,
+-2,
+14,
+-17,
+-86,
+35,
+-25,
+36,
+2,
+5,
+38,
+-9,
+16,
+-12,
+-41,
+-24,
+-31,
+-72,
+-68,
+-9,
+5,
+11,
+-63,
+15,
+12,
+21,
+22,
+19,
+62,
+-28,
+122,
+12,
+28,
+-59,
+-19,
+49,
+57,
+-12,
+29,
+-1,
+-23,
+126,
+75,
+-29,
+-21,
+-37,
+8,
+-70,
+-8,
+91,
+-34,
+-10,
+31,
+78,
+44,
+59,
+23,
+66,
+42,
+38,
+6,
+19,
+10,
+1,
+57,
+38,
+71,
+-10,
+-12,
+-7,
+-13,
+-15,
+6,
+};
+
+#define SIZE 24
+__attribute__ ((section(".heapsram"))) short g_mA[SIZE][SIZE];
+__attribute__ ((section(".heapsram"))) short g_mB[SIZE][SIZE];
+__attribute__ ((section(".heapsram"))) short g_mC[SIZE][SIZE];
+__attribute__ ((section(".heapsram"))) short g_mB_tmp[SIZE][SIZE];
diff --git a/carfield/parMatrixMul32/Makefile b/carfield/parMatrixMul32/Makefile
new file mode 100755
index 0000000..bd55e15
--- /dev/null
+++ b/carfield/parMatrixMul32/Makefile
@@ -0,0 +1,8 @@
+PULP_APP = test
+PULP_APP_SRCS = matrixMul.c
+
+PULP_CFLAGS = -O3
+
+include $(PULP_SDK_HOME)/install/rules/pulp.mk
+
+#pulp-bench-reg --name=parMatrixMul32.cycles --module=pulp_rtl_testset --pipeline=$(PIPELINE) --artefact=pulp_rtl_testset --cmd="make run -f Makefile.sdk" --probe-regexp='matrixMul -> success, nr. of errors: 0, execution time: (\d+)' --params="platform($(platformName)),compiler($(OR1K_TOOLCHAIN_TYPE)),nbCores(4),elemSize(32)" --probe-regexp='matrixMulTranspose -> success, nr. of errors: 0, execution time: (\d+)' --params="platform($(platformName)),compiler($(OR1K_TOOLCHAIN_TYPE)),nbCores(4),elemSize(32),transposed"
diff --git a/carfield/parMatrixMul32/gen_stimuli.py b/carfield/parMatrixMul32/gen_stimuli.py
new file mode 100755
index 0000000..32926e9
--- /dev/null
+++ b/carfield/parMatrixMul32/gen_stimuli.py
@@ -0,0 +1,54 @@
+#!/usr/bin/env python
+
+import sys
+import random
+
+
+def write_arr(f, name, arr):
+    f.write('const int %s[] = {\n' % name)
+    for v in arr:
+        f.write('%d,\n' % (v))
+    f.write('};\n\n')
+    return
+
+################################################################################
+f = open('parMatrixMul32_stimuli.h', 'w')
+
+
+SIZE = 24
+RANGE = int(2**15/SIZE)
+
+m_a   = []
+m_b   = []
+m_exp = []
+
+for i in range(0,SIZE):
+    for j in range(0,SIZE):
+        a = random.randint(-RANGE, RANGE-1)
+        b = random.randint(-RANGE, RANGE-1)
+
+        m_a.append(a)
+        m_b.append(b)
+
+for i in range(0,SIZE):
+    for j in range(0,SIZE):
+        r = 0
+
+        for k in range (0,SIZE):
+            r = r + m_a[i * SIZE + k] * m_b[k * SIZE + j]
+
+        m_exp.append(r)
+
+
+write_arr(f, 'm_a',   m_a)
+write_arr(f, 'm_b',   m_b)
+write_arr(f, 'm_exp', m_exp)
+
+f.write('#define SIZE %d\n' % SIZE)
+
+
+f.write('__attribute__ ((section(".heapsram"))) int g_mA[SIZE][SIZE];\n')
+f.write('__attribute__ ((section(".heapsram"))) int g_mB[SIZE][SIZE];\n')
+f.write('__attribute__ ((section(".heapsram"))) int g_mC[SIZE][SIZE];\n')
+f.write('__attribute__ ((section(".heapsram"))) int g_mB_tmp[SIZE][SIZE];\n')
+
diff --git a/carfield/parMatrixMul32/matrixMul.c b/carfield/parMatrixMul32/matrixMul.c
new file mode 100644
index 0000000..990d411
--- /dev/null
+++ b/carfield/parMatrixMul32/matrixMul.c
@@ -0,0 +1,177 @@
+/*
+ * Copyright (C) 2018 ETH Zurich and University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+/* 
+ * Mantainer: Luca Valente, luca.valente2@unibo.it
+ */
+
+#include "pulp.h"
+
+#include "parMatrixMul32_stimuli.h"
+
+void check_matrix_mul(testresult_t *result, void (*start)(), void (*stop)());
+void check_matrix_mul_transpose(testresult_t *result, void (*start)(), void (*stop)());
+
+testcase_t testcases[] = {
+  { .name = "matrixMul",          .test = check_matrix_mul           },
+  { .name = "matrixMulTranspose", .test = check_matrix_mul_transpose },
+  {0, 0}
+};
+
+unsigned int num_cores;
+
+int main()
+{
+  if (rt_cluster_id() != 0)
+    return bench_cluster_forward(0);
+
+  num_cores = get_core_num();
+
+  if(rt_core_id() < num_cores) {
+    run_suite(testcases);
+  }
+
+  synch_barrier();
+
+  return 0;
+}
+
+void matrix_init();
+unsigned int matrix_check();
+
+void check_matrix_mul(testresult_t *result, void (*start)(), void (*stop)()) {
+  int core_id;
+  unsigned int i, j, k;
+  unsigned int chunk;
+  unsigned int lb, ub;
+
+  core_id = get_core_id();
+
+  // number of rows each core has to multiply
+  chunk = SIZE / num_cores;
+  // lower bound
+  lb = core_id * chunk;
+  // upper bound
+  ub = lb + chunk;
+
+  if(core_id == 0) {
+    matrix_init();
+  }
+
+  if(num_cores != 1) synch_barrier();
+
+  // start benchmark
+  start();
+
+  for(i = lb; i < ub; i++) {
+    for(j = 0; j < SIZE; j++) {
+      g_mC[i][j] = 0;
+  
+      for(k = 0; k < SIZE; k++) {
+        g_mC[i][j] += g_mA[i][k] * g_mB[k][j];
+      }
+    }
+  }
+
+  if(num_cores != 1) synch_barrier();
+
+  stop();
+
+  if(core_id == 0) {
+    result->errors = matrix_check();
+  }
+}
+
+void check_matrix_mul_transpose(testresult_t *result, void (*start)(), void (*stop)()) {
+  int core_id;
+  unsigned int i, j, k;
+  unsigned int chunk;
+  unsigned int lb, ub;
+
+  core_id = get_core_id();
+
+  // number of rows each core has to multiply
+  chunk = SIZE / num_cores;
+  // lower bound
+  lb = core_id * chunk;
+  // upper bound
+  ub = lb + chunk;
+
+  if(core_id == 0) {
+    matrix_init();
+  }
+
+  if(num_cores != 1) synch_barrier();
+
+  // start benchmark
+  start();
+
+  // transpose array before using it
+  for(i = lb; i < ub; i++) {
+    for(j = 0; j < SIZE; j++) {
+      g_mB_tmp[i][j] = g_mB[j][i];
+    }
+  }
+
+  if(num_cores != 1) synch_barrier();
+
+  for(i = lb; i < ub; i++) {
+    for(j = 0; j < SIZE; j++) {
+      g_mC[i][j] = 0;
+
+      for(k = 0; k < SIZE; k++) {
+        g_mC[i][j] += g_mA[i][k] * g_mB_tmp[j][k];
+      }
+    }
+  }
+
+  if(num_cores != 1) synch_barrier();
+
+  stop();
+
+  if(core_id == 0) {
+    result->errors = matrix_check();
+  }
+}
+
+void matrix_init() {
+  unsigned int i, j;
+
+  // init, copy to TCDM
+  for(i = 0; i < SIZE; i++) {
+    for(j = 0; j < SIZE; j++) {
+      g_mA[i][j] = m_a[i * SIZE + j];
+      g_mB[i][j] = m_b[i * SIZE + j];
+      g_mC[i][j] = 0;
+    }
+  }
+}
+
+unsigned int matrix_check() {
+  unsigned int errors = 0;
+  unsigned int i, j;
+  // check
+  for(i = 0; i < SIZE; i++) {
+    for(j = 0; j < SIZE; j++) {
+      if(g_mC[i][j] != m_exp[i * SIZE + j]) {
+        printf("At index %d, %d\n", i, j, 0, 0);
+        errors++;
+      }
+    }
+  }
+
+  return errors;
+}
diff --git a/carfield/parMatrixMul32/parMatrixMul32_stimuli.h b/carfield/parMatrixMul32/parMatrixMul32_stimuli.h
new file mode 100644
index 0000000..d5c1af0
--- /dev/null
+++ b/carfield/parMatrixMul32/parMatrixMul32_stimuli.h
@@ -0,0 +1,1742 @@
+const int m_a[] = {
+-1344,
+-778,
+-350,
+1240,
+950,
+940,
+1262,
+285,
+738,
+-37,
+257,
+704,
+87,
+-1135,
+1175,
+960,
+984,
+611,
+-1170,
+1243,
+-729,
+-1235,
+-1229,
+8,
+52,
+1156,
+-804,
+-129,
+-112,
+-307,
+1045,
+-954,
+-944,
+477,
+1104,
+-600,
+-31,
+1364,
+-950,
+-153,
+967,
+-446,
+235,
+-197,
+763,
+-660,
+-1289,
+116,
+-458,
+994,
+1086,
+156,
+-549,
+102,
+-532,
+943,
+299,
+340,
+516,
+1117,
+515,
+1335,
+708,
+-306,
+-589,
+517,
+655,
+-279,
+-595,
+-255,
+-1104,
+-16,
+507,
+199,
+-297,
+-905,
+-1084,
+982,
+20,
+-1364,
+407,
+333,
+-798,
+-711,
+1301,
+175,
+490,
+-165,
+-1097,
+1251,
+1203,
+-884,
+419,
+-1262,
+-950,
+-1200,
+759,
+-205,
+-1365,
+-870,
+109,
+-1135,
+36,
+796,
+-1233,
+-1117,
+-826,
+241,
+588,
+-513,
+-353,
+791,
+1071,
+-1073,
+220,
+-1094,
+1096,
+-723,
+280,
+-505,
+151,
+399,
+319,
+-1120,
+-213,
+-966,
+679,
+497,
+-290,
+-300,
+-290,
+-599,
+243,
+-752,
+604,
+1196,
+-715,
+-177,
+-329,
+1337,
+596,
+1027,
+509,
+-301,
+-1211,
+-936,
+-281,
+446,
+-356,
+841,
+-1123,
+-1343,
+-140,
+-1300,
+-828,
+-237,
+1206,
+1274,
+-1092,
+-922,
+913,
+1201,
+-422,
+782,
+-230,
+633,
+-1055,
+-1160,
+329,
+1255,
+1038,
+770,
+123,
+934,
+670,
+-492,
+-24,
+-779,
+-1129,
+47,
+555,
+1214,
+-232,
+-716,
+-322,
+-126,
+178,
+827,
+710,
+-1057,
+-313,
+996,
+1149,
+-532,
+570,
+1171,
+899,
+-313,
+-790,
+1071,
+154,
+-303,
+492,
+-918,
+-1139,
+292,
+129,
+1347,
+-309,
+751,
+1262,
+142,
+-1062,
+-1305,
+250,
+657,
+238,
+-141,
+1308,
+-37,
+-514,
+-591,
+-611,
+852,
+-653,
+-640,
+91,
+254,
+-1145,
+-1263,
+-838,
+-10,
+266,
+-444,
+1129,
+762,
+-713,
+-326,
+-88,
+1063,
+-442,
+-177,
+365,
+-740,
+-1219,
+1085,
+783,
+-725,
+-1112,
+426,
+660,
+6,
+-440,
+513,
+687,
+1078,
+212,
+-434,
+-953,
+1337,
+160,
+622,
+-950,
+-943,
+288,
+-136,
+-1103,
+-223,
+1271,
+211,
+251,
+-271,
+-26,
+704,
+1177,
+544,
+699,
+-885,
+-864,
+-1280,
+877,
+-461,
+995,
+-623,
+-121,
+-146,
+-484,
+-225,
+-978,
+163,
+-278,
+-502,
+-505,
+-567,
+-771,
+1279,
+699,
+-1337,
+544,
+1145,
+1271,
+640,
+277,
+-164,
+458,
+-1280,
+-602,
+-2,
+1136,
+1203,
+-699,
+-195,
+659,
+-472,
+1230,
+1151,
+-97,
+-77,
+-772,
+-381,
+-295,
+636,
+-1341,
+-445,
+-806,
+531,
+-1186,
+-1313,
+-274,
+835,
+-446,
+558,
+-1307,
+-235,
+43,
+-254,
+-109,
+911,
+-1189,
+559,
+-854,
+-218,
+149,
+580,
+1158,
+-14,
+181,
+-1120,
+-947,
+-542,
+1142,
+631,
+-893,
+-614,
+-257,
+-365,
+-951,
+1,
+-762,
+268,
+382,
+-131,
+808,
+-234,
+839,
+346,
+-733,
+1251,
+496,
+-566,
+-751,
+581,
+-1292,
+1068,
+-932,
+-855,
+1336,
+-280,
+523,
+1294,
+-1251,
+1284,
+-1276,
+87,
+1264,
+-274,
+-922,
+-289,
+-458,
+-117,
+196,
+-79,
+-707,
+1233,
+-385,
+-620,
+-617,
+703,
+-995,
+-374,
+660,
+145,
+821,
+1289,
+582,
+-201,
+447,
+116,
+759,
+-615,
+834,
+268,
+-1114,
+-1016,
+-227,
+-589,
+-910,
+-244,
+-660,
+764,
+219,
+1165,
+506,
+-673,
+799,
+-1355,
+-872,
+491,
+689,
+176,
+-285,
+1151,
+1080,
+-319,
+286,
+833,
+217,
+-621,
+478,
+539,
+-109,
+-1273,
+-564,
+-240,
+504,
+518,
+256,
+-124,
+74,
+949,
+-912,
+-1341,
+965,
+-774,
+634,
+1009,
+1304,
+200,
+-1041,
+-1262,
+-865,
+-1065,
+-635,
+-357,
+-928,
+806,
+1148,
+-411,
+56,
+686,
+-644,
+1241,
+-430,
+297,
+127,
+457,
+-1313,
+741,
+861,
+220,
+-540,
+772,
+265,
+1066,
+679,
+177,
+-734,
+29,
+-149,
+181,
+-1042,
+-1139,
+271,
+-326,
+-29,
+1298,
+643,
+-890,
+-136,
+-1015,
+-565,
+-964,
+894,
+-312,
+698,
+159,
+-222,
+-1322,
+578,
+945,
+1124,
+1278,
+54,
+-389,
+1101,
+362,
+-543,
+380,
+959,
+-399,
+-1105,
+1308,
+338,
+-198,
+-1111,
+-278,
+-752,
+668,
+1156,
+-1226,
+579,
+184,
+-1084,
+-917,
+-498,
+-466,
+316,
+-788,
+-718,
+468,
+367,
+-1333,
+-1146,
+828,
+1329,
+311,
+-1346,
+54,
+-976,
+854,
+-658,
+-198,
+-979,
+156,
+385,
+-659,
+1326,
+1351,
+-1173,
+-648,
+720,
+-40,
+313,
+729,
+-416,
+351,
+452,
+-413,
+-4,
+-1113,
+-612,
+-28,
+-721,
+400,
+1072,
+-1010,
+};
+
+const int m_b[] = {
+-1316,
+319,
+963,
+-608,
+519,
+-783,
+-676,
+181,
+172,
+203,
+-1351,
+-935,
+-12,
+758,
+-746,
+1226,
+127,
+-1346,
+1251,
+-377,
+889,
+-23,
+-417,
+-122,
+680,
+1363,
+729,
+-907,
+-1263,
+-431,
+363,
+1355,
+-566,
+-517,
+-1186,
+1318,
+-1104,
+-1245,
+950,
+687,
+252,
+-270,
+1081,
+-1290,
+656,
+8,
+60,
+1171,
+915,
+-500,
+678,
+-953,
+307,
+-35,
+-1334,
+-888,
+598,
+1160,
+722,
+850,
+-268,
+988,
+635,
+-340,
+252,
+1208,
+420,
+82,
+1283,
+-319,
+-666,
+172,
+583,
+174,
+471,
+-1063,
+452,
+-191,
+-1188,
+116,
+-927,
+1086,
+119,
+-245,
+-717,
+-657,
+417,
+319,
+1133,
+1338,
+141,
+-546,
+567,
+-1089,
+-191,
+-1138,
+-201,
+-1286,
+-820,
+-1356,
+1177,
+-317,
+191,
+67,
+164,
+-306,
+-1015,
+1147,
+-482,
+1229,
+-259,
+-207,
+1309,
+847,
+-399,
+-1005,
+-995,
+140,
+-567,
+-1220,
+-427,
+180,
+-571,
+997,
+-783,
+-316,
+-1360,
+736,
+75,
+-1251,
+-307,
+-902,
+1181,
+1057,
+-141,
+-1098,
+776,
+1096,
+-923,
+914,
+1049,
+-28,
+-742,
+-804,
+-467,
+567,
+329,
+-309,
+-161,
+-157,
+-430,
+-639,
+1138,
+-165,
+292,
+-20,
+777,
+-715,
+60,
+-1359,
+35,
+307,
+-1092,
+271,
+548,
+822,
+-50,
+-475,
+-103,
+784,
+537,
+152,
+-517,
+-1097,
+117,
+-619,
+538,
+941,
+172,
+-223,
+1161,
+-1004,
+-1145,
+-455,
+255,
+363,
+859,
+403,
+-861,
+-657,
+-537,
+-1084,
+-1042,
+541,
+1283,
+-356,
+1298,
+-1254,
+-303,
+203,
+104,
+1123,
+-72,
+-171,
+-1122,
+-533,
+440,
+275,
+613,
+-846,
+-189,
+884,
+704,
+-570,
+-440,
+-1157,
+-200,
+-80,
+616,
+799,
+757,
+-264,
+-1256,
+-690,
+152,
+184,
+-810,
+-221,
+-821,
+-243,
+508,
+-709,
+574,
+-693,
+315,
+-952,
+952,
+697,
+875,
+-480,
+-691,
+422,
+-413,
+-1199,
+441,
+-751,
+821,
+1303,
+-410,
+416,
+566,
+-131,
+-551,
+46,
+978,
+-228,
+1117,
+-251,
+-537,
+874,
+-882,
+260,
+-213,
+248,
+-1296,
+1343,
+-626,
+-812,
+629,
+-601,
+-378,
+-1314,
+-889,
+774,
+-307,
+692,
+-1125,
+-692,
+923,
+947,
+1158,
+-939,
+1284,
+35,
+1299,
+369,
+-8,
+43,
+768,
+524,
+137,
+659,
+285,
+-1315,
+-457,
+871,
+-768,
+1107,
+-695,
+488,
+-527,
+-161,
+414,
+-526,
+-1164,
+1059,
+-1108,
+560,
+-622,
+898,
+-50,
+-286,
+-170,
+513,
+952,
+433,
+237,
+584,
+-665,
+-960,
+585,
+-434,
+1223,
+-130,
+1035,
+430,
+202,
+1312,
+1152,
+1059,
+-1082,
+-1295,
+805,
+-18,
+613,
+-94,
+557,
+548,
+1354,
+116,
+289,
+-1358,
+-1234,
+1237,
+451,
+820,
+-102,
+974,
+832,
+-1019,
+914,
+-512,
+-267,
+1329,
+-910,
+-1341,
+862,
+-381,
+-23,
+-658,
+40,
+-71,
+-782,
+1240,
+-956,
+1241,
+-291,
+-884,
+1250,
+699,
+834,
+190,
+960,
+-1260,
+177,
+464,
+155,
+-1105,
+768,
+424,
+621,
+740,
+-1357,
+1186,
+-594,
+1329,
+829,
+126,
+1101,
+1146,
+-95,
+605,
+-673,
+1334,
+440,
+-10,
+12,
+-745,
+20,
+19,
+-793,
+999,
+1083,
+487,
+-657,
+-356,
+654,
+-326,
+-250,
+-718,
+-947,
+-235,
+558,
+974,
+-981,
+637,
+-861,
+-768,
+1045,
+-583,
+-910,
+128,
+734,
+896,
+-1156,
+223,
+284,
+272,
+634,
+-473,
+363,
+359,
+-1185,
+14,
+-33,
+-1122,
+140,
+900,
+439,
+-944,
+-770,
+663,
+865,
+1056,
+-238,
+86,
+-1294,
+-44,
+-603,
+602,
+20,
+397,
+-423,
+-703,
+-209,
+-906,
+-1236,
+945,
+-737,
+578,
+904,
+645,
+1225,
+-877,
+-425,
+-493,
+-1326,
+424,
+965,
+1300,
+-1210,
+823,
+1345,
+626,
+-427,
+592,
+-869,
+-1055,
+-938,
+-427,
+1066,
+472,
+-1055,
+48,
+-1200,
+-349,
+313,
+-1227,
+-228,
+783,
+839,
+187,
+1021,
+-1355,
+1284,
+68,
+-1321,
+-997,
+1286,
+-887,
+772,
+-156,
+-105,
+1329,
+1141,
+-377,
+-881,
+-341,
+1316,
+-391,
+-1249,
+-205,
+53,
+-266,
+-540,
+-289,
+-1011,
+602,
+-1032,
+-1097,
+-202,
+-467,
+-1047,
+-867,
+-340,
+-109,
+-496,
+967,
+1147,
+108,
+384,
+-12,
+1216,
+137,
+1318,
+151,
+219,
+-543,
+391,
+668,
+-1348,
+-1244,
+-810,
+-676,
+321,
+-1258,
+1343,
+1214,
+791,
+35,
+1219,
+1278,
+1037,
+-1282,
+661,
+585,
+921,
+-880,
+-989,
+-1192,
+-207,
+273,
+-382,
+690,
+165,
+271,
+-212,
+739,
+-343,
+-42,
+226,
+40,
+859,
+-153,
+622,
+-1059,
+};
+
+const int m_exp[] = {
+-4942391,
+2289133,
+-1363225,
+1978230,
+1580032,
+-625813,
+-3230128,
+2236653,
+19494,
+3242695,
+-1080745,
+-34154,
+4086860,
+-1370876,
+3997221,
+-1812380,
+4705498,
+7690207,
+-4068140,
+3595067,
+-1103308,
+-939857,
+-4249710,
+-8650816,
+-2013119,
+2933624,
+235162,
+-453807,
+-4447391,
+3527041,
+2046492,
+411956,
+-994117,
+-1411344,
+1333704,
+-519761,
+3026373,
+-564969,
+3749147,
+2447173,
+-557628,
+1138674,
+-1426096,
+-4033488,
+-1829685,
+2815607,
+2382958,
+1714081,
+-1470484,
+3379876,
+3660759,
+-2439960,
+-1180478,
+-3300785,
+-5104533,
+-309753,
+-1667400,
+3258850,
+1805449,
+2481948,
+-944985,
+-363123,
+4227063,
+3022289,
+2763211,
+5114077,
+-1534394,
+-2957168,
+3401637,
+-1195822,
+-747480,
+-2915318,
+-2505013,
+-174927,
+276733,
+2899369,
+-6702856,
+923396,
+-2741169,
+4270685,
+-1020657,
+-2562887,
+2074098,
+-2382784,
+1366504,
+691209,
+4127820,
+400356,
+-7415505,
+823772,
+-3848400,
+-158560,
+3759990,
+2298445,
+323394,
+873625,
+474364,
+2617120,
+-1382444,
+1735284,
+-5799715,
+1915577,
+7016057,
+-1212904,
+919286,
+2949768,
+1228832,
+572192,
+4145710,
+-6809520,
+-2199597,
+677764,
+-7169579,
+-4904277,
+6902014,
+231123,
+-4797299,
+3093608,
+989455,
+4324476,
+3121268,
+810907,
+-2457323,
+2299211,
+-1625774,
+-141013,
+3343022,
+-2044657,
+4089375,
+-291323,
+-1950307,
+2480885,
+2846731,
+-2139146,
+-2718414,
+-1997531,
+-2399245,
+-4060224,
+423228,
+-205276,
+-1602384,
+910872,
+-3535114,
+6008729,
+-559984,
+-301205,
+-5407307,
+2981269,
+1079061,
+6602535,
+-857708,
+2756391,
+-5304566,
+-3769267,
+2620777,
+-4409088,
+502077,
+7568647,
+-37918,
+2315061,
+-2540065,
+8296540,
+-7465282,
+1553910,
+-4736227,
+-2139045,
+95614,
+342546,
+1309722,
+1777391,
+2981296,
+-736899,
+-2572111,
+-972463,
+-2793724,
+-2893912,
+-1230264,
+-871649,
+-1439985,
+3074445,
+3339004,
+2423842,
+1751086,
+4064832,
+1550243,
+6225792,
+-503750,
+-2567772,
+5671219,
+-2054796,
+-551487,
+-2787790,
+-3835027,
+-4272806,
+2725813,
+-2982521,
+1803437,
+3024675,
+-201092,
+-1626608,
+1548043,
+2303810,
+3032912,
+398283,
+1704371,
+1860306,
+-4101665,
+5187913,
+4233418,
+2054883,
+3603470,
+1935132,
+-17548,
+-4362444,
+-2806918,
+-5651039,
+-1853372,
+-1707208,
+-153048,
+-2791834,
+402265,
+2815962,
+3391662,
+-4833520,
+-1190520,
+2302448,
+80738,
+2089586,
+174096,
+2837490,
+-5514606,
+2138871,
+336249,
+-378675,
+5833977,
+4367000,
+-2445147,
+-3652299,
+-1794451,
+-1471577,
+-1263012,
+3719274,
+-3404819,
+5765304,
+-4256415,
+3558206,
+-1884441,
+-475244,
+3659623,
+-2914867,
+689238,
+-2576754,
+7739914,
+1823902,
+2077002,
+-2365242,
+2023481,
+1663749,
+-4973435,
+-694558,
+1118078,
+2260786,
+-3256285,
+4596746,
+-5421599,
+594942,
+-1730692,
+-4626077,
+3077882,
+-2232009,
+2672161,
+3135747,
+-4602601,
+859784,
+-3530668,
+21600,
+-4690786,
+2023164,
+-496745,
+-2728919,
+281474,
+-108745,
+-809613,
+1445687,
+-5781458,
+2097169,
+1594266,
+-4504019,
+2460482,
+6259537,
+-700848,
+413263,
+-1212884,
+-5695130,
+2094147,
+-750529,
+-1379008,
+6029072,
+722889,
+1719449,
+1682336,
+-4219755,
+1971162,
+66575,
+-1195119,
+141466,
+-1083536,
+-718558,
+-4039954,
+-168429,
+-2026861,
+2025800,
+-761083,
+-4194692,
+2013337,
+-1156936,
+3823019,
+4081732,
+-3145845,
+-1733615,
+-1371947,
+-3811245,
+-1584663,
+-3547009,
+-3267886,
+8255291,
+-3232160,
+3404636,
+3248369,
+3233853,
+671601,
+-1009897,
+1821121,
+-3517645,
+2005444,
+-2768741,
+115998,
+533867,
+4717709,
+1315923,
+-3510545,
+-3539595,
+-538461,
+4529529,
+2792584,
+-107486,
+-1840413,
+-1474849,
+1579605,
+-4197602,
+-34825,
+-462678,
+1294881,
+-1730927,
+-2549709,
+-1531672,
+-271859,
+-1181904,
+-1680154,
+-2321723,
+-6641222,
+1127764,
+893535,
+-2804646,
+5653509,
+2657606,
+-1751466,
+-4669812,
+-827592,
+-126901,
+-2599752,
+-845148,
+1390838,
+8975481,
+-7663778,
+3572438,
+5920790,
+5233883,
+-613590,
+-881500,
+-3974422,
+-5523348,
+-3243204,
+-6405765,
+-4376438,
+1352634,
+-105650,
+2650174,
+1442151,
+5088231,
+2974595,
+-4501663,
+-841006,
+-3101819,
+-1265401,
+-2756903,
+2579743,
+2045040,
+-5328835,
+2801176,
+-386694,
+-3068782,
+3147225,
+-248211,
+-662659,
+-1112717,
+2733193,
+336344,
+3107302,
+2244003,
+4285762,
+1998904,
+1888720,
+-1174981,
+-2567532,
+-5588952,
+-101948,
+4004848,
+-610048,
+793760,
+3345423,
+716318,
+1033698,
+4011882,
+-965219,
+1258434,
+1579522,
+-4249500,
+3233648,
+-424838,
+2640541,
+1020028,
+4933599,
+-1964947,
+3237309,
+-1251962,
+-437406,
+-2749192,
+-2943112,
+-117113,
+778507,
+2757711,
+3478291,
+-661571,
+1077087,
+-3821174,
+2731860,
+3035264,
+-4424379,
+295413,
+3873542,
+-1272809,
+4145370,
+-363272,
+2240544,
+88954,
+-2016552,
+-862779,
+-844808,
+3142493,
+2019692,
+3648148,
+3857820,
+593190,
+1285134,
+-4257140,
+-1476035,
+-1951773,
+-2334649,
+1355368,
+-4390456,
+3666652,
+562848,
+-8226958,
+1134896,
+1136697,
+-2132899,
+3300228,
+1855661,
+6476864,
+5097743,
+-1373818,
+3287769,
+1709294,
+-2926119,
+2463141,
+400199,
+3051372,
+1815531,
+1746372,
+398117,
+2333959,
+-708565,
+-4241370,
+51697,
+-1626217,
+26865,
+2248300,
+3357859,
+-325912,
+194201,
+612298,
+388227,
+256018,
+-5630155,
+-1085451,
+653494,
+-1966315,
+-273079,
+-4296295,
+-2813232,
+2079672,
+2378463,
+-3869089,
+-438799,
+-725265,
+3152791,
+3461913,
+-777750,
+47521,
+2588203,
+1888001,
+-4445421,
+654349,
+811737,
+418334,
+-1854075,
+-5194402,
+-1571674,
+-622026,
+-1091628,
+1787463,
+3439585,
+2923276,
+-1997884,
+-193963,
+-731696,
+3686658,
+-1311796,
+-5219031,
+-2906251,
+2140229,
+-1846978,
+2541247,
+-3677377,
+-3935140,
+3605308,
+4807232,
+-1633864,
+344286,
+-2051894,
+2498349,
+-3085,
+379207,
+-701595,
+-1080351,
+3161365,
+-1606976,
+1640595,
+3757649,
+798095,
+-3167055,
+-2288739,
+2301831,
+-3324819,
+3219538,
+516049,
+-3153835,
+7342606,
+1098913,
+-2522436,
+376783,
+47367,
+530901,
+-395499,
+304200,
+};
+
+#define SIZE 24
+__attribute__ ((section(".heapsram"))) int g_mA[SIZE][SIZE];
+__attribute__ ((section(".heapsram"))) int g_mB[SIZE][SIZE];
+__attribute__ ((section(".heapsram"))) int g_mC[SIZE][SIZE];
+__attribute__ ((section(".heapsram"))) int g_mB_tmp[SIZE][SIZE];
diff --git a/carfield/parMatrixMul8/Makefile b/carfield/parMatrixMul8/Makefile
new file mode 100755
index 0000000..7c755b8
--- /dev/null
+++ b/carfield/parMatrixMul8/Makefile
@@ -0,0 +1,8 @@
+PULP_APP = test
+PULP_APP_SRCS = matrixMul.c
+
+PULP_CFLAGS = -O3
+
+include $(PULP_SDK_HOME)/install/rules/pulp.mk
+
+#pulp-bench-reg --name=parMatrixMul8.cycles --module=pulp_rtl_testset --pipeline=$(PIPELINE) --artefact=pulp_rtl_testset --cmd="make run -f Makefile.sdk" --probe-regexp='matrixMul -> success, nr. of errors: 0, execution time: (\d+)' --params="platform($(platformName)),compiler($(OR1K_TOOLCHAIN_TYPE)),nbCores(4),elemSize(8)" --probe-regexp='matrixMulTranspose -> success, nr. of errors: 0, execution time: (\d+)' --params="platform($(platformName)),compiler($(OR1K_TOOLCHAIN_TYPE)),nbCores(4),elemSize(8),transposed"
diff --git a/carfield/parMatrixMul8/gen_stimuli.py b/carfield/parMatrixMul8/gen_stimuli.py
new file mode 100755
index 0000000..153d5c3
--- /dev/null
+++ b/carfield/parMatrixMul8/gen_stimuli.py
@@ -0,0 +1,54 @@
+#!/usr/bin/env python
+
+import sys
+import random
+
+
+def write_arr(f, name, arr):
+    f.write('const char %s[] = {\n' % name)
+    for v in arr:
+        f.write('%d,\n' % (v))
+    f.write('};\n\n')
+    return
+
+################################################################################
+f = open('parMatrixMul8_stimuli.h', 'w')
+
+
+SIZE = 24 
+RANGE = 4
+
+m_a   = []
+m_b   = []
+m_exp = []
+
+for i in range(0,SIZE):
+    for j in range(0,SIZE):
+        a = random.randint(-RANGE, RANGE-1)
+        b = random.randint(-RANGE, RANGE-1)
+
+        m_a.append(a)
+        m_b.append(b)
+
+for i in range(0,SIZE):
+    for j in range(0,SIZE):
+        r = 0
+
+        for k in range (0,SIZE):
+            r = r + m_a[i * SIZE + k] * m_b[k * SIZE + j]
+
+        m_exp.append(r)
+
+
+write_arr(f, 'm_a',   m_a)
+write_arr(f, 'm_b',   m_b)
+write_arr(f, 'm_exp', m_exp)
+
+f.write('#define SIZE %d\n' % SIZE)
+
+
+f.write('__attribute__ ((section(".heapsram"))) char g_mA[SIZE][SIZE];\n')
+f.write('__attribute__ ((section(".heapsram"))) char g_mB[SIZE][SIZE];\n')
+f.write('__attribute__ ((section(".heapsram"))) char g_mC[SIZE][SIZE];\n')
+f.write('__attribute__ ((section(".heapsram"))) char g_mB_tmp[SIZE][SIZE];\n')
+
diff --git a/carfield/parMatrixMul8/matrixMul.c b/carfield/parMatrixMul8/matrixMul.c
new file mode 100644
index 0000000..357fdf0
--- /dev/null
+++ b/carfield/parMatrixMul8/matrixMul.c
@@ -0,0 +1,177 @@
+/*
+ * Copyright (C) 2018 ETH Zurich and University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+/* 
+ * Mantainer: Luca Valente, luca.valente2@unibo.it
+ */
+
+#include "pulp.h"
+
+#include "parMatrixMul8_stimuli.h"
+
+void check_matrix_mul(testresult_t *result, void (*start)(), void (*stop)());
+void check_matrix_mul_transpose(testresult_t *result, void (*start)(), void (*stop)());
+
+testcase_t testcases[] = {
+  { .name = "matrixMul",          .test = check_matrix_mul           },
+  { .name = "matrixMulTranspose", .test = check_matrix_mul_transpose },
+  {0, 0}
+};
+
+unsigned int num_cores;
+
+int main()
+{
+  if (rt_cluster_id() != 0)
+    return bench_cluster_forward(0);
+
+  num_cores = get_core_num();
+
+  if(rt_core_id() < num_cores) {
+    run_suite(testcases);
+  }
+
+  synch_barrier();
+
+  return 0;
+}
+
+void matrix_init();
+unsigned int matrix_check();
+
+void check_matrix_mul(testresult_t *result, void (*start)(), void (*stop)()) {
+  int core_id;
+  unsigned int i, j, k;
+  unsigned int chunk;
+  unsigned int lb, ub;
+
+  core_id = get_core_id();
+
+  // number of rows each core has to multiply
+  chunk = SIZE / num_cores;
+  // lower bound
+  lb = core_id * chunk;
+  // upper bound
+  ub = lb + chunk;
+
+  if(core_id == 0) {
+    matrix_init();
+  }
+
+  if(num_cores != 1) synch_barrier();
+
+  // start benchmark
+  start();
+
+  for(i = lb; i < ub; i++) {
+    for(j = 0; j < SIZE; j++) {
+      g_mC[i][j] = 0;
+  
+      for(k = 0; k < SIZE; k++) {
+        g_mC[i][j] += g_mA[i][k] * g_mB[k][j];
+      }
+    }
+  }
+
+  if(num_cores != 1) synch_barrier();
+
+  stop();
+
+  if(core_id == 0) {
+    result->errors = matrix_check();
+  }
+}
+
+void check_matrix_mul_transpose(testresult_t *result, void (*start)(), void (*stop)()) {
+  int core_id;
+  unsigned int i, j, k;
+  unsigned int chunk;
+  unsigned int lb, ub;
+
+  core_id = get_core_id();
+
+  // number of rows each core has to multiply
+  chunk = SIZE / num_cores;
+  // lower bound
+  lb = core_id * chunk;
+  // upper bound
+  ub = lb + chunk;
+
+  if(core_id == 0) {
+    matrix_init();
+  }
+
+  if(num_cores != 1) synch_barrier();
+
+  // start benchmark
+  start();
+
+  // transpose array before using it
+  for(i = lb; i < ub; i++) {
+    for(j = 0; j < SIZE; j++) {
+      g_mB_tmp[i][j] = g_mB[j][i];
+    }
+  }
+
+  if(num_cores != 1) synch_barrier();
+
+  for(i = lb; i < ub; i++) {
+    for(j = 0; j < SIZE; j++) {
+      g_mC[i][j] = 0;
+
+      for(k = 0; k < SIZE; k++) {
+        g_mC[i][j] += g_mA[i][k] * g_mB_tmp[j][k];
+      }
+    }
+  }
+
+  if(num_cores != 1) synch_barrier();
+
+  stop();
+
+  if(core_id == 0) {
+    result->errors = matrix_check();
+  }
+}
+
+void matrix_init() {
+  unsigned int i, j;
+
+  // init, copy to TCDM
+  for(i = 0; i < SIZE; i++) {
+    for(j = 0; j < SIZE; j++) {
+      g_mA[i][j] = m_a[i * SIZE + j];
+      g_mB[i][j] = m_b[i * SIZE + j];
+      g_mC[i][j] = 0;
+    }
+  }
+}
+
+unsigned int matrix_check() {
+  unsigned int errors = 0;
+  unsigned int i, j;
+  // check
+  for(i = 0; i < SIZE; i++) {
+    for(j = 0; j < SIZE; j++) {
+      if(g_mC[i][j] != m_exp[i * SIZE + j]) {
+        printf("At index %d, %d\n", i, j, 0, 0);
+        errors++;
+      }
+    }
+  }
+
+  return errors;
+}
diff --git a/carfield/parMatrixMul8/parMatrixMul8_stimuli.h b/carfield/parMatrixMul8/parMatrixMul8_stimuli.h
new file mode 100644
index 0000000..78d026c
--- /dev/null
+++ b/carfield/parMatrixMul8/parMatrixMul8_stimuli.h
@@ -0,0 +1,1742 @@
+const char m_a[] = {
+-1,
+0,
+0,
+0,
+-4,
+-4,
+-4,
+1,
+-1,
+-2,
+-3,
+3,
+2,
+-4,
+1,
+-2,
+-3,
+0,
+-3,
+-4,
+-3,
+-1,
+-3,
+-3,
+2,
+-2,
+2,
+3,
+0,
+2,
+0,
+-1,
+-3,
+-2,
+-1,
+-3,
+1,
+-1,
+-1,
+0,
+-1,
+-2,
+-1,
+-4,
+-2,
+-3,
+-1,
+-2,
+1,
+1,
+2,
+-1,
+0,
+2,
+1,
+-3,
+2,
+-3,
+0,
+0,
+0,
+2,
+-2,
+3,
+-4,
+2,
+-2,
+-3,
+-2,
+-3,
+3,
+-3,
+0,
+1,
+-1,
+-1,
+1,
+-2,
+-1,
+-4,
+-2,
+-1,
+-4,
+-4,
+-2,
+-1,
+-2,
+-3,
+-4,
+-4,
+1,
+0,
+-2,
+1,
+-1,
+1,
+2,
+-3,
+-4,
+0,
+2,
+-3,
+-3,
+0,
+0,
+-2,
+-3,
+-1,
+-2,
+-1,
+-4,
+-1,
+-4,
+-3,
+-3,
+0,
+1,
+-3,
+-4,
+1,
+2,
+0,
+2,
+-1,
+-4,
+-1,
+-4,
+2,
+2,
+-1,
+3,
+0,
+0,
+-1,
+1,
+3,
+0,
+2,
+0,
+3,
+2,
+2,
+2,
+-3,
+2,
+2,
+-4,
+0,
+-3,
+-1,
+-1,
+-3,
+-4,
+3,
+2,
+0,
+2,
+1,
+0,
+-2,
+-1,
+3,
+3,
+0,
+1,
+-3,
+3,
+0,
+-3,
+2,
+-3,
+1,
+2,
+0,
+3,
+-1,
+0,
+-2,
+-4,
+0,
+-3,
+0,
+3,
+0,
+-1,
+-3,
+0,
+-2,
+-4,
+-4,
+-1,
+1,
+2,
+-4,
+1,
+-4,
+-4,
+-3,
+-3,
+3,
+-1,
+-2,
+-4,
+2,
+1,
+2,
+-1,
+1,
+3,
+0,
+2,
+-4,
+-3,
+1,
+-2,
+0,
+1,
+1,
+3,
+-2,
+-4,
+2,
+-1,
+-1,
+1,
+2,
+-2,
+1,
+1,
+2,
+0,
+3,
+-4,
+2,
+-1,
+-1,
+-3,
+-3,
+-3,
+-3,
+2,
+-3,
+1,
+0,
+3,
+-3,
+3,
+3,
+-4,
+1,
+-4,
+1,
+2,
+-3,
+1,
+-4,
+-1,
+-4,
+-3,
+-2,
+2,
+3,
+0,
+-4,
+1,
+3,
+0,
+3,
+-1,
+-4,
+3,
+-2,
+-2,
+2,
+-4,
+-1,
+-1,
+3,
+-2,
+-4,
+-1,
+-1,
+2,
+-1,
+2,
+0,
+-2,
+-4,
+0,
+2,
+-4,
+0,
+1,
+-4,
+2,
+0,
+-1,
+-1,
+-3,
+2,
+3,
+2,
+-2,
+1,
+-4,
+-3,
+1,
+2,
+-1,
+-4,
+3,
+-1,
+2,
+-2,
+-3,
+0,
+-1,
+3,
+-4,
+-3,
+-3,
+-3,
+0,
+0,
+3,
+1,
+-2,
+-2,
+-4,
+2,
+2,
+0,
+3,
+-2,
+1,
+3,
+-1,
+3,
+-4,
+-2,
+0,
+-4,
+0,
+-1,
+-3,
+-2,
+-4,
+-4,
+-2,
+-4,
+-3,
+-3,
+0,
+2,
+-4,
+1,
+0,
+-3,
+0,
+-4,
+-4,
+3,
+3,
+-2,
+1,
+3,
+3,
+1,
+-3,
+-4,
+-4,
+1,
+1,
+2,
+-4,
+0,
+3,
+-1,
+3,
+1,
+-4,
+-2,
+1,
+-1,
+3,
+3,
+-1,
+-1,
+0,
+-2,
+2,
+-2,
+-2,
+1,
+3,
+1,
+3,
+-1,
+2,
+-1,
+0,
+1,
+2,
+-4,
+-3,
+1,
+-3,
+-1,
+-4,
+-1,
+-4,
+-4,
+-3,
+-3,
+3,
+0,
+3,
+0,
+-2,
+3,
+-3,
+3,
+-1,
+-1,
+2,
+-2,
+1,
+-4,
+0,
+-3,
+3,
+0,
+-2,
+-1,
+3,
+2,
+-2,
+3,
+-3,
+0,
+-2,
+1,
+0,
+-1,
+1,
+3,
+-4,
+3,
+0,
+3,
+-4,
+3,
+-1,
+-1,
+1,
+3,
+1,
+0,
+-4,
+3,
+-3,
+-4,
+-1,
+2,
+-3,
+-4,
+-2,
+-3,
+2,
+3,
+-2,
+-2,
+-4,
+-1,
+1,
+-4,
+-2,
+2,
+-3,
+1,
+-3,
+-2,
+2,
+3,
+-2,
+0,
+1,
+-1,
+2,
+-2,
+1,
+-1,
+-4,
+-3,
+-4,
+-3,
+-3,
+-3,
+-4,
+3,
+1,
+1,
+3,
+-4,
+0,
+-4,
+2,
+2,
+3,
+2,
+-4,
+1,
+-1,
+-2,
+-2,
+2,
+-2,
+2,
+-3,
+2,
+0,
+0,
+0,
+-4,
+-2,
+-2,
+2,
+1,
+0,
+2,
+3,
+2,
+3,
+-3,
+0,
+-3,
+-2,
+2,
+-1,
+2,
+-3,
+1,
+0,
+1,
+3,
+-4,
+0,
+-3,
+-1,
+1,
+-2,
+-2,
+0,
+0,
+1,
+-3,
+-2,
+0,
+3,
+2,
+0,
+3,
+-4,
+2,
+1,
+3,
+-2,
+-2,
+-3,
+2,
+2,
+2,
+-1,
+3,
+2,
+-3,
+};
+
+const char m_b[] = {
+3,
+-2,
+-3,
+-4,
+-3,
+2,
+-4,
+-1,
+-4,
+-4,
+-1,
+-3,
+3,
+0,
+2,
+1,
+2,
+-2,
+0,
+3,
+-4,
+-2,
+2,
+-1,
+1,
+-4,
+-4,
+-1,
+1,
+-3,
+-2,
+1,
+-3,
+1,
+-2,
+2,
+1,
+-3,
+-1,
+-4,
+1,
+-1,
+-3,
+3,
+0,
+0,
+-2,
+0,
+-4,
+-3,
+0,
+-3,
+0,
+3,
+-1,
+-4,
+-2,
+2,
+-4,
+-1,
+-3,
+1,
+-4,
+-4,
+0,
+2,
+3,
+2,
+0,
+-3,
+-1,
+-1,
+2,
+2,
+-2,
+-2,
+-2,
+-4,
+3,
+-3,
+2,
+1,
+-1,
+1,
+-3,
+-1,
+2,
+-1,
+1,
+-1,
+0,
+-4,
+-2,
+3,
+0,
+-4,
+-2,
+3,
+-3,
+1,
+-1,
+-1,
+-2,
+1,
+-2,
+-4,
+0,
+0,
+0,
+0,
+-4,
+-3,
+3,
+-2,
+3,
+3,
+2,
+3,
+-4,
+1,
+-2,
+-3,
+-3,
+2,
+3,
+-4,
+-1,
+3,
+0,
+3,
+2,
+1,
+-1,
+-1,
+-4,
+-1,
+3,
+2,
+-1,
+3,
+0,
+3,
+0,
+1,
+-4,
+0,
+3,
+2,
+-4,
+-2,
+-2,
+-4,
+-3,
+-2,
+1,
+0,
+-2,
+1,
+3,
+-2,
+-1,
+-3,
+0,
+3,
+0,
+0,
+0,
+2,
+3,
+-4,
+-4,
+2,
+2,
+3,
+1,
+0,
+0,
+-2,
+1,
+-2,
+-4,
+-3,
+0,
+-2,
+2,
+3,
+2,
+3,
+3,
+1,
+1,
+1,
+3,
+1,
+3,
+2,
+-3,
+1,
+-1,
+-2,
+-2,
+-2,
+-1,
+2,
+2,
+0,
+-3,
+-1,
+2,
+-4,
+-2,
+2,
+-3,
+-2,
+-4,
+3,
+-1,
+3,
+2,
+3,
+-1,
+-1,
+-1,
+-2,
+-2,
+2,
+2,
+-1,
+-3,
+-1,
+3,
+-3,
+3,
+3,
+2,
+-3,
+2,
+-3,
+-3,
+-2,
+3,
+-1,
+0,
+-2,
+-1,
+-1,
+2,
+-1,
+-2,
+1,
+-4,
+1,
+3,
+2,
+1,
+2,
+-1,
+-1,
+-1,
+-4,
+0,
+3,
+1,
+-2,
+0,
+2,
+-2,
+-2,
+3,
+-1,
+-1,
+0,
+3,
+3,
+-1,
+0,
+0,
+-4,
+1,
+-4,
+-4,
+0,
+2,
+3,
+-3,
+-2,
+-3,
+-3,
+-2,
+0,
+-4,
+-1,
+0,
+-2,
+1,
+-1,
+-4,
+1,
+2,
+0,
+-2,
+0,
+2,
+2,
+2,
+3,
+-3,
+1,
+0,
+-2,
+2,
+3,
+-1,
+-2,
+1,
+-3,
+-1,
+2,
+2,
+1,
+3,
+-2,
+0,
+-2,
+-4,
+1,
+1,
+1,
+-2,
+-1,
+2,
+0,
+1,
+-1,
+-3,
+-1,
+1,
+-1,
+-1,
+-3,
+-4,
+-2,
+-2,
+-1,
+0,
+3,
+3,
+0,
+-2,
+-2,
+-2,
+-3,
+2,
+2,
+1,
+3,
+0,
+3,
+0,
+-1,
+-1,
+3,
+3,
+-4,
+1,
+1,
+-2,
+-4,
+-4,
+3,
+1,
+0,
+-1,
+-4,
+-2,
+2,
+0,
+1,
+-1,
+0,
+-3,
+-2,
+-1,
+1,
+-3,
+-2,
+2,
+-4,
+-3,
+-3,
+0,
+0,
+-3,
+-3,
+2,
+0,
+1,
+-2,
+-3,
+-1,
+3,
+-1,
+3,
+-3,
+-4,
+-4,
+0,
+-4,
+-1,
+2,
+1,
+0,
+0,
+2,
+2,
+3,
+2,
+-1,
+0,
+-3,
+-3,
+3,
+0,
+-4,
+0,
+-2,
+-2,
+-1,
+1,
+3,
+3,
+1,
+-3,
+1,
+-2,
+-1,
+2,
+0,
+0,
+2,
+-1,
+-4,
+-1,
+-3,
+0,
+2,
+-4,
+-1,
+0,
+-2,
+-1,
+2,
+-2,
+-2,
+3,
+3,
+0,
+3,
+1,
+0,
+2,
+3,
+2,
+-2,
+-4,
+-1,
+3,
+3,
+-1,
+2,
+0,
+-2,
+2,
+-4,
+-3,
+-3,
+-3,
+-3,
+-3,
+0,
+3,
+3,
+3,
+0,
+-3,
+-1,
+2,
+3,
+-3,
+-3,
+-4,
+-1,
+3,
+0,
+-4,
+-2,
+-3,
+1,
+3,
+3,
+-2,
+2,
+-4,
+-2,
+1,
+-3,
+-3,
+-2,
+3,
+-3,
+0,
+-4,
+0,
+2,
+-4,
+0,
+1,
+-1,
+3,
+3,
+3,
+1,
+3,
+-4,
+-4,
+-1,
+-3,
+-3,
+0,
+1,
+-1,
+-3,
+-4,
+2,
+3,
+0,
+-4,
+-2,
+-3,
+0,
+3,
+2,
+0,
+-2,
+-4,
+-3,
+-3,
+-3,
+-2,
+2,
+-4,
+-4,
+2,
+3,
+-3,
+2,
+-2,
+-2,
+-1,
+-2,
+-2,
+-1,
+3,
+-4,
+-4,
+3,
+-3,
+-3,
+-1,
+-1,
+2,
+3,
+-3,
+-1,
+-3,
+-4,
+};
+
+const char m_exp[] = {
+27,
+-26,
+-19,
+13,
+47,
+17,
+-5,
+14,
+88,
+66,
+2,
+14,
+6,
+-39,
+56,
+35,
+-19,
+61,
+-13,
+4,
+46,
+1,
+70,
+27,
+-22,
+-28,
+-13,
+-5,
+-4,
+-6,
+-10,
+5,
+26,
+14,
+3,
+-1,
+-3,
+0,
+10,
+43,
+4,
+32,
+-3,
+6,
+36,
+26,
+66,
+20,
+-20,
+-10,
+21,
+-17,
+-5,
+36,
+-9,
+11,
+10,
+17,
+-32,
+12,
+20,
+9,
+-1,
+15,
+-42,
+-21,
+-51,
+24,
+20,
+26,
+10,
+42,
+-4,
+2,
+-11,
+41,
+22,
+4,
+-17,
+10,
+28,
+-10,
+2,
+10,
+25,
+32,
+18,
+43,
+-3,
+-9,
+-9,
+-4,
+25,
+-5,
+9,
+24,
+54,
+43,
+14,
+59,
+18,
+36,
+-7,
+30,
+50,
+-30,
+21,
+3,
+27,
+7,
+21,
+55,
+-17,
+-2,
+9,
+17,
+31,
+8,
+12,
+29,
+43,
+-17,
+14,
+-55,
+13,
+50,
+24,
+-13,
+-33,
+15,
+-45,
+9,
+36,
+-9,
+-6,
+14,
+-1,
+2,
+-27,
+-37,
+-40,
+-14,
+2,
+-10,
+6,
+27,
+-8,
+-13,
+-17,
+-4,
+28,
+51,
+-5,
+-14,
+2,
+-23,
+17,
+35,
+41,
+-3,
+-37,
+-12,
+-49,
+-31,
+7,
+15,
+36,
+-15,
+-21,
+6,
+2,
+36,
+-12,
+-27,
+-34,
+9,
+44,
+8,
+57,
+20,
+-22,
+-22,
+-11,
+-15,
+-8,
+-15,
+-9,
+49,
+37,
+26,
+6,
+30,
+-17,
+-43,
+-3,
+-31,
+36,
+49,
+-15,
+8,
+51,
+-8,
+17,
+-51,
+-16,
+-21,
+12,
+49,
+-11,
+41,
+36,
+15,
+-5,
+-69,
+60,
+14,
+-24,
+-21,
+23,
+3,
+41,
+41,
+-27,
+0,
+33,
+63,
+18,
+1,
+-36,
+-20,
+18,
+7,
+-1,
+37,
+-33,
+30,
+36,
+-29,
+38,
+57,
+-8,
+-21,
+-26,
+15,
+-10,
+-4,
+-17,
+-19,
+8,
+-9,
+-13,
+0,
+10,
+-30,
+59,
+12,
+11,
+51,
+60,
+19,
+35,
+6,
+3,
+-5,
+-19,
+-10,
+20,
+6,
+-12,
+-5,
+5,
+11,
+33,
+-33,
+-27,
+-2,
+-2,
+3,
+53,
+-1,
+-15,
+-10,
+13,
+-5,
+24,
+-17,
+10,
+8,
+-3,
+41,
+-9,
+9,
+0,
+30,
+7,
+7,
+-10,
+-28,
+21,
+13,
+4,
+-3,
+44,
+9,
+-49,
+-31,
+-51,
+23,
+16,
+-13,
+-12,
+36,
+23,
+34,
+-39,
+8,
+20,
+2,
+32,
+73,
+24,
+-16,
+0,
+-30,
+17,
+52,
+-10,
+28,
+-43,
+-1,
+-34,
+16,
+-4,
+39,
+31,
+5,
+5,
+3,
+36,
+48,
+5,
+28,
+12,
+57,
+25,
+-28,
+14,
+34,
+49,
+11,
+41,
+20,
+-12,
+-18,
+-43,
+-29,
+58,
+26,
+44,
+36,
+-100,
+-46,
+-24,
+0,
+54,
+4,
+-2,
+21,
+5,
+-12,
+22,
+9,
+-35,
+-5,
+-36,
+-22,
+-10,
+-11,
+-15,
+34,
+20,
+-33,
+17,
+39,
+-13,
+-32,
+-36,
+-22,
+-3,
+-38,
+-39,
+-6,
+0,
+15,
+-6,
+-15,
+9,
+-10,
+45,
+16,
+17,
+13,
+-8,
+15,
+11,
+-4,
+59,
+18,
+-9,
+12,
+69,
+27,
+-30,
+16,
+7,
+30,
+41,
+-30,
+9,
+-4,
+23,
+-5,
+8,
+3,
+-75,
+-25,
+5,
+14,
+12,
+-21,
+-21,
+6,
+-13,
+15,
+8,
+-6,
+30,
+24,
+41,
+-12,
+5,
+39,
+18,
+-4,
+-60,
+19,
+-1,
+-3,
+-7,
+16,
+-11,
+-51,
+2,
+-10,
+-17,
+-32,
+39,
+-35,
+-46,
+10,
+27,
+1,
+21,
+13,
+78,
+-12,
+-2,
+-28,
+-20,
+11,
+75,
+85,
+-18,
+21,
+-7,
+-9,
+33,
+19,
+71,
+23,
+25,
+2,
+47,
+32,
+8,
+24,
+46,
+-22,
+23,
+18,
+-58,
+66,
+52,
+22,
+32,
+25,
+-34,
+-26,
+-21,
+-69,
+6,
+27,
+-28,
+-13,
+-15,
+10,
+-20,
+-16,
+20,
+25,
+7,
+22,
+52,
+-4,
+-8,
+-28,
+-18,
+-6,
+-36,
+6,
+-26,
+12,
+32,
+8,
+11,
+16,
+0,
+-18,
+-33,
+-10,
+11,
+-40,
+-6,
+-29,
+-34,
+-30,
+15,
+15,
+-27,
+9,
+8,
+-43,
+0,
+-5,
+4,
+12,
+60,
+-11,
+-4,
+-32,
+-25,
+-38,
+-37,
+-17,
+-1,
+20,
+37,
+43,
+7,
+-2,
+12,
+-7,
+-7,
+-14,
+-30,
+9,
+50,
+-18,
+-9,
+-1,
+-9,
+6,
+16,
+-38,
+-13,
+30,
+};
+
+#define SIZE 24
+__attribute__ ((section(".heapsram"))) char g_mA[SIZE][SIZE];
+__attribute__ ((section(".heapsram"))) char g_mB[SIZE][SIZE];
+__attribute__ ((section(".heapsram"))) char g_mC[SIZE][SIZE];
+__attribute__ ((section(".heapsram"))) char g_mB_tmp[SIZE][SIZE];
diff --git a/carfield/redmule b/carfield/redmule
new file mode 120000
index 0000000..bd0da93
--- /dev/null
+++ b/carfield/redmule
@@ -0,0 +1 @@
+../hwpe/redmule/
\ No newline at end of file
diff --git a/fpu_tests.yaml b/fpu_tests.yaml
new file mode 100644
index 0000000..963d381
--- /dev/null
+++ b/fpu_tests.yaml
@@ -0,0 +1,11 @@
+fpu_tests:
+  fp32:
+    path: ./fpu_tests/matmul/FP32
+    command: make clean all cores=8 platform=rtl fmt_A=FP32 fmt_B=FP32 fmt_OUT=FP32 thr=0.004f check=1 run
+  fp16:
+    path: ./fpu_tests/matmul/FP16
+    command: make clean all cores=8 platform=rtl fmt_A=FP16 fmt_B=FP16 fmt_OUT=FP16 thr=0.004f check=1 vect=1 run
+  fp16alt:
+    path: ./fpu_tests/matmul/FP16ALT
+    command: make clean all cores=8 platform=rtl fmt_A=FP16ALT fmt_B=FP16ALT fmt_OUT=FP16ALT thr=0.04f check=1 vect=1 run
+    
\ No newline at end of file
diff --git a/fpu_tests/matmul/FP16/Makefile b/fpu_tests/matmul/FP16/Makefile
new file mode 100644
index 0000000..f9116c0
--- /dev/null
+++ b/fpu_tests/matmul/FP16/Makefile
@@ -0,0 +1,82 @@
+PULP_APP = test
+
+PULP_APP_FC_SRCS = main.c
+PULP_APP_SRCS = support_func.c matmul.c
+
+PULP_CFLAGS += -O3 -g3 
+PULP_CFLAGS += -mno-memcpy
+
+ifdef cores
+PULP_CFLAGS += -DNUM_CORES=${cores} #-flto -DFABRIC=1
+else
+PULP_CFLAGS += -DNUM_CORES=1
+endif
+
+
+ifdef FABRIC
+PULP_CFLAGS += -DFABRIC
+endif
+
+ifdef cores
+PULP_CFLAGS += -DUSE_INTRINSICS
+endif
+
+ifdef thr
+PULP_CFLAGS += -DTHR=${thr}
+endif
+
+PULP_CFLAGS += -fno-tree-vectorize
+
+
+ifdef fmt
+PULP_CFLAGS += -D${fmt} -DFIXED
+
+else
+# FP FORMAT
+#INPUT DATA TYPE
+ifdef fmt_A
+PULP_CFLAGS += -DMA${fmt_A}
+else
+PULP_CFLAGS += -DMAFP32
+endif
+
+#FILTER DATA TYPE
+ifdef fmt_B
+PULP_CFLAGS += -DMB${fmt_B}
+else
+PULP_CFLAGS += -DMBFP32
+endif
+
+# OUTPUT DATA TYPE
+
+ifdef fmt_OUT
+PULP_CFLAGS += -DOUT${fmt_OUT}
+else
+PULP_CFLAGS += -DOUTFP32
+endif
+endif
+
+# VECTORIAL FORMAT for half-precision FP
+ifdef vec
+PULP_CFLAGS += -DVECTORIAL
+endif
+
+# CHECK RESULTS
+ifdef check
+PULP_CFLAGS += -DCHECK
+endif
+
+ifdef PRINT_RESULTS
+PULP_CFLAGS += -DPRINT_RESULTS
+endif
+
+ifdef verbose
+PULP_CFLAGS += -DVERBOSE
+endif
+
+# STATISTICS
+ifdef stats
+PULP_CFLAGS += -DSTATS
+endif
+
+include $(PULP_SDK_HOME)/install/rules/pulp.mk
diff --git a/fpu_tests/matmul/FP16/config.h b/fpu_tests/matmul/FP16/config.h
new file mode 100644
index 0000000..9505082
--- /dev/null
+++ b/fpu_tests/matmul/FP16/config.h
@@ -0,0 +1,96 @@
+#ifndef _CONFIG_MATMUL_
+#define _CONFIG_MATMUL_
+
+#ifdef FABRIC
+#define DATA_LOCATION
+#else
+#define DATA_LOCATION L1_DATA
+#endif
+
+//Define INPUT data types
+
+#ifdef FIXED
+    #ifdef FP16
+        typedef signed short      v2s    __attribute__((vector_size (4)));
+        typedef float16      MA_TYPE;
+        typedef float16      MB_TYPE;
+        typedef float16      OUT_TYPE;
+        typedef float16      MA_VTYPE    __attribute__((vector_size (4)));
+        typedef float16      MB_VTYPE    __attribute__((vector_size (4)));
+        typedef float16     OUT_VTYPE    __attribute__((vector_size (4)));
+        #undef USE_INTRINSICS
+    #elif defined(FP16ALT)
+        typedef signed short      v2s    __attribute__((vector_size (4)));
+        typedef float16alt      MA_TYPE;
+        typedef float16alt      MB_TYPE;
+        typedef float16alt      OUT_TYPE;
+        typedef float16alt      MA_VTYPE    __attribute__((vector_size (4)));
+        typedef float16alt      MB_VTYPE    __attribute__((vector_size (4)));
+        typedef float16alt     OUT_VTYPE    __attribute__((vector_size (4)));
+        #undef USE_INTRINSICS
+    #elif defined(FP32)
+        typedef float  MA_TYPE;
+        typedef float  MB_TYPE;
+        typedef float  OUT_TYPE;
+    #endif
+
+#else // MIXED
+    #ifdef MAFP32
+        typedef float      MA_TYPE;
+    #elif MAFP16
+            typedef signed short      v2s    __attribute__((vector_size (4)));
+            typedef float16      MA_TYPE;
+            typedef float16      MA_VTYPE    __attribute__((vector_size (4)));
+            #undef USE_INTRINSICS
+    #elif MAFP16ALT
+            typedef signed short      v2s    __attribute__((vector_size (4)));
+            typedef float16alt      MA_TYPE;
+            typedef float16alt     MA_VTYPE    __attribute__((vector_size (4)));
+            #undef USE_INTRINSICS
+    #endif
+        
+    #ifdef MBFP32
+            typedef float   MB_TYPE;
+    #elif MBFP16
+            typedef signed short      v2s    __attribute__((vector_size (4)));
+            typedef float16      MB_TYPE;
+            typedef float16      MB_VTYPE    __attribute__((vector_size (4)));
+            #undef USE_INTRINSICS
+    #elif MBFP16ALT
+            typedef signed short      v2s    __attribute__((vector_size (4)));
+            typedef float16alt      MB_TYPE;
+            typedef float16alt     MB_VTYPE    __attribute__((vector_size (4)));
+            #undef USE_INTRINSICS
+    #endif
+        // Define output data types
+    #ifdef OUTFP32
+            typedef float OUT_TYPE;
+    #elif OUTFP16
+            typedef signed short      v2s    __attribute__((vector_size (4)));
+            typedef float16      OUT_TYPE;
+            typedef float16      OUT_VTYPE    __attribute__((vector_size (4)));
+    #elif OUTFP16ALT
+            typedef signed short      v2s    __attribute__((vector_size (4)));
+            typedef float16alt      OUT_TYPE;
+            typedef float16alt     OUT_VTYPE    __attribute__((vector_size (4)));
+    #endif
+
+#endif
+
+#ifndef THR
+    #define THR 0.004f
+#endif
+#ifdef VECTORIAL
+    #if defined(MAFP16) && defined (MBFP16ALT) || defined (MAFP16ALT) && defined (MBFP16) 
+        #error "Vecotrization does not work for different data types...!!!" 
+    #endif
+
+    #if defined (MAFP32) || defined (MBFP32) || defined (OUTFP32)
+
+        #error "Vecotrization does not work for FP32 data type...!!!" 
+    #endif
+#endif
+
+void matMul(MA_TYPE * __restrict__ A, MB_TYPE * __restrict__ B, OUT_TYPE * __restrict__ C, int M, int N, int P);
+
+#endif
diff --git a/fpu_tests/matmul/FP16/data.h b/fpu_tests/matmul/FP16/data.h
new file mode 100644
index 0000000..5c74f33
--- /dev/null
+++ b/fpu_tests/matmul/FP16/data.h
@@ -0,0 +1,7 @@
+#define M 8
+#define N 8
+#define P 8
+
+PI_L2 MA_TYPE A_mat[] = {-1.4462890625, -0.329345703125, 0.3203125, -1.083984375, 0.25146484375, 0.15234375, -1.6630859375, 0.71728515625, 0.1917724609375, 0.5859375, 2.029296875, -0.1700439453125, 0.2003173828125, -0.1534423828125, 1.2353515625, 0.73681640625, 0.38427734375, -0.6748046875, 0.60791015625, -0.253662109375, 1.62890625, 1.1982421875, -0.134521484375, -0.3447265625, -1.3662109375, 0.039459228515625, 0.1907958984375, -1.96875, -0.18115234375, -0.11669921875, -0.1326904296875, 0.281494140625, 0.70556640625, -0.58837890625, -0.7724609375, 0.07135009765625, -0.1737060546875, 0.486328125, 0.018463134765625, -0.364501953125, 0.478271484375, 0.98828125, 0.6455078125, -0.2303466796875, 0.306640625, 0.22119140625, 0.364013671875, 0.0013647079467773438, -0.43408203125, -0.59375, 0.168701171875, 0.9921875, -0.04827880859375, -0.6396484375, 0.9873046875, -0.09088134765625, -0.61376953125, 1.4345703125, 0.400634765625, 0.259033203125, 0.93115234375, 1.4326171875, -1.08984375, -1.767578125, };
+PI_L2 MB_TYPE B_mat[] = {-0.345703125, 0.41845703125, -1.3466796875, -0.826171875, 1.490234375, -0.86474609375, 1.185546875, -0.12469482421875, -0.51025390625, -0.724609375, -0.057769775390625, -1.701171875, -0.92236328125, -0.422119140625, 0.21533203125, -0.0709228515625, -0.497314453125, -0.81298828125, 1.9052734375, -0.60546875, -1.1884765625, 1.6962890625, -0.306884765625, 0.1558837890625, 0.53759765625, 0.360107421875, 0.128662109375, -1.00390625, -0.8955078125, -0.16064453125, 0.73583984375, -1.662109375, 0.64404296875, 1.5068359375, 0.03680419921875, -0.92431640625, -0.62255859375, -1.263671875, -0.92529296875, 1.1123046875, -1.1728515625, 2.85546875, 1.126953125, -1.2001953125, 1.0205078125, 1.5361328125, 1.1337890625, -1.283203125, 1.568359375, -0.95703125, -0.2474365234375, -1.115234375, -0.416748046875, 1.0478515625, 1.419921875, 1.8359375, -0.41845703125, 0.007236480712890625, -1.646484375, -0.01568603515625, 1.05078125, -0.89013671875, 0.55126953125, -1.8427734375, };
+PI_L2 OUT_TYPE ref[] = {-2.998046875, 1.3935546875, 1.8505859375, 4.07421875, 0.18408203125, -0.357421875, -4.7109375, -2.236328125, 0.471923828125, -3.369140625, 1.869140625, -3.60546875, -2.53515625, 3.203125, 1.4072265625, 1.86328125, -0.650390625, 6.06640625, 2.66015625, -2.0703125, 0.6025390625, 0.97265625, -0.591796875, 1.1796875, -1.0068359375, -1.943359375, 1.37890625, 3.375, -0.19140625, 1.46484375, -3.115234375, 2.65625, -0.0216064453125, 2.482421875, -1.240234375, 0.376220703125, 2.66015625, -0.372802734375, 1.5380859375, -0.396484375, -0.607421875, -0.378662109375, 0.66748046875, -3.19140625, -0.87548828125, 0.6337890625, 0.8974609375, 1.076171875, 3.20703125, -2.375, 0.2509765625, -0.0181121826171875, -2.318359375, 0.94677734375, 0.70751953125, 1.220703125, -2.62890625, 4.99609375, 6.3671875, -3.77734375, -3.46875, 2.01953125, -2.111328125, 0.0625, };
diff --git a/fpu_tests/matmul/FP16/main.c b/fpu_tests/matmul/FP16/main.c
new file mode 100644
index 0000000..7d12782
--- /dev/null
+++ b/fpu_tests/matmul/FP16/main.c
@@ -0,0 +1,122 @@
+// License, Version 0.51 (the "License"); you may not use this file except in
+// or agreed to in writing, software, hardware and materials distributed under
+// this License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR
+// CONDITIONS OF ANY KIND, either express or implied. See the License for the
+// specific language governing permissions and limitations under the License.
+
+#include "config.h"
+#include "pulp.h"
+
+#include <stdio.h>
+#include <stdint.h>
+#include <limits.h> /* for CHAR_BIT */
+#include <math.h>
+
+#include "data.h"
+
+#define STACK_SIZE 2048
+
+void main_fn(testresult_t *result, void (*start)(), void (*stop)());
+
+testcase_t testcases[] = {
+  { .name = "Matrix Multiplication", .test = main_fn },
+  {0, 0}
+};
+
+DATA_LOCATION MA_TYPE matA[M*N] __attribute__ ((aligned (4)));
+DATA_LOCATION MB_TYPE matB[N*P] __attribute__ ((aligned (4)));
+DATA_LOCATION OUT_TYPE matC[M*P] __attribute__ ((aligned (4)));
+
+// End of computation
+int done = 0;
+
+int retval = 0;
+
+void __attribute__ ((noinline)) matrix_init(MA_TYPE * __restrict__ A, MB_TYPE * __restrict__ B, OUT_TYPE * __restrict__ C) {
+  for (int i = 0; i < M; i++) 
+    for (int j = 0; j < N; j++){
+      A[i*N+j] = A_mat[i*N+j];
+
+
+    } 
+      
+  for (int i = 0; i < N; i++) 
+    for (int j = 0; j < P; j++){
+      B[i*P+j] = B_mat[i*P+j];
+    }
+  for (int i = 0; i < M; i++) 
+    for (int j = 0; j < P; j++)  
+      C[i*P+j] = 0;
+  
+}
+
+int __attribute ((noinline)) check_result(OUT_TYPE * __restrict__ result) {
+  #ifndef FABRIC
+    synch_barrier();
+  #endif
+
+  if(get_core_id() == 0) {
+    float diff;
+    int err = 0;
+
+    for (int i = 0; i < (M*P); i++) {
+      diff = fabs(result[i] - ref[i]);
+      if(diff > THR) {
+        err++;
+      #ifdef VERBOSE
+
+        printf("Error at index %d:\t refrence %f\t output %f\t error %.4f\n", i, ref[i], result[i], diff);
+      #endif
+      
+      }
+
+      #ifdef PRINT_RESULTS
+
+        printf("index %d:\t refrence %f\t output %f\t error %f\n", i, ref[i], result[i], diff);
+      #endif
+    }
+
+    return err;
+
+  }
+}
+
+void main_fn(testresult_t *result, void (*start)(), void (*stop)()){
+  
+  if (get_core_id() == 0)
+    matrix_init(matA, matB, matC);
+ 
+  #ifndef FABRIC
+  synch_barrier();
+  #endif
+  
+  #ifdef STATS
+  start();
+  #endif
+  matMul(matA, matB, matC, M, N, P);
+
+  #ifdef STATS
+  stop();
+  #endif
+
+  #ifdef CHECK
+  result->errors = check_result(matC);
+  #endif
+};
+
+int main()
+{
+  #ifdef FABRIC
+      main_fn();
+  #else
+  if (rt_cluster_id() != 0)
+    return bench_cluster_forward(0);
+
+  int nbErrors = run_suite(testcases);
+
+  synch_barrier();
+  #endif
+  retval = nbErrors;
+
+  return retval;
+}
diff --git a/fpu_tests/matmul/FP16/matmul.c b/fpu_tests/matmul/FP16/matmul.c
new file mode 100644
index 0000000..0a616ca
--- /dev/null
+++ b/fpu_tests/matmul/FP16/matmul.c
@@ -0,0 +1,103 @@
+#include "config.h"
+#include "pulp.h"
+
+#ifdef VECTORIAL
+
+void __attribute__ ((noinline)) matMul(MA_TYPE * __restrict__ A, MB_TYPE * __restrict__ B, OUT_TYPE * __restrict__ C, int M, int N, int P){
+
+  OUT_VTYPE temp;
+  MA_VTYPE Av;
+  MB_VTYPE Bv0;
+  MB_VTYPE Bv1;
+  OUT_VTYPE *Cv;
+  int blockSize = (M+NUM_CORES-1)/NUM_CORES;
+  int start = get_core_id()*blockSize;
+  int end = start + blockSize < M? start + blockSize : M;
+
+  for (int i = start; i < end; i++) {
+    for (int j=0; j < (P & 0xfffffffe); j+=2) {
+        
+      temp = (OUT_VTYPE) {0, 0};
+
+			// Manual unrolling
+			for (int k=0; k<(N & 0xfffffffe); k+=2){
+				Av  = *((MA_VTYPE *) &A[i*N+k]);
+				Bv0 = *((MB_VTYPE *) &B[k*P+j]);
+				Bv1 = *((MB_VTYPE *) &B[k*P+j+P]);
+				temp += (OUT_VTYPE)(__builtin_shuffle(Av, (v2s){0,0})) * Bv0;
+				temp += (OUT_VTYPE)(__builtin_shuffle(Av, (v2s){1,1})) * Bv1;
+			}
+
+    if (N & 0x00000001) 
+     {
+      temp[0] += A[i*N+N-1] * B[(N-1)*P+j];
+      temp[1] += A[i*N+N-1] * B[(N-1)*P+j+1];
+     }
+			Cv = (OUT_VTYPE *) &C[i*P+j];
+
+			*Cv = temp;
+    }
+  }
+  /// Leftover in P
+    if (P & 0x00000001) 
+    {
+      for (int i = start; i < end; i++) {
+
+            OUT_TYPE temp1 = 0;
+
+            // Manual unrolling
+            for (int k=0; k<(N & 0xfffffffe); k+=2){
+              temp1 += A[i*N+k] * B[k*P+P-1];
+              temp1 += A[i*N+k+1] * B[k*P+P-1+P];
+            }
+          if (N & 0x00000001) 
+               {
+                temp1 += A[i*N+N-1] * B[(N-1)*P+P-1];
+               }
+            C[i*P+(P-1)]=temp1;
+          }
+      }
+      
+  #if NUM_CORES > 1
+  synch_barrier();
+  #endif  
+}
+#else
+
+void __attribute__ ((noinline)) matMul(MA_TYPE * __restrict__ A, MB_TYPE * __restrict__ B, OUT_TYPE * __restrict__ C, int M, int N, int P) {
+
+  int blockSize = (M+NUM_CORES-1)/NUM_CORES;
+  int start = get_core_id()*blockSize;
+  int end = start + blockSize < M? start + blockSize : M;
+  
+  for (int i = start; i < end; i++) {
+    for (int j = 0; j < P; j++) {
+      OUT_TYPE temp = 0;
+
+      //Manual unrolling
+      for (int k = 0; k < (N & 0xfffffffe); k+=2) {
+        temp += (OUT_TYPE)(A[i*N+k]   * B[k*P+j]);
+        temp += (OUT_TYPE)(A[i*N+k+1] * B[k*P+j+P]);
+
+      }
+         C[i*P+j] = (OUT_TYPE)(temp);
+    }
+  }
+      // Leftover on N
+
+      if (N & 0x00000001) 
+      {
+        for (int i=start; i<end; i++) 
+        {
+          for (int j=0; j<P; j++) 
+          {
+            C[i*P+j] += (OUT_TYPE)(A[i*N+N-1] * B[(N-1)*P+j]);
+          }
+        }
+      } 
+
+  #if NUM_CORES > 1
+  synch_barrier();
+  #endif
+}
+#endif
diff --git a/fpu_tests/matmul/FP16/support_func.c b/fpu_tests/matmul/FP16/support_func.c
new file mode 100644
index 0000000..6d24d2e
--- /dev/null
+++ b/fpu_tests/matmul/FP16/support_func.c
@@ -0,0 +1,15 @@
+#include "config.h"
+#include "pulp.h"
+double __extendohfdf2(float16alt value)
+{
+  float result;
+  __asm__ __volatile__ ("fcvt.s.ah %0, %1": "=f"(result): "f"(value) :);
+  return (double) result;
+}
+
+double __extendhfdf2(float16 value)
+{
+  float result;
+  __asm__ __volatile__ ("fcvt.s.h %0, %1": "=f"(result): "f"(value) :);
+  return (double) result;
+}
diff --git a/fpu_tests/matmul/FP16ALT/Makefile b/fpu_tests/matmul/FP16ALT/Makefile
new file mode 100644
index 0000000..f9116c0
--- /dev/null
+++ b/fpu_tests/matmul/FP16ALT/Makefile
@@ -0,0 +1,82 @@
+PULP_APP = test
+
+PULP_APP_FC_SRCS = main.c
+PULP_APP_SRCS = support_func.c matmul.c
+
+PULP_CFLAGS += -O3 -g3 
+PULP_CFLAGS += -mno-memcpy
+
+ifdef cores
+PULP_CFLAGS += -DNUM_CORES=${cores} #-flto -DFABRIC=1
+else
+PULP_CFLAGS += -DNUM_CORES=1
+endif
+
+
+ifdef FABRIC
+PULP_CFLAGS += -DFABRIC
+endif
+
+ifdef cores
+PULP_CFLAGS += -DUSE_INTRINSICS
+endif
+
+ifdef thr
+PULP_CFLAGS += -DTHR=${thr}
+endif
+
+PULP_CFLAGS += -fno-tree-vectorize
+
+
+ifdef fmt
+PULP_CFLAGS += -D${fmt} -DFIXED
+
+else
+# FP FORMAT
+#INPUT DATA TYPE
+ifdef fmt_A
+PULP_CFLAGS += -DMA${fmt_A}
+else
+PULP_CFLAGS += -DMAFP32
+endif
+
+#FILTER DATA TYPE
+ifdef fmt_B
+PULP_CFLAGS += -DMB${fmt_B}
+else
+PULP_CFLAGS += -DMBFP32
+endif
+
+# OUTPUT DATA TYPE
+
+ifdef fmt_OUT
+PULP_CFLAGS += -DOUT${fmt_OUT}
+else
+PULP_CFLAGS += -DOUTFP32
+endif
+endif
+
+# VECTORIAL FORMAT for half-precision FP
+ifdef vec
+PULP_CFLAGS += -DVECTORIAL
+endif
+
+# CHECK RESULTS
+ifdef check
+PULP_CFLAGS += -DCHECK
+endif
+
+ifdef PRINT_RESULTS
+PULP_CFLAGS += -DPRINT_RESULTS
+endif
+
+ifdef verbose
+PULP_CFLAGS += -DVERBOSE
+endif
+
+# STATISTICS
+ifdef stats
+PULP_CFLAGS += -DSTATS
+endif
+
+include $(PULP_SDK_HOME)/install/rules/pulp.mk
diff --git a/fpu_tests/matmul/FP16ALT/config.h b/fpu_tests/matmul/FP16ALT/config.h
new file mode 100644
index 0000000..9505082
--- /dev/null
+++ b/fpu_tests/matmul/FP16ALT/config.h
@@ -0,0 +1,96 @@
+#ifndef _CONFIG_MATMUL_
+#define _CONFIG_MATMUL_
+
+#ifdef FABRIC
+#define DATA_LOCATION
+#else
+#define DATA_LOCATION L1_DATA
+#endif
+
+//Define INPUT data types
+
+#ifdef FIXED
+    #ifdef FP16
+        typedef signed short      v2s    __attribute__((vector_size (4)));
+        typedef float16      MA_TYPE;
+        typedef float16      MB_TYPE;
+        typedef float16      OUT_TYPE;
+        typedef float16      MA_VTYPE    __attribute__((vector_size (4)));
+        typedef float16      MB_VTYPE    __attribute__((vector_size (4)));
+        typedef float16     OUT_VTYPE    __attribute__((vector_size (4)));
+        #undef USE_INTRINSICS
+    #elif defined(FP16ALT)
+        typedef signed short      v2s    __attribute__((vector_size (4)));
+        typedef float16alt      MA_TYPE;
+        typedef float16alt      MB_TYPE;
+        typedef float16alt      OUT_TYPE;
+        typedef float16alt      MA_VTYPE    __attribute__((vector_size (4)));
+        typedef float16alt      MB_VTYPE    __attribute__((vector_size (4)));
+        typedef float16alt     OUT_VTYPE    __attribute__((vector_size (4)));
+        #undef USE_INTRINSICS
+    #elif defined(FP32)
+        typedef float  MA_TYPE;
+        typedef float  MB_TYPE;
+        typedef float  OUT_TYPE;
+    #endif
+
+#else // MIXED
+    #ifdef MAFP32
+        typedef float      MA_TYPE;
+    #elif MAFP16
+            typedef signed short      v2s    __attribute__((vector_size (4)));
+            typedef float16      MA_TYPE;
+            typedef float16      MA_VTYPE    __attribute__((vector_size (4)));
+            #undef USE_INTRINSICS
+    #elif MAFP16ALT
+            typedef signed short      v2s    __attribute__((vector_size (4)));
+            typedef float16alt      MA_TYPE;
+            typedef float16alt     MA_VTYPE    __attribute__((vector_size (4)));
+            #undef USE_INTRINSICS
+    #endif
+        
+    #ifdef MBFP32
+            typedef float   MB_TYPE;
+    #elif MBFP16
+            typedef signed short      v2s    __attribute__((vector_size (4)));
+            typedef float16      MB_TYPE;
+            typedef float16      MB_VTYPE    __attribute__((vector_size (4)));
+            #undef USE_INTRINSICS
+    #elif MBFP16ALT
+            typedef signed short      v2s    __attribute__((vector_size (4)));
+            typedef float16alt      MB_TYPE;
+            typedef float16alt     MB_VTYPE    __attribute__((vector_size (4)));
+            #undef USE_INTRINSICS
+    #endif
+        // Define output data types
+    #ifdef OUTFP32
+            typedef float OUT_TYPE;
+    #elif OUTFP16
+            typedef signed short      v2s    __attribute__((vector_size (4)));
+            typedef float16      OUT_TYPE;
+            typedef float16      OUT_VTYPE    __attribute__((vector_size (4)));
+    #elif OUTFP16ALT
+            typedef signed short      v2s    __attribute__((vector_size (4)));
+            typedef float16alt      OUT_TYPE;
+            typedef float16alt     OUT_VTYPE    __attribute__((vector_size (4)));
+    #endif
+
+#endif
+
+#ifndef THR
+    #define THR 0.004f
+#endif
+#ifdef VECTORIAL
+    #if defined(MAFP16) && defined (MBFP16ALT) || defined (MAFP16ALT) && defined (MBFP16) 
+        #error "Vecotrization does not work for different data types...!!!" 
+    #endif
+
+    #if defined (MAFP32) || defined (MBFP32) || defined (OUTFP32)
+
+        #error "Vecotrization does not work for FP32 data type...!!!" 
+    #endif
+#endif
+
+void matMul(MA_TYPE * __restrict__ A, MB_TYPE * __restrict__ B, OUT_TYPE * __restrict__ C, int M, int N, int P);
+
+#endif
diff --git a/fpu_tests/matmul/FP16ALT/data.h b/fpu_tests/matmul/FP16ALT/data.h
new file mode 100644
index 0000000..f01329d
--- /dev/null
+++ b/fpu_tests/matmul/FP16ALT/data.h
@@ -0,0 +1,7 @@
+#define M 8
+#define N 8
+#define P 8
+
+PI_L2 MA_TYPE A_mat[] = {-0.1435546875, -2.203125, 0.20703125, -0.00146484375, -0.10498046875, -0.255859375, 1.3515625, -0.7109375, 1.171875, 0.283203125, 0.390625, 1.7890625, 1.4921875, 0.59375, 0.5625, -2.40625, -0.76953125, 0.83203125, 0.89453125, 0.1142578125, 0.365234375, -0.55859375, 2.15625, -0.5859375, -1.1328125, -0.55859375, -0.1259765625, 1.4921875, 0.41015625, 0.11669921875, -0.78515625, -1.0859375, -0.71484375, -0.67578125, -0.76171875, 1.2109375, 1.1484375, 0.66015625, 0.6875, -0.74609375, -0.03564453125, 0.72265625, 0.0849609375, -1.234375, 0.72265625, -2.140625, -0.97265625, -2.328125, -0.9453125, 0.09912109375, -0.00640869140625, -0.59765625, -0.5078125, -0.75, 0.04248046875, -0.431640625, -0.326171875, -0.52734375, 0.5078125, -1.0078125, -0.07275390625, 0.125, 0.055419921875, -1.09375, };
+PI_L2 MB_TYPE B_mat[] = {-3.21875, 0.76953125, -0.875, -0.44921875, 1.0078125, -1.7265625, 0.369140625, 0.287109375, 0.13671875, -0.50390625, -0.486328125, 0.71484375, 0.392578125, -0.1494140625, 0.5625, 0.451171875, -0.16796875, 0.40234375, 0.3671875, 1.4921875, -0.796875, 1.1015625, 0.388671875, 0.97265625, -0.609375, 1.3046875, -0.0189208984375, -1.328125, 0.0986328125, 0.67578125, -0.09716796875, -0.1962890625, 0.671875, 0.271484375, 0.7421875, -0.8828125, -1.0625, -2.171875, 0.30859375, 1.4453125, -1.5703125, -0.57421875, 1.1015625, 0.953125, 0.396484375, 0.77734375, 0.435546875, 1.9375, -0.0255126953125, 1.8046875, 1.3203125, 0.80078125, -2.140625, 0.5625, 0.08203125, -0.6796875, -1.6328125, -0.08544921875, -1.3046875, 0.3828125, 0.3515625, 0.333984375, -0.89453125, -0.1181640625, };
+PI_L2 OUT_TYPE ref[] = {1.59375, 3.703125, 3.609375, -0.5390625, -4.3125, 1.359375, -0.61328125, -2.3125, -0.890625, 4.53125, 4.59375, -3.34375, -2.25, -3.6875, 3.5, 3.71875, 4.40625, 3.84375, 3.859375, 2.78125, -6.59375, 2.0625, 1.09375, -0.953125, 4.5625, 0.0224609375, 2.0, -3.359375, -0.201171875, 1.3125, 0.15234375, 0.4921875, 2.53125, 2.296875, 4.09375, -3.03125, -2.953125, -0.5390625, 0.30078125, 1.0703125, 8.625, -2.109375, -0.34375, -2.0625, -0.3046875, -5.34375, 1.8359375, -1.5390625, 4.9375, -1.15625, 0.203125, 0.8828125, -0.96875, 1.609375, -0.328125, -2.296875, 3.046875, -0.9921875, 2.328125, 1.671875, -1.4140625, 0.443359375, 0.890625, 0.5859375, };
diff --git a/fpu_tests/matmul/FP16ALT/main.c b/fpu_tests/matmul/FP16ALT/main.c
new file mode 100644
index 0000000..7d12782
--- /dev/null
+++ b/fpu_tests/matmul/FP16ALT/main.c
@@ -0,0 +1,122 @@
+// License, Version 0.51 (the "License"); you may not use this file except in
+// or agreed to in writing, software, hardware and materials distributed under
+// this License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR
+// CONDITIONS OF ANY KIND, either express or implied. See the License for the
+// specific language governing permissions and limitations under the License.
+
+#include "config.h"
+#include "pulp.h"
+
+#include <stdio.h>
+#include <stdint.h>
+#include <limits.h> /* for CHAR_BIT */
+#include <math.h>
+
+#include "data.h"
+
+#define STACK_SIZE 2048
+
+void main_fn(testresult_t *result, void (*start)(), void (*stop)());
+
+testcase_t testcases[] = {
+  { .name = "Matrix Multiplication", .test = main_fn },
+  {0, 0}
+};
+
+DATA_LOCATION MA_TYPE matA[M*N] __attribute__ ((aligned (4)));
+DATA_LOCATION MB_TYPE matB[N*P] __attribute__ ((aligned (4)));
+DATA_LOCATION OUT_TYPE matC[M*P] __attribute__ ((aligned (4)));
+
+// End of computation
+int done = 0;
+
+int retval = 0;
+
+void __attribute__ ((noinline)) matrix_init(MA_TYPE * __restrict__ A, MB_TYPE * __restrict__ B, OUT_TYPE * __restrict__ C) {
+  for (int i = 0; i < M; i++) 
+    for (int j = 0; j < N; j++){
+      A[i*N+j] = A_mat[i*N+j];
+
+
+    } 
+      
+  for (int i = 0; i < N; i++) 
+    for (int j = 0; j < P; j++){
+      B[i*P+j] = B_mat[i*P+j];
+    }
+  for (int i = 0; i < M; i++) 
+    for (int j = 0; j < P; j++)  
+      C[i*P+j] = 0;
+  
+}
+
+int __attribute ((noinline)) check_result(OUT_TYPE * __restrict__ result) {
+  #ifndef FABRIC
+    synch_barrier();
+  #endif
+
+  if(get_core_id() == 0) {
+    float diff;
+    int err = 0;
+
+    for (int i = 0; i < (M*P); i++) {
+      diff = fabs(result[i] - ref[i]);
+      if(diff > THR) {
+        err++;
+      #ifdef VERBOSE
+
+        printf("Error at index %d:\t refrence %f\t output %f\t error %.4f\n", i, ref[i], result[i], diff);
+      #endif
+      
+      }
+
+      #ifdef PRINT_RESULTS
+
+        printf("index %d:\t refrence %f\t output %f\t error %f\n", i, ref[i], result[i], diff);
+      #endif
+    }
+
+    return err;
+
+  }
+}
+
+void main_fn(testresult_t *result, void (*start)(), void (*stop)()){
+  
+  if (get_core_id() == 0)
+    matrix_init(matA, matB, matC);
+ 
+  #ifndef FABRIC
+  synch_barrier();
+  #endif
+  
+  #ifdef STATS
+  start();
+  #endif
+  matMul(matA, matB, matC, M, N, P);
+
+  #ifdef STATS
+  stop();
+  #endif
+
+  #ifdef CHECK
+  result->errors = check_result(matC);
+  #endif
+};
+
+int main()
+{
+  #ifdef FABRIC
+      main_fn();
+  #else
+  if (rt_cluster_id() != 0)
+    return bench_cluster_forward(0);
+
+  int nbErrors = run_suite(testcases);
+
+  synch_barrier();
+  #endif
+  retval = nbErrors;
+
+  return retval;
+}
diff --git a/fpu_tests/matmul/FP16ALT/matmul.c b/fpu_tests/matmul/FP16ALT/matmul.c
new file mode 100644
index 0000000..0a616ca
--- /dev/null
+++ b/fpu_tests/matmul/FP16ALT/matmul.c
@@ -0,0 +1,103 @@
+#include "config.h"
+#include "pulp.h"
+
+#ifdef VECTORIAL
+
+void __attribute__ ((noinline)) matMul(MA_TYPE * __restrict__ A, MB_TYPE * __restrict__ B, OUT_TYPE * __restrict__ C, int M, int N, int P){
+
+  OUT_VTYPE temp;
+  MA_VTYPE Av;
+  MB_VTYPE Bv0;
+  MB_VTYPE Bv1;
+  OUT_VTYPE *Cv;
+  int blockSize = (M+NUM_CORES-1)/NUM_CORES;
+  int start = get_core_id()*blockSize;
+  int end = start + blockSize < M? start + blockSize : M;
+
+  for (int i = start; i < end; i++) {
+    for (int j=0; j < (P & 0xfffffffe); j+=2) {
+        
+      temp = (OUT_VTYPE) {0, 0};
+
+			// Manual unrolling
+			for (int k=0; k<(N & 0xfffffffe); k+=2){
+				Av  = *((MA_VTYPE *) &A[i*N+k]);
+				Bv0 = *((MB_VTYPE *) &B[k*P+j]);
+				Bv1 = *((MB_VTYPE *) &B[k*P+j+P]);
+				temp += (OUT_VTYPE)(__builtin_shuffle(Av, (v2s){0,0})) * Bv0;
+				temp += (OUT_VTYPE)(__builtin_shuffle(Av, (v2s){1,1})) * Bv1;
+			}
+
+    if (N & 0x00000001) 
+     {
+      temp[0] += A[i*N+N-1] * B[(N-1)*P+j];
+      temp[1] += A[i*N+N-1] * B[(N-1)*P+j+1];
+     }
+			Cv = (OUT_VTYPE *) &C[i*P+j];
+
+			*Cv = temp;
+    }
+  }
+  /// Leftover in P
+    if (P & 0x00000001) 
+    {
+      for (int i = start; i < end; i++) {
+
+            OUT_TYPE temp1 = 0;
+
+            // Manual unrolling
+            for (int k=0; k<(N & 0xfffffffe); k+=2){
+              temp1 += A[i*N+k] * B[k*P+P-1];
+              temp1 += A[i*N+k+1] * B[k*P+P-1+P];
+            }
+          if (N & 0x00000001) 
+               {
+                temp1 += A[i*N+N-1] * B[(N-1)*P+P-1];
+               }
+            C[i*P+(P-1)]=temp1;
+          }
+      }
+      
+  #if NUM_CORES > 1
+  synch_barrier();
+  #endif  
+}
+#else
+
+void __attribute__ ((noinline)) matMul(MA_TYPE * __restrict__ A, MB_TYPE * __restrict__ B, OUT_TYPE * __restrict__ C, int M, int N, int P) {
+
+  int blockSize = (M+NUM_CORES-1)/NUM_CORES;
+  int start = get_core_id()*blockSize;
+  int end = start + blockSize < M? start + blockSize : M;
+  
+  for (int i = start; i < end; i++) {
+    for (int j = 0; j < P; j++) {
+      OUT_TYPE temp = 0;
+
+      //Manual unrolling
+      for (int k = 0; k < (N & 0xfffffffe); k+=2) {
+        temp += (OUT_TYPE)(A[i*N+k]   * B[k*P+j]);
+        temp += (OUT_TYPE)(A[i*N+k+1] * B[k*P+j+P]);
+
+      }
+         C[i*P+j] = (OUT_TYPE)(temp);
+    }
+  }
+      // Leftover on N
+
+      if (N & 0x00000001) 
+      {
+        for (int i=start; i<end; i++) 
+        {
+          for (int j=0; j<P; j++) 
+          {
+            C[i*P+j] += (OUT_TYPE)(A[i*N+N-1] * B[(N-1)*P+j]);
+          }
+        }
+      } 
+
+  #if NUM_CORES > 1
+  synch_barrier();
+  #endif
+}
+#endif
diff --git a/fpu_tests/matmul/FP16ALT/support_func.c b/fpu_tests/matmul/FP16ALT/support_func.c
new file mode 100644
index 0000000..6d24d2e
--- /dev/null
+++ b/fpu_tests/matmul/FP16ALT/support_func.c
@@ -0,0 +1,15 @@
+#include "config.h"
+#include "pulp.h"
+double __extendohfdf2(float16alt value)
+{
+  float result;
+  __asm__ __volatile__ ("fcvt.s.ah %0, %1": "=f"(result): "f"(value) :);
+  return (double) result;
+}
+
+double __extendhfdf2(float16 value)
+{
+  float result;
+  __asm__ __volatile__ ("fcvt.s.h %0, %1": "=f"(result): "f"(value) :);
+  return (double) result;
+}
diff --git a/fpu_tests/matmul/FP32/Makefile b/fpu_tests/matmul/FP32/Makefile
new file mode 100644
index 0000000..f9116c0
--- /dev/null
+++ b/fpu_tests/matmul/FP32/Makefile
@@ -0,0 +1,82 @@
+PULP_APP = test
+
+PULP_APP_FC_SRCS = main.c
+PULP_APP_SRCS = support_func.c matmul.c
+
+PULP_CFLAGS += -O3 -g3 
+PULP_CFLAGS += -mno-memcpy
+
+ifdef cores
+PULP_CFLAGS += -DNUM_CORES=${cores} #-flto -DFABRIC=1
+else
+PULP_CFLAGS += -DNUM_CORES=1
+endif
+
+
+ifdef FABRIC
+PULP_CFLAGS += -DFABRIC
+endif
+
+ifdef cores
+PULP_CFLAGS += -DUSE_INTRINSICS
+endif
+
+ifdef thr
+PULP_CFLAGS += -DTHR=${thr}
+endif
+
+PULP_CFLAGS += -fno-tree-vectorize
+
+
+ifdef fmt
+PULP_CFLAGS += -D${fmt} -DFIXED
+
+else
+# FP FORMAT
+#INPUT DATA TYPE
+ifdef fmt_A
+PULP_CFLAGS += -DMA${fmt_A}
+else
+PULP_CFLAGS += -DMAFP32
+endif
+
+#FILTER DATA TYPE
+ifdef fmt_B
+PULP_CFLAGS += -DMB${fmt_B}
+else
+PULP_CFLAGS += -DMBFP32
+endif
+
+# OUTPUT DATA TYPE
+
+ifdef fmt_OUT
+PULP_CFLAGS += -DOUT${fmt_OUT}
+else
+PULP_CFLAGS += -DOUTFP32
+endif
+endif
+
+# VECTORIAL FORMAT for half-precision FP
+ifdef vec
+PULP_CFLAGS += -DVECTORIAL
+endif
+
+# CHECK RESULTS
+ifdef check
+PULP_CFLAGS += -DCHECK
+endif
+
+ifdef PRINT_RESULTS
+PULP_CFLAGS += -DPRINT_RESULTS
+endif
+
+ifdef verbose
+PULP_CFLAGS += -DVERBOSE
+endif
+
+# STATISTICS
+ifdef stats
+PULP_CFLAGS += -DSTATS
+endif
+
+include $(PULP_SDK_HOME)/install/rules/pulp.mk
diff --git a/fpu_tests/matmul/FP32/config.h b/fpu_tests/matmul/FP32/config.h
new file mode 100644
index 0000000..9505082
--- /dev/null
+++ b/fpu_tests/matmul/FP32/config.h
@@ -0,0 +1,96 @@
+#ifndef _CONFIG_MATMUL_
+#define _CONFIG_MATMUL_
+
+#ifdef FABRIC
+#define DATA_LOCATION
+#else
+#define DATA_LOCATION L1_DATA
+#endif
+
+//Define INPUT data types
+
+#ifdef FIXED
+    #ifdef FP16
+        typedef signed short      v2s    __attribute__((vector_size (4)));
+        typedef float16      MA_TYPE;
+        typedef float16      MB_TYPE;
+        typedef float16      OUT_TYPE;
+        typedef float16      MA_VTYPE    __attribute__((vector_size (4)));
+        typedef float16      MB_VTYPE    __attribute__((vector_size (4)));
+        typedef float16     OUT_VTYPE    __attribute__((vector_size (4)));
+        #undef USE_INTRINSICS
+    #elif defined(FP16ALT)
+        typedef signed short      v2s    __attribute__((vector_size (4)));
+        typedef float16alt      MA_TYPE;
+        typedef float16alt      MB_TYPE;
+        typedef float16alt      OUT_TYPE;
+        typedef float16alt      MA_VTYPE    __attribute__((vector_size (4)));
+        typedef float16alt      MB_VTYPE    __attribute__((vector_size (4)));
+        typedef float16alt     OUT_VTYPE    __attribute__((vector_size (4)));
+        #undef USE_INTRINSICS
+    #elif defined(FP32)
+        typedef float  MA_TYPE;
+        typedef float  MB_TYPE;
+        typedef float  OUT_TYPE;
+    #endif
+
+#else // MIXED
+    #ifdef MAFP32
+        typedef float      MA_TYPE;
+    #elif MAFP16
+            typedef signed short      v2s    __attribute__((vector_size (4)));
+            typedef float16      MA_TYPE;
+            typedef float16      MA_VTYPE    __attribute__((vector_size (4)));
+            #undef USE_INTRINSICS
+    #elif MAFP16ALT
+            typedef signed short      v2s    __attribute__((vector_size (4)));
+            typedef float16alt      MA_TYPE;
+            typedef float16alt     MA_VTYPE    __attribute__((vector_size (4)));
+            #undef USE_INTRINSICS
+    #endif
+        
+    #ifdef MBFP32
+            typedef float   MB_TYPE;
+    #elif MBFP16
+            typedef signed short      v2s    __attribute__((vector_size (4)));
+            typedef float16      MB_TYPE;
+            typedef float16      MB_VTYPE    __attribute__((vector_size (4)));
+            #undef USE_INTRINSICS
+    #elif MBFP16ALT
+            typedef signed short      v2s    __attribute__((vector_size (4)));
+            typedef float16alt      MB_TYPE;
+            typedef float16alt     MB_VTYPE    __attribute__((vector_size (4)));
+            #undef USE_INTRINSICS
+    #endif
+        // Define output data types
+    #ifdef OUTFP32
+            typedef float OUT_TYPE;
+    #elif OUTFP16
+            typedef signed short      v2s    __attribute__((vector_size (4)));
+            typedef float16      OUT_TYPE;
+            typedef float16      OUT_VTYPE    __attribute__((vector_size (4)));
+    #elif OUTFP16ALT
+            typedef signed short      v2s    __attribute__((vector_size (4)));
+            typedef float16alt      OUT_TYPE;
+            typedef float16alt     OUT_VTYPE    __attribute__((vector_size (4)));
+    #endif
+
+#endif
+
+#ifndef THR
+    #define THR 0.004f
+#endif
+#ifdef VECTORIAL
+    #if defined(MAFP16) && defined (MBFP16ALT) || defined (MAFP16ALT) && defined (MBFP16) 
+        #error "Vecotrization does not work for different data types...!!!" 
+    #endif
+
+    #if defined (MAFP32) || defined (MBFP32) || defined (OUTFP32)
+
+        #error "Vecotrization does not work for FP32 data type...!!!" 
+    #endif
+#endif
+
+void matMul(MA_TYPE * __restrict__ A, MB_TYPE * __restrict__ B, OUT_TYPE * __restrict__ C, int M, int N, int P);
+
+#endif
diff --git a/fpu_tests/matmul/FP32/data.h b/fpu_tests/matmul/FP32/data.h
new file mode 100644
index 0000000..18b3c69
--- /dev/null
+++ b/fpu_tests/matmul/FP32/data.h
@@ -0,0 +1,7 @@
+#define M 8
+#define N 8
+#define P 8
+
+PI_L2 MA_TYPE A_mat[] = {-1.321346640586853, -0.14075148105621338, -1.3874461650848389, 0.28156477212905884, -0.9179959893226624, -1.2968281507492065, 0.19895713031291962, -0.1636660248041153, 0.09745144098997116, -0.4712808132171631, -1.2986916303634644, -1.3279190063476562, -0.6729908585548401, 0.21538715064525604, 0.0033319147769361734, -0.3368145227432251, -0.4880366027355194, -1.6056562662124634, -0.06749758124351501, 0.537982702255249, -1.518497109413147, -0.14922605454921722, 1.2041049003601074, 0.08387433737516403, 1.0984026193618774, -0.27249911427497864, -0.6649801731109619, -0.6570414304733276, -0.3877118229866028, -0.819801926612854, 0.21040719747543335, 2.10652232170105, -0.1445361226797104, 0.8136182427406311, -0.6583006978034973, -0.3472365438938141, -1.0367175340652466, 1.103456974029541, -0.19918836653232574, 0.05155167728662491, 1.4965381622314453, 0.9884462356567383, 1.1312751770019531, -0.5285311341285706, -0.37252330780029297, -0.5133379697799683, 0.3439998924732208, -0.6833076477050781, -1.3308806419372559, -0.0043663098476827145, 1.0170615911483765, -0.2980661988258362, -0.6847434043884277, 0.132366344332695, 1.3589580059051514, -0.03737796097993851, -0.5342410802841187, -0.7537646293640137, -1.2974865436553955, -0.5541737079620361, -0.4874458611011505, 1.8318110704421997, 0.20764854550361633, 1.5493804216384888, };
+PI_L2 MB_TYPE B_mat[] = {-0.8827683925628662, -0.5542697310447693, 0.14389587938785553, -1.975010633468628, 0.20784620940685272, 0.013685223646461964, -0.9441406726837158, 0.2295272946357727, 0.18610191345214844, -0.022595224902033806, -2.596245288848877, 0.2539166808128357, 0.181132510304451, 0.5449540615081787, -0.025274118408560753, -0.07331778109073639, -0.7438264489173889, -0.23840615153312683, 1.428622841835022, -0.34748250246047974, -1.5984361171722412, 0.20564231276512146, -1.1283481121063232, -0.9706998467445374, 1.064945936203003, 0.5016964077949524, 1.2105423212051392, -0.5837512612342834, -0.983349084854126, 1.118739128112793, -0.43573465943336487, -0.2066168338060379, 2.088926315307617, 0.21492750942707062, 0.019352668896317482, -1.092134952545166, 2.3897173404693604, 0.4519634544849396, -1.5889497995376587, -1.6419037580490112, 0.4415695369243622, -2.7164535522460938, 0.9438610076904297, -0.4197200834751129, -0.4402216672897339, -1.895782709121704, 1.4696519374847412, -0.043221864849328995, -0.6337109208106995, 0.13316427171230316, 1.1930891275405884, -1.0255396366119385, 0.17155316472053528, 0.0703289583325386, -0.42407506704330444, 0.22795960307121277, 0.6689052581787109, 1.656773567199707, -0.28926777839660645, 0.4480336606502533, -0.4898584485054016, 0.8577049374580383, -0.40668636560440063, -1.1225101947784424, };
+PI_L2 OUT_TYPE ref[] = {-0.25370487570762634, 4.288405895233154, -2.4230763912200928, 4.161197662353516, 0.1321820169687271, 1.8521130084991455, 2.2288806438446045, 2.7880353927612305, -2.160022497177124, -1.68727707862854, -1.93358314037323, 1.4045865535736084, 1.7790610790252686, -3.0093014240264893, 3.4853577613830566, 3.0664453506469727, -3.1897361278533936, 0.9710832238197327, 5.895380020141602, 0.7893388867378235, -4.211012363433838, -0.540488064289093, 1.99186372756958, 2.640082359313965, -1.1216058731079102, 4.887918949127197, -2.019449234008789, -0.1283983588218689, 0.3265409469604492, 2.1952335834503174, -1.5282257795333862, -0.5912652015686035, -1.118781328201294, -3.1169605255126953, -2.7250733375549316, 1.8199642896652222, -1.511614203453064, -2.6127007007598877, 4.342489242553711, 2.1691524982452393, -4.221386909484863, -1.1585693359375, -1.2581945657730103, -2.8258895874023438, -1.068955421447754, 0.44341498613357544, -2.514594793319702, 0.7613731622695923, -2.158013105392456, -0.041946373879909515, 2.655857563018799, 1.7298434972763062, -3.0531845092773438, -0.6418022513389587, 0.9603833556175232, 0.23947691917419434, 1.4017069339752197, -2.1417248249053955, 0.8747122287750244, 1.8828234672546387, -0.3232908248901367, -3.6543807983398438, 4.977419853210449, 0.33591794967651367, };
diff --git a/fpu_tests/matmul/FP32/main.c b/fpu_tests/matmul/FP32/main.c
new file mode 100644
index 0000000..7d12782
--- /dev/null
+++ b/fpu_tests/matmul/FP32/main.c
@@ -0,0 +1,122 @@
+// License, Version 0.51 (the "License"); you may not use this file except in
+// or agreed to in writing, software, hardware and materials distributed under
+// this License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR
+// CONDITIONS OF ANY KIND, either express or implied. See the License for the
+// specific language governing permissions and limitations under the License.
+
+#include "config.h"
+#include "pulp.h"
+
+#include <stdio.h>
+#include <stdint.h>
+#include <limits.h> /* for CHAR_BIT */
+#include <math.h>
+
+#include "data.h"
+
+#define STACK_SIZE 2048
+
+void main_fn(testresult_t *result, void (*start)(), void (*stop)());
+
+testcase_t testcases[] = {
+  { .name = "Matrix Multiplication", .test = main_fn },
+  {0, 0}
+};
+
+DATA_LOCATION MA_TYPE matA[M*N] __attribute__ ((aligned (4)));
+DATA_LOCATION MB_TYPE matB[N*P] __attribute__ ((aligned (4)));
+DATA_LOCATION OUT_TYPE matC[M*P] __attribute__ ((aligned (4)));
+
+// End of computation
+int done = 0;
+
+int retval = 0;
+
+void __attribute__ ((noinline)) matrix_init(MA_TYPE * __restrict__ A, MB_TYPE * __restrict__ B, OUT_TYPE * __restrict__ C) {
+  for (int i = 0; i < M; i++) 
+    for (int j = 0; j < N; j++){
+      A[i*N+j] = A_mat[i*N+j];
+
+
+    } 
+      
+  for (int i = 0; i < N; i++) 
+    for (int j = 0; j < P; j++){
+      B[i*P+j] = B_mat[i*P+j];
+    }
+  for (int i = 0; i < M; i++) 
+    for (int j = 0; j < P; j++)  
+      C[i*P+j] = 0;
+  
+}
+
+int __attribute ((noinline)) check_result(OUT_TYPE * __restrict__ result) {
+  #ifndef FABRIC
+    synch_barrier();
+  #endif
+
+  if(get_core_id() == 0) {
+    float diff;
+    int err = 0;
+
+    for (int i = 0; i < (M*P); i++) {
+      diff = fabs(result[i] - ref[i]);
+      if(diff > THR) {
+        err++;
+      #ifdef VERBOSE
+
+        printf("Error at index %d:\t refrence %f\t output %f\t error %.4f\n", i, ref[i], result[i], diff);
+      #endif
+      
+      }
+
+      #ifdef PRINT_RESULTS
+
+        printf("index %d:\t refrence %f\t output %f\t error %f\n", i, ref[i], result[i], diff);
+      #endif
+    }
+
+    return err;
+
+  }
+}
+
+void main_fn(testresult_t *result, void (*start)(), void (*stop)()){
+  
+  if (get_core_id() == 0)
+    matrix_init(matA, matB, matC);
+ 
+  #ifndef FABRIC
+  synch_barrier();
+  #endif
+  
+  #ifdef STATS
+  start();
+  #endif
+  matMul(matA, matB, matC, M, N, P);
+
+  #ifdef STATS
+  stop();
+  #endif
+
+  #ifdef CHECK
+  result->errors = check_result(matC);
+  #endif
+};
+
+int main()
+{
+  #ifdef FABRIC
+      main_fn();
+  #else
+  if (rt_cluster_id() != 0)
+    return bench_cluster_forward(0);
+
+  int nbErrors = run_suite(testcases);
+
+  synch_barrier();
+  #endif
+  retval = nbErrors;
+
+  return retval;
+}
diff --git a/fpu_tests/matmul/FP32/matmul.c b/fpu_tests/matmul/FP32/matmul.c
new file mode 100644
index 0000000..0a616ca
--- /dev/null
+++ b/fpu_tests/matmul/FP32/matmul.c
@@ -0,0 +1,103 @@
+#include "config.h"
+#include "pulp.h"
+
+#ifdef VECTORIAL
+
+void __attribute__ ((noinline)) matMul(MA_TYPE * __restrict__ A, MB_TYPE * __restrict__ B, OUT_TYPE * __restrict__ C, int M, int N, int P){
+
+  OUT_VTYPE temp;
+  MA_VTYPE Av;
+  MB_VTYPE Bv0;
+  MB_VTYPE Bv1;
+  OUT_VTYPE *Cv;
+  int blockSize = (M+NUM_CORES-1)/NUM_CORES;
+  int start = get_core_id()*blockSize;
+  int end = start + blockSize < M? start + blockSize : M;
+
+  for (int i = start; i < end; i++) {
+    for (int j=0; j < (P & 0xfffffffe); j+=2) {
+        
+      temp = (OUT_VTYPE) {0, 0};
+
+			// Manual unrolling
+			for (int k=0; k<(N & 0xfffffffe); k+=2){
+				Av  = *((MA_VTYPE *) &A[i*N+k]);
+				Bv0 = *((MB_VTYPE *) &B[k*P+j]);
+				Bv1 = *((MB_VTYPE *) &B[k*P+j+P]);
+				temp += (OUT_VTYPE)(__builtin_shuffle(Av, (v2s){0,0})) * Bv0;
+				temp += (OUT_VTYPE)(__builtin_shuffle(Av, (v2s){1,1})) * Bv1;
+			}
+
+    if (N & 0x00000001) 
+     {
+      temp[0] += A[i*N+N-1] * B[(N-1)*P+j];
+      temp[1] += A[i*N+N-1] * B[(N-1)*P+j+1];
+     }
+			Cv = (OUT_VTYPE *) &C[i*P+j];
+
+			*Cv = temp;
+    }
+  }
+  /// Leftover in P
+    if (P & 0x00000001) 
+    {
+      for (int i = start; i < end; i++) {
+
+            OUT_TYPE temp1 = 0;
+
+            // Manual unrolling
+            for (int k=0; k<(N & 0xfffffffe); k+=2){
+              temp1 += A[i*N+k] * B[k*P+P-1];
+              temp1 += A[i*N+k+1] * B[k*P+P-1+P];
+            }
+          if (N & 0x00000001) 
+               {
+                temp1 += A[i*N+N-1] * B[(N-1)*P+P-1];
+               }
+            C[i*P+(P-1)]=temp1;
+          }
+      }
+      
+  #if NUM_CORES > 1
+  synch_barrier();
+  #endif  
+}
+#else
+
+void __attribute__ ((noinline)) matMul(MA_TYPE * __restrict__ A, MB_TYPE * __restrict__ B, OUT_TYPE * __restrict__ C, int M, int N, int P) {
+
+  int blockSize = (M+NUM_CORES-1)/NUM_CORES;
+  int start = get_core_id()*blockSize;
+  int end = start + blockSize < M? start + blockSize : M;
+  
+  for (int i = start; i < end; i++) {
+    for (int j = 0; j < P; j++) {
+      OUT_TYPE temp = 0;
+
+      //Manual unrolling
+      for (int k = 0; k < (N & 0xfffffffe); k+=2) {
+        temp += (OUT_TYPE)(A[i*N+k]   * B[k*P+j]);
+        temp += (OUT_TYPE)(A[i*N+k+1] * B[k*P+j+P]);
+
+      }
+         C[i*P+j] = (OUT_TYPE)(temp);
+    }
+  }
+      // Leftover on N
+
+      if (N & 0x00000001) 
+      {
+        for (int i=start; i<end; i++) 
+        {
+          for (int j=0; j<P; j++) 
+          {
+            C[i*P+j] += (OUT_TYPE)(A[i*N+N-1] * B[(N-1)*P+j]);
+          }
+        }
+      } 
+
+  #if NUM_CORES > 1
+  synch_barrier();
+  #endif
+}
+#endif
diff --git a/fpu_tests/matmul/FP32/out.txt b/fpu_tests/matmul/FP32/out.txt
new file mode 100644
index 0000000..9950041
--- /dev/null
+++ b/fpu_tests/matmul/FP32/out.txt
@@ -0,0 +1,15 @@
+GNU Make 3.82
+Built for x86_64-redhat-linux-gnu
+Copyright (C) 2010  Free Software Foundation, Inc.
+License GPLv3+: GNU GPL version 3 or later <http://gnu.org/licenses/gpl.html>
+This is free software: you are free to change and redistribute it.
+There is NO WARRANTY, to the extent permitted by law.
+Lettura dei makefile...
+/scratch2/rtedeschi/pulp_cluster/alsaqr_cluster/pulp-runtime/rules/pulpos/targets/pulp_cluster.mk:74: warning: overriding recipe for target `run'
+/scratch2/rtedeschi/pulp_cluster/alsaqr_cluster/pulp-runtime/rules/pulpos/default_rules.mk:322: warning: ignoring old recipe for target `run'
+Aggiornamento degli obbiettivi....
+ Il file «all» non esiste.
+   Il file «/scratch2/rtedeschi/pulp_cluster/alsaqr_cluster/regression-tests/fpu_tests/matmul/FP32/build/fp_matmul/fp_matmul» non esiste.
+     Il file «/scratch2/rtedeschi/pulp_cluster/alsaqr_cluster/regression-tests/fpu_tests/matmul/FP32/build/fp_matmul/support_func.o» non esiste.
+    L'obiettivo «/scratch2/rtedeschi/pulp_cluster/alsaqr_cluster/regression-tests/fpu_tests/matmul/FP32/build/fp_matmul/support_func.o» deve essere rigenerato.
+make: *** Nessuna regola per generare l'obiettivo «/scratch2/rtedeschi/pulp_cluster/alsaqr_cluster/regression-tests/fpu_tests/matmul/FP32/build/fp_matmul/support_func.o», necessario per «/scratch2/rtedeschi/pulp_cluster/alsaqr_cluster/regression-tests/fpu_tests/matmul/FP32/build/fp_matmul/fp_matmul».  Stop.
diff --git a/fpu_tests/matmul/FP32/support_func.c b/fpu_tests/matmul/FP32/support_func.c
new file mode 100644
index 0000000..6d24d2e
--- /dev/null
+++ b/fpu_tests/matmul/FP32/support_func.c
@@ -0,0 +1,15 @@
+#include "config.h"
+#include "pulp.h"
+double __extendohfdf2(float16alt value)
+{
+  float result;
+  __asm__ __volatile__ ("fcvt.s.ah %0, %1": "=f"(result): "f"(value) :);
+  return (double) result;
+}
+
+double __extendhfdf2(float16 value)
+{
+  float result;
+  __asm__ __volatile__ ("fcvt.s.h %0, %1": "=f"(result): "f"(value) :);
+  return (double) result;
+}
diff --git a/fpu_tests/matmul/README.md b/fpu_tests/matmul/README.md
new file mode 100644
index 0000000..538deba
--- /dev/null
+++ b/fpu_tests/matmul/README.md
@@ -0,0 +1,38 @@
+# MatMul test
+This test performs a matrix multiplication on FP32/FP16/FP16ALT data and can also be used to measure performances.
+In this folder you can find pre-generated golden models.
+
+## Running a test
+After the platform and the SDK setup you can run the test:
+
+~~~~~shell
+make clean all [platform=rtl] run
+~~~~~
+
+If you want to run this test on RTL, remember to specify the platform which is gvsoc by default.
+There are several flags useful to activate some functionalities:
+
+- `cores=N_CORES` set the number of cores used for the execution to `N_CORES`, by default `cores=1`. There is also the ability to run on the Fabric controller by using `FABRIC=1` instead of `cores=N_CORE`.
+- `fmt=FP_FMT` specifies the floating-point format for data, by deafult it is set to `FP32` but you can also choose `FP16` or `FP16ALT` formats. **For this application you can use mixed-precision in the C code by using `fmt_A=FP_A   fmt_B=FP_B   fmt_OUT=FP_OUT` instead of `fmt`.**
+- `vec=1` activates vectorial format **only for half-precision floating point (FP16 and FP16ALT)**
+- `check=1` activates the result check
+- `verbose=1` prints the wrong results
+- `stats=1` activates performance measurement
+- `PRINT_RESULTS=1` print outputs of C code
+
+
+## Generating the golden model
+If you want to re-generate a golden model, you can use the [data_generator.py](./data_generator.py) script with the following command:
+
+~~~~~shell
+./data_generator.py --M=m --N=n --P=p --float_type=fmt --MAC_flag=MAC_FLAG
+~~~~~
+- specifies the floating-point format for data, by deafult it is set to `FP32` but you can also choose `FP16` and `FP16ALT` formats. **Also, you can run the mixed-precision golden model by using `--float_type=FP_A,FP_B,FP_OUT`.**
+- `MAC_flag` is used to emulate the multiply-and-add operator available on most DSP instruction sets for embedded devices. It can be true or false. To emulate `FP16` and `FP16ALT` behavior on PULP, true this flag.
+
+The script will generate three floating-point matrices fo format `fmt` (FP32/FP16/FP16ALT):
+- A_mat[m,n] input matrix
+- B_mat[n,p] input matrix
+- ref[m,p] output matrix
+
+The generated header file will be written in the [references](./references) folder.
diff --git a/fpu_tests/matmul/data_generator.py b/fpu_tests/matmul/data_generator.py
new file mode 100755
index 0000000..26905a2
--- /dev/null
+++ b/fpu_tests/matmul/data_generator.py
@@ -0,0 +1,206 @@
+#!/bin/python3
+
+import os
+import argparse
+import sys
+
+import torch
+from torch import nn
+
+
+def relative_absolute_error(true, pred):
+    true_mean = torch.mean(true)
+    squared_error_num = torch.sum(torch.abs(true - pred))
+    squared_error_den = torch.sum(torch.abs(true - true_mean))
+    rae_loss = squared_error_num / squared_error_den
+    return rae_loss
+
+
+def mean_squared_error(true, pred):
+    squared_error = torch.square(true - pred)
+    sum_squared_error = torch.sum(squared_error)
+    size = true.size(dim=0) * true.size(dim=1)
+    mse_loss = sum_squared_error / size
+    return mse_loss
+
+
+def matrix_init(IN, dt):
+    temp = torch.zeros((IN.shape[0], IN.shape[1]), dtype=dt)
+    # iterate through rows of IN
+    for i in range(IN.shape[0]):
+        # iterate through columns of IN
+        for j in range(IN.shape[1]):
+            temp[i][j] = IN[i][j]
+    return temp
+
+
+def error_metric(ref, res):
+
+    # calculate manually because metrics doesn't supprt bfloat16
+    d = ref - res
+    mse_f = torch.mean(d**2)
+    mae_f = torch.mean(abs(d))
+    rmse_f = torch.sqrt(mse_f)
+    r2_f = 1-(torch.sum(d**2)/torch.sum((ref-torch.mean(ref))**2))
+    print("Results of metrics:")
+    print("MAE:",mae_f.item())
+    print("MSE:", mse_f.item())
+    print("RMSE:", rmse_f.item())
+    print("R-Squared:", r2_f.item())
+    rae = relative_absolute_error(ref, res)
+    print("RAE is", rae.item())
+
+
+def matrix_mult(Xs, Ys, dt, mac_flag, cast_flag, cast_to):
+    Rs = torch.zeros((Xs.shape[0], Ys.shape[1]), dtype=dt)
+    # iterate through rows of X
+    for i in range(Xs.shape[0]):
+        # iterate through columns of Y
+        for j in range(Ys.shape[1]):
+            temp = torch.tensor([0], dtype=dt)
+            # iterate through rows of Y
+            for k in range(Ys.shape[0]):
+                a = Xs[i][k]
+                b = Ys[k][j]
+                if cast_flag == "true":
+                    if cast_to == "FP16":
+                        a = a.type(torch.float16)
+                        b = b.type(torch.float16)
+                    elif cast_to == "FP16ALT":
+                        a = a.type(torch.bfloat16)
+                        b = b.type(torch.bfloat16)
+                if mac_flag == "true":
+                    a = a.type(torch.float32)
+                    b = b.type(torch.float32)
+                    temp = temp.type(torch.float32)
+                temp += a * b
+                if mac_flag == "true":
+                    temp = temp.type(dt)
+
+            Rs[i][j] = temp
+    return Rs
+
+
+def write_matrix(matrix_to_write, name, file_pointer, float_type):
+    matrix_string = ''
+    sz0 = matrix_to_write.size()[0]
+    sz1 = matrix_to_write.size()[1]
+    if 'ref' in name:
+        file_pointer.write("PI_L2 OUT_TYPE %s[] = {" % name)
+    elif 'A_mat' in name:
+        file_pointer.write("PI_L2 MA_TYPE %s[] = {" % name)
+    else:
+        file_pointer.write("PI_L2 MB_TYPE %s[] = {" % name)
+    if float_type == torch.float32:
+        name = ")"
+    elif float_type == torch.float16:
+        name = ", dtype=torch.float16)"
+    elif float_type == torch.bfloat16:
+        name = ", dtype=torch.bfloat16)"
+    for i in range(sz0):
+        for j in range(sz1):
+            matrix_string += str(matrix_to_write[i][j].item()).replace('tensor(', '').replace(name, '')
+            matrix_string += ', '
+    file_pointer.write("%s" % matrix_string)
+    file_pointer.write("};\n")
+
+
+def get_inital_config():
+    # get arguments  and data format
+    parser = argparse.ArgumentParser()
+    parser.add_argument('--M')
+    parser.add_argument('--N')
+    parser.add_argument('--P')
+
+    parser.add_argument('--MAC_flag', default="true")
+    parser.add_argument('--float_type', default='FP32')
+    args = parser.parse_args()
+
+    M = int(args.M)
+    N = int(args.N)
+    P = int(args.P)
+    mac_flag = str(args.MAC_flag)
+    bits = args.float_type.split(",")
+    return M, N, P, bits, mac_flag
+
+
+def select_dtypes(user_dtypes, num_param):
+    types_dict = {
+        "FP32": torch.float32,
+        "FP16": torch.float16,
+        "FP16ALT": torch.bfloat16
+    }
+    dtypes = []
+    if len(user_dtypes) == 1:
+        for i in range(num_param):
+            dtypes.append(types_dict[user_dtypes[0]])
+    elif len(user_dtypes) == num_param:
+        for i in range(num_param):
+            dtypes.append(types_dict[user_dtypes[i]])
+    else:
+        for i in range(len(user_dtypes)):
+            dtypes.append(types_dict[user_dtypes[i]])
+        if 'FP32' in user_dtypes:
+            for i in range(len(user_dtypes), num_param):
+                dtypes.append(types_dict["FP32"])
+        elif 'FP16' in user_dtypes:
+            for i in range(len(user_dtypes), num_param):
+                dtypes.append(types_dict["FP16"])
+        else:
+            for i in range(len(user_dtypes), num_param):
+                dtypes.append(types_dict["FP16ALT"])
+    return dtypes
+
+def check_cast(datatypes):
+    result = len(set(datatypes)) == 1  
+    if result : #All Elements in List are Equal
+        return "false"
+    else: #All Elements in List are Not Equal
+        if torch.float32 in datatypes:
+            return "false"
+        else:
+            return "true"
+
+def save_data_into_hfile(M, N, P, A_mat, B_mat, res):
+    # Generate header file
+    f = open('data.h', 'w')
+    f.write('\
+#define M %s\n\
+#define N %s\n\
+#define P %s\n\n' % (M, N, P))
+    write_matrix(A_mat, 'A_mat', f, A_mat.dtype)
+    write_matrix(B_mat, 'B_mat', f, B_mat.dtype)
+    write_matrix(res, 'ref', f, res.dtype)
+
+    f.close()
+
+
+def main():
+    M, N, P, bits, mac_flag = get_inital_config()
+
+    # Create reference matrices
+    A_ref = torch.randn((M, N), dtype=torch.float32)
+    B_ref = torch.randn((N, P), dtype=torch.float32)
+
+    # calculate reference output
+    ref = matrix_mult(Xs=A_ref, Ys=B_ref, dt=torch.float32, mac_flag=mac_flag, cast_flag="false",cast_to="false")
+
+    # set the data types based on the parser input
+    datatypes = select_dtypes(bits, 3)
+
+    cast_flag = check_cast(datatypes[0:2])
+    cast_to = "FP16ALT"
+    A_mat = matrix_init(A_ref, dt=datatypes[0])
+    B_mat = matrix_init(B_ref, dt=datatypes[1])
+
+    res = matrix_mult(Xs=A_mat, Ys=B_mat, dt=datatypes[2], mac_flag=mac_flag, cast_flag=cast_flag, cast_to = cast_to)
+
+    error_metric(ref, res)
+    save_data_into_hfile(M, N, P, A_mat, B_mat, res)
+    print("############################## Done! ###################################")
+    return None
+
+
+if __name__ == "__main__":
+    main()
+    pass
diff --git a/hwpe/neureka/Makefile b/hwpe/neureka/Makefile
new file mode 100644
index 0000000..22ba1d0
--- /dev/null
+++ b/hwpe/neureka/Makefile
@@ -0,0 +1,49 @@
+ACCELERATOR ?= neureka
+
+LIBDIR := $(abspath ./pulp-nnx)
+ACC_DIR := $(LIBDIR)/$(ACCELERATOR)
+
+## Test
+INC_DIRS += inc
+PULP_APP_SRCS += $(wildcard src/*.c)
+SRC_DIRS += src
+
+## Library
+INC_DIRS += $(LIBDIR)/inc $(LIBDIR)/util
+PULP_APP_SRCS += $(LIBDIR)/src/pulp_nnx_$(ACCELERATOR).c $(wildcard $(LIBDIR)/util/*.c)
+SRC_DIRS += $(LIBDIR)/src $(LIBDIR)/util
+
+## Accelerator
+INC_DIRS += $(ACC_DIR)/hal $(ACC_DIR)/bsp $(ACC_DIR)/bsp/pulp_cluster
+PULP_APP_SRCS += $(wildcard $(ACC_DIR)/hal/*.c) $(wildcard $(ACC_DIR)/bsp/pulp_cluster/*.c)
+SRC_DIRS += $(ACC_DIR)/hal $(ACC_DIR)/bsp/pulp_cluster
+
+## Generated 
+INC_DIRS += gen/inc
+SRC_DIRS += gen/src
+
+INC_FLAGS += $(addprefix -I,$(INC_DIRS))
+
+# Flags
+ACCELERATOR_UPPERCASE := $(shell echo $(ACCELERATOR) | tr [:lower:] [:upper:])
+PULP_CFLAGS += -DNNX_ACCELERATOR=\"$(ACCELERATOR)\" -DNNX_$(ACCELERATOR_UPPERCASE) -DNNX_NEUREKA_PULP_CLUSTER -DNNX_NEUREKA_PE_H=4 -DNNX_NEUREKA_PE_W=4
+PULP_CFLAGS += $(INC_FLAGS) -O3
+
+PULP_APP = test
+
+ifeq ($(no_ecc),1)
+ PULP_CFLAGS += -DNO_ECC
+endif
+
+ifeq ($(fault_inject),1)
+  export FAULT_INJECTION=1
+  export FAULT_INJECTION_SCRIPT=$(CURDIR)/pulp_inject_fault.tcl
+endif
+
+ifeq ($(multi_bit_upset),1)
+  export MULTI_BIT_UPSET=1
+else
+  export MULTI_BIT_UPSET=0
+endif
+
+include $(PULP_SDK_HOME)/install/rules/pulp_rt.mk
diff --git a/hwpe/neureka/gen/inc/bias.h b/hwpe/neureka/gen/inc/bias.h
new file mode 100644
index 0000000..30d072b
--- /dev/null
+++ b/hwpe/neureka/gen/inc/bias.h
@@ -0,0 +1,14 @@
+#ifndef __BIAS_H__
+#define __BIAS_H__
+
+#include <pmsis.h>
+
+#define BIAS_SIZE (32)
+PI_L1 int32_t bias[BIAS_SIZE] = {
+    -0x65c83656, 0x79b5657f, -0x57f742d4, 0x48370263, -0x322008d7, -0x40c1a507, -0x6cb54499, -0x76a3f065, -0x13494905, -0x17d3f6d0, 
+    -0x64732a81, 0x60b4468a, 0x1176b544, -0x1168daef, -0x40e1a4fd, -0x3746219b, 0x18afd65e, 0x6f6ff1d7, -0x61e9d6e3, 0x45b33266, 
+    0x79d3867b, 0x1c699c9e, 0x6ebc54cb, -0x7f508e3e, -0x1bb33223, 0x6e771d3c, -0x42009579, 0x2c809ab3, -0x3fe5e70c, -0x468955b7, 
+    -0x4405229, -0x79a9c840
+};
+
+#endif  // __BIAS_H__
diff --git a/hwpe/neureka/gen/inc/input.h b/hwpe/neureka/gen/inc/input.h
new file mode 100644
index 0000000..2c7c3e5
--- /dev/null
+++ b/hwpe/neureka/gen/inc/input.h
@@ -0,0 +1,65 @@
+#ifndef __INPUT_H__
+#define __INPUT_H__
+
+#include <pmsis.h>
+
+#define INPUT_SIZE (512)
+
+PI_L1 uint8_t input[INPUT_SIZE] = {
+    0xc2, 0x9c, 0xad, 0x2f, 0x37, 0xd4, 0x88, 0x59, 0x67, 0xd8, 
+    0x73, 0x01, 0x9c, 0x5f, 0xc4, 0x64, 0xd9, 0x74, 0xa9, 0xe7, 
+    0x18, 0x9a, 0x6e, 0xd7, 0x77, 0xc2, 0x71, 0xa6, 0xdc, 0x3f, 
+    0xab, 0x2b, 0x36, 0x6c, 0x87, 0x3d, 0x9b, 0x68, 0x64, 0x62, 
+    0xc4, 0x13, 0x65, 0x5c, 0x3c, 0xcf, 0xd0, 0x0a, 0x12, 0xc1, 
+    0x18, 0xb4, 0x82, 0x15, 0x77, 0x44, 0x92, 0xb6, 0x6d, 0x09, 
+    0xc7, 0x0a, 0x41, 0xc1, 0xe9, 0xfe, 0x82, 0x73, 0x1c, 0x1d, 
+    0xd4, 0xb8, 0x49, 0x81, 0x3e, 0x3a, 0xa7, 0x32, 0x41, 0x19, 
+    0xcb, 0x37, 0xc8, 0x56, 0x72, 0x34, 0x91, 0x11, 0xa4, 0x99, 
+    0x9d, 0xc5, 0x7a, 0x50, 0x22, 0x65, 0x87, 0xc6, 0xe1, 0x95, 
+    0xe8, 0x3a, 0x61, 0xdb, 0x01, 0xb3, 0x56, 0x46, 0xa9, 0xdd, 
+    0x21, 0xb2, 0xb9, 0xbf, 0x93, 0x3f, 0x33, 0x87, 0x86, 0x65, 
+    0x20, 0xef, 0xdf, 0xd3, 0xdc, 0xb6, 0xe8, 0x5e, 0xdf, 0xdd, 
+    0x43, 0xad, 0xbc, 0x81, 0x6e, 0x8b, 0x61, 0x28, 0xef, 0xc6, 
+    0x5c, 0x7f, 0x01, 0x5a, 0x66, 0x52, 0x5a, 0x14, 0x0f, 0x32, 
+    0x12, 0x07, 0xb9, 0x98, 0xc8, 0x81, 0x2b, 0x67, 0x07, 0x7f, 
+    0xcc, 0x2c, 0x68, 0x1f, 0xd9, 0xda, 0xc9, 0x5a, 0x49, 0x8b, 
+    0x24, 0x03, 0x16, 0xa8, 0x5b, 0xc8, 0x4b, 0x77, 0xfd, 0x50, 
+    0x0a, 0x3d, 0x7d, 0x1e, 0x99, 0xe4, 0x5b, 0x3b, 0xb2, 0x89, 
+    0xf5, 0x12, 0x43, 0xec, 0x65, 0x1a, 0x4d, 0x13, 0x3d, 0x61, 
+    0x95, 0x48, 0x24, 0xf9, 0x0c, 0xed, 0xba, 0xad, 0x0d, 0x93, 
+    0xed, 0x35, 0x51, 0x30, 0x48, 0x11, 0x04, 0x5b, 0xcd, 0xa7, 
+    0xc1, 0x1d, 0x5c, 0xd3, 0xa3, 0xd1, 0x67, 0xb9, 0x14, 0xdf, 
+    0x96, 0x54, 0x5b, 0x07, 0x27, 0x39, 0xef, 0x81, 0x75, 0x0b, 
+    0x2a, 0x5b, 0x78, 0x3d, 0x19, 0x5e, 0x20, 0x69, 0x38, 0xc8, 
+    0x68, 0x4a, 0x2d, 0x39, 0xb5, 0x93, 0x8e, 0xb3, 0x8f, 0x17, 
+    0xc3, 0xb8, 0x59, 0x2d, 0x9b, 0x2a, 0x4c, 0x93, 0x13, 0xe9, 
+    0x25, 0x52, 0xf4, 0x6d, 0x9e, 0x84, 0x9c, 0x47, 0x29, 0xdc, 
+    0xa5, 0x0d, 0x27, 0xf1, 0x74, 0x51, 0x0b, 0x1e, 0xf3, 0xce, 
+    0x7b, 0xc4, 0x19, 0x71, 0x2d, 0x58, 0x3b, 0x1f, 0x46, 0xd7, 
+    0xe9, 0x4e, 0xe8, 0x2a, 0x7f, 0x30, 0x06, 0xbb, 0x0a, 0x33, 
+    0x5e, 0xce, 0x69, 0x46, 0x08, 0x44, 0xca, 0xb2, 0x55, 0xab, 
+    0x65, 0x6d, 0xe2, 0xce, 0xe4, 0xbe, 0x99, 0x76, 0x3c, 0xd1, 
+    0x96, 0x71, 0xa5, 0x24, 0x48, 0x69, 0xee, 0xe9, 0x10, 0xf5, 
+    0x69, 0xbe, 0x66, 0x5e, 0x87, 0x23, 0x2f, 0xe1, 0xdc, 0x05, 
+    0x28, 0x2b, 0x92, 0x3b, 0x04, 0x2f, 0x56, 0xb2, 0x38, 0x70, 
+    0x5e, 0x6d, 0xca, 0x08, 0x5b, 0x23, 0x30, 0xf0, 0x34, 0xd1, 
+    0xc7, 0x7a, 0x56, 0xb8, 0x5a, 0x57, 0x01, 0xe4, 0xc8, 0x80, 
+    0x7a, 0x2e, 0xed, 0xee, 0x51, 0xf9, 0x10, 0x8b, 0x54, 0x61, 
+    0x20, 0x3a, 0x90, 0x0b, 0xaa, 0x65, 0x81, 0xc6, 0x99, 0x40, 
+    0x23, 0xc2, 0x24, 0xf2, 0x61, 0x8c, 0x42, 0x06, 0xb4, 0xa3, 
+    0x0f, 0xed, 0x6b, 0xd5, 0x2d, 0xb5, 0x09, 0x96, 0xe3, 0x5b, 
+    0x2e, 0x3a, 0x62, 0xd7, 0x7f, 0x29, 0x03, 0x8a, 0x26, 0x71, 
+    0x10, 0xae, 0x28, 0x3b, 0x1a, 0xf3, 0x9d, 0xf6, 0x37, 0x23, 
+    0x84, 0x3c, 0x1f, 0x13, 0xdf, 0x74, 0x63, 0xa2, 0xb2, 0x88, 
+    0xee, 0xf3, 0x6a, 0xcd, 0x8c, 0xa9, 0x5a, 0xf5, 0xc9, 0xc8, 
+    0x53, 0xc8, 0xd4, 0x58, 0x44, 0x27, 0x0e, 0x3d, 0xb0, 0x72, 
+    0xd2, 0x3d, 0x31, 0xf3, 0x76, 0xf2, 0x83, 0x03, 0xfc, 0xd1, 
+    0x29, 0xc0, 0x19, 0x81, 0xb9, 0x65, 0x8b, 0xf4, 0xe1, 0x8f, 
+    0xf3, 0x80, 0xc2, 0xe1, 0xb2, 0xd0, 0x27, 0x9a, 0x0d, 0xd7, 
+    0xf5, 0x46, 0x6d, 0x37, 0xdf, 0x16, 0xe5, 0x40, 0x63, 0xac, 
+    0xdd, 0xe2
+};
+
+
+
+#endif  // __INPUT_H__
diff --git a/hwpe/neureka/gen/inc/layer_conf.h b/hwpe/neureka/gen/inc/layer_conf.h
new file mode 100644
index 0000000..42e2de0
--- /dev/null
+++ b/hwpe/neureka/gen/inc/layer_conf.h
@@ -0,0 +1,42 @@
+#ifndef __LAYER_CONF_H__
+#define __LAYER_CONF_H__
+
+#define TEST_NAME "test"
+#define INPUT_HEIGHT (4)
+#define INPUT_WIDTH (4)
+#define INPUT_CHANNEL (32)
+#define INPUT_SIGNED (0)
+#define INPUT_BITS (8)
+
+#define OUTPUT_HEIGHT (4)
+#define OUTPUT_WIDTH (4)
+#define OUTPUT_CHANNEL (32)
+#define OUTPUT_BITS (8)
+
+#define WEIGHT_HEIGHT (1)
+#define WEIGHT_WIDTH (1)
+#define WEIGHT_CHANNEL_IN (32)
+#define WEIGHT_CHANNEL_OUT (32)
+#define WEIGHT_BITS (8)
+#define WEIGHT_OFFSET (-128)
+
+#define SCALE_BITS (8)
+
+#define BIAS_BITS (32)
+
+#define PADDING_TOP (0)
+#define PADDING_BOTTOM (0)
+#define PADDING_LEFT (0)
+#define PADDING_RIGHT (0)
+#define PADDING_VALUE (0)
+
+#define STRIDE_HEIGHT (1)
+#define STRIDE_WIDTH (1)
+
+#define GROUPS (1)
+#define OUTSHIFT (23)
+#define HAS_NORM_QUANT (1)
+#define HAS_BIAS (1)
+#define HAS_RELU (1)
+
+#endif  // __LAYER_CONF_H__
diff --git a/hwpe/neureka/gen/inc/output.h b/hwpe/neureka/gen/inc/output.h
new file mode 100644
index 0000000..472a3bb
--- /dev/null
+++ b/hwpe/neureka/gen/inc/output.h
@@ -0,0 +1,10 @@
+#ifndef __OUTPUT_H__
+#define __OUTPUT_H__
+
+#include <pmsis.h>
+
+#define OUTPUT_SIZE (512)
+PI_L1 uint8_t output[OUTPUT_SIZE];
+
+
+#endif  // __OUTPUT_H__
diff --git a/hwpe/neureka/gen/inc/scale.h b/hwpe/neureka/gen/inc/scale.h
new file mode 100644
index 0000000..8ae79a5
--- /dev/null
+++ b/hwpe/neureka/gen/inc/scale.h
@@ -0,0 +1,15 @@
+#ifndef __SCALE_H__
+#define __SCALE_H__
+
+#include <pmsis.h>
+
+#define SCALE_SIZE (32)
+PI_L1 uint8_t scale[SCALE_SIZE] = {
+    0x2c, 0xca, 0xa6, 0xb9, 0x5f, 0xfb, 0xb4, 0x1e, 0x98, 0x1b, 
+    0xf1, 0xdf, 0xdd, 0xcf, 0xde, 0xf4, 0x82, 0xfd, 0x38, 0x59, 
+    0x2b, 0xa6, 0x10, 0x67, 0xa4, 0x2a, 0xf7, 0x12, 0x89, 0xcc, 
+    0x53, 0x1c
+};
+
+
+#endif  // __SCALE_H__
diff --git a/hwpe/neureka/gen/inc/weight.h b/hwpe/neureka/gen/inc/weight.h
new file mode 100644
index 0000000..dfa4788
--- /dev/null
+++ b/hwpe/neureka/gen/inc/weight.h
@@ -0,0 +1,115 @@
+#ifndef __WEIGHT_H__
+#define __WEIGHT_H__
+
+#include <pmsis.h>
+
+#define WEIGHT_SIZE (1024)
+PI_L1 uint8_t weight[WEIGHT_SIZE] = {
+    0x6e, 0xe7, 0xcc, 0xf1, 0x1d, 0x56, 0xfe, 0xa4, 0xdd, 0x6c, 
+    0x22, 0xe7, 0x2b, 0x7d, 0x5a, 0xc1, 0x16, 0x58, 0x21, 0xc1, 
+    0x16, 0x95, 0x10, 0x54, 0x9c, 0xdd, 0xc9, 0x66, 0x89, 0x93, 
+    0xc7, 0xe1, 0x73, 0x89, 0xad, 0x4d, 0x48, 0x2e, 0xe5, 0x3e, 
+    0x29, 0x36, 0x3c, 0xae, 0xc3, 0xa4, 0x4a, 0x2b, 0x00, 0x54, 
+    0x42, 0xed, 0x55, 0x7d, 0x54, 0x94, 0x70, 0x1e, 0x3c, 0x46, 
+    0x86, 0x01, 0xd3, 0xa4, 0xe2, 0x52, 0x0c, 0x87, 0x0b, 0x5e, 
+    0x02, 0xfb, 0xcd, 0x76, 0xb6, 0x83, 0xc5, 0x9c, 0xf1, 0x6a, 
+    0x7f, 0x5f, 0xc0, 0x9c, 0x13, 0xde, 0xdd, 0x70, 0xb6, 0x38, 
+    0x98, 0x87, 0xe5, 0x71, 0xc7, 0x37, 0x9b, 0x47, 0x54, 0xfa, 
+    0xeb, 0xc7, 0x72, 0x0f, 0x61, 0x72, 0xff, 0x63, 0xc5, 0x59, 
+    0xb5, 0xeb, 0x0f, 0x38, 0xff, 0x68, 0x18, 0x4e, 0xdb, 0x0a, 
+    0x0e, 0xc6, 0x77, 0x17, 0xa5, 0x2e, 0x36, 0x00, 0x8e, 0xe2, 
+    0x41, 0x4f, 0x1e, 0x13, 0x83, 0x6a, 0x9b, 0x08, 0xc1, 0x63, 
+    0x4e, 0x14, 0xd8, 0xd0, 0x00, 0x8b, 0x06, 0x30, 0xcc, 0x11, 
+    0x44, 0x09, 0x1e, 0x9a, 0x07, 0xde, 0x4e, 0x6d, 0x45, 0x36, 
+    0xdd, 0x98, 0x27, 0x32, 0x60, 0xfa, 0xfb, 0x7b, 0x29, 0x73, 
+    0xe6, 0x6c, 0xec, 0xa1, 0x3f, 0x36, 0x9b, 0x8c, 0xa4, 0x03, 
+    0x08, 0x60, 0x2b, 0x06, 0xff, 0x89, 0x4d, 0x26, 0x90, 0x0e, 
+    0x34, 0xcf, 0xf4, 0x1a, 0xd6, 0xa6, 0x55, 0x06, 0xdf, 0x83, 
+    0x46, 0x0b, 0xec, 0x7a, 0x27, 0xb4, 0x88, 0x8e, 0xae, 0xa4, 
+    0xad, 0xcd, 0x3e, 0x2a, 0xc7, 0xba, 0xb6, 0x1b, 0x1f, 0xe3, 
+    0x26, 0x65, 0xc8, 0x2a, 0x0b, 0xaa, 0xfa, 0xf9, 0x5c, 0xe3, 
+    0x07, 0x50, 0xc9, 0x51, 0x26, 0x10, 0x83, 0xa6, 0x3e, 0x54, 
+    0x7b, 0xe3, 0x46, 0x79, 0x9a, 0xbe, 0x25, 0xe9, 0x83, 0x2b, 
+    0xbb, 0xec, 0x29, 0x7c, 0xe9, 0xc7, 0x28, 0x99, 0x1d, 0x7f, 
+    0xb3, 0x1b, 0x01, 0xa6, 0xe2, 0xed, 0x8b, 0xce, 0x0e, 0xe1, 
+    0x8a, 0x93, 0x44, 0x13, 0xae, 0xee, 0x19, 0x7d, 0x87, 0xea, 
+    0xd7, 0x70, 0x9f, 0x3c, 0xbf, 0x2f, 0x5c, 0x60, 0x10, 0xae, 
+    0x4b, 0x6b, 0x2d, 0x7e, 0xe5, 0x8a, 0xa9, 0xa0, 0xf9, 0x8e, 
+    0x4f, 0x98, 0x19, 0xf9, 0x3e, 0x4b, 0x0d, 0xf0, 0x81, 0xe0, 
+    0x52, 0xa8, 0xb5, 0x29, 0x40, 0x0d, 0x71, 0x90, 0x39, 0xe8, 
+    0x7d, 0x15, 0xf7, 0x21, 0x59, 0x34, 0xf3, 0xfb, 0xc2, 0x51, 
+    0x37, 0xee, 0x78, 0x32, 0x06, 0xdd, 0xfb, 0xf4, 0x39, 0xb7, 
+    0x96, 0x1f, 0x17, 0x94, 0x36, 0x2e, 0xb7, 0xd2, 0x6b, 0xa3, 
+    0x5d, 0x35, 0xcb, 0x76, 0xb2, 0xa9, 0x01, 0xaa, 0xd1, 0xd1, 
+    0x8a, 0xb7, 0x86, 0x17, 0x21, 0x18, 0xba, 0xa2, 0x38, 0xbe, 
+    0xc9, 0x64, 0x2d, 0xf4, 0x81, 0xc2, 0xd1, 0x97, 0xbb, 0xcd, 
+    0x04, 0x50, 0x20, 0xbb, 0x67, 0x20, 0xb4, 0x8a, 0x96, 0x30, 
+    0x8f, 0x35, 0xa4, 0xb0, 0xc2, 0x8c, 0x4e, 0x9c, 0xc7, 0x2d, 
+    0xbb, 0x90, 0x2d, 0x4a, 0x96, 0xbb, 0xb9, 0x71, 0x35, 0x37, 
+    0xe7, 0xc8, 0x54, 0x1c, 0x91, 0x15, 0xb7, 0x46, 0x28, 0x3e, 
+    0x90, 0x85, 0xa6, 0x09, 0xba, 0xd4, 0x5a, 0x51, 0x5d, 0x4a, 
+    0xb6, 0xa0, 0xb1, 0x84, 0x12, 0x9d, 0x85, 0x6d, 0xa7, 0x07, 
+    0x7d, 0x3b, 0x24, 0x44, 0x61, 0xb7, 0x78, 0x41, 0xba, 0x29, 
+    0x8b, 0xc9, 0x67, 0x2e, 0x54, 0x72, 0x79, 0x11, 0x61, 0xe9, 
+    0x31, 0x84, 0x21, 0x70, 0x1e, 0xcc, 0x5c, 0x03, 0x75, 0x19, 
+    0xab, 0x3d, 0xea, 0x1d, 0x15, 0xae, 0xba, 0xe5, 0xa2, 0x0b, 
+    0x73, 0xf4, 0xa3, 0x55, 0x3c, 0x81, 0x3c, 0x00, 0x30, 0xb0, 
+    0x2e, 0xca, 0x8f, 0xd8, 0xd3, 0x71, 0x2a, 0x9e, 0x40, 0xaf, 
+    0x0d, 0x20, 0xa1, 0x22, 0xf8, 0x7e, 0xa4, 0x8d, 0x5a, 0x7a, 
+    0xfc, 0x23, 0x05, 0x4c, 0xa3, 0xc1, 0x16, 0x95, 0x5e, 0xa9, 
+    0x6e, 0xbf, 0xf8, 0xaf, 0x5f, 0xe3, 0xb0, 0x22, 0x62, 0x39, 
+    0x45, 0xc9, 0x47, 0x84, 0x0e, 0x53, 0x79, 0xa2, 0xa8, 0x20, 
+    0x43, 0x65, 0x63, 0x4a, 0x6d, 0x10, 0x7c, 0x03, 0x5f, 0xd0, 
+    0xbe, 0x6f, 0x5a, 0x91, 0x41, 0xef, 0x9e, 0x06, 0xc9, 0x43, 
+    0x1b, 0xd9, 0x6f, 0xe5, 0x69, 0xac, 0xb2, 0xe4, 0x35, 0xea, 
+    0x8a, 0x8b, 0xb1, 0x5c, 0x20, 0x53, 0x38, 0x40, 0x29, 0xc6, 
+    0x0e, 0x64, 0xa3, 0xf7, 0xbf, 0x75, 0xad, 0x84, 0x6d, 0x9b, 
+    0xdd, 0x17, 0x10, 0x3e, 0x85, 0x6a, 0xc2, 0xc2, 0xc1, 0x22, 
+    0xb0, 0xf7, 0xae, 0x65, 0x43, 0x72, 0x4a, 0xbd, 0xeb, 0xe0, 
+    0x4e, 0xf9, 0x88, 0xa7, 0x00, 0x02, 0x8d, 0xbc, 0x05, 0x2f, 
+    0x6c, 0x8d, 0xa5, 0xb6, 0x02, 0x64, 0x32, 0x55, 0x1d, 0xde, 
+    0x76, 0xc6, 0x54, 0x95, 0xf9, 0xe7, 0xa5, 0x71, 0x66, 0xc0, 
+    0xe6, 0xab, 0xd7, 0xb7, 0x03, 0xa1, 0x38, 0xc4, 0x6f, 0x90, 
+    0x0a, 0x87, 0x25, 0x39, 0x42, 0x8f, 0xc7, 0x6b, 0x62, 0x2e, 
+    0x13, 0xde, 0x07, 0x1e, 0xa7, 0x00, 0xf0, 0x8d, 0xeb, 0xd6, 
+    0xc2, 0x9e, 0x20, 0x58, 0xe7, 0x7f, 0xc6, 0x20, 0xd6, 0xb1, 
+    0xfe, 0xf3, 0xee, 0x34, 0x9c, 0x14, 0x1e, 0xf7, 0xa4, 0x18, 
+    0x2c, 0xdd, 0xfb, 0xde, 0x4a, 0x95, 0xf8, 0xb0, 0x73, 0x14, 
+    0x09, 0x89, 0x75, 0x4c, 0xe8, 0xc8, 0x00, 0x73, 0xdc, 0x49, 
+    0x90, 0x58, 0x15, 0x53, 0x22, 0xca, 0x16, 0xc5, 0x49, 0xd5, 
+    0x2b, 0x98, 0x2f, 0x05, 0x2d, 0x8d, 0x6b, 0x0a, 0x0e, 0xac, 
+    0xfb, 0x7d, 0x42, 0x12, 0x83, 0xe8, 0x8d, 0x72, 0xa3, 0xf2, 
+    0xf7, 0x04, 0xc6, 0x87, 0x97, 0xa5, 0x82, 0x5b, 0x85, 0xea, 
+    0x8f, 0xd3, 0x68, 0x74, 0x3e, 0xf7, 0x36, 0x0c, 0xa0, 0x83, 
+    0x51, 0x68, 0xf9, 0xfb, 0xbe, 0x10, 0xde, 0xad, 0xec, 0x91, 
+    0xea, 0x12, 0x2e, 0x4a, 0xf5, 0x35, 0x03, 0x50, 0x3b, 0x5f, 
+    0x1e, 0xdf, 0xdc, 0x95, 0x68, 0xeb, 0x87, 0x8e, 0x24, 0xa4, 
+    0x70, 0x13, 0x9e, 0x09, 0x50, 0xa5, 0x15, 0x78, 0x7f, 0x18, 
+    0x06, 0x61, 0x49, 0x8a, 0xb4, 0x65, 0xca, 0x33, 0xe2, 0xa7, 
+    0x57, 0x51, 0xa1, 0xcf, 0x58, 0x82, 0xa5, 0xa0, 0xab, 0xd1, 
+    0x5f, 0xf5, 0x68, 0x2c, 0xad, 0x4e, 0xe5, 0xd1, 0x96, 0x4f, 
+    0xe0, 0x51, 0x65, 0x0c, 0x44, 0xb8, 0xc6, 0x5b, 0xc8, 0x0e, 
+    0xe5, 0x2a, 0x7e, 0x74, 0xd7, 0x45, 0x7b, 0xb5, 0x03, 0x4a, 
+    0x77, 0x7f, 0x8e, 0xda, 0x26, 0x5c, 0xe3, 0x58, 0xbc, 0x36, 
+    0xa5, 0x6e, 0xa3, 0x7e, 0x87, 0xb9, 0x80, 0xd6, 0x93, 0x5e, 
+    0xf5, 0x24, 0xd4, 0x2d, 0xbc, 0x4d, 0xc8, 0xe7, 0xdc, 0x85, 
+    0x44, 0x00, 0x4b, 0x8b, 0x2d, 0x05, 0x32, 0x97, 0x4c, 0x9c, 
+    0xb9, 0xb4, 0x97, 0x97, 0x55, 0x60, 0x0b, 0x98, 0x21, 0x4f, 
+    0xa2, 0x2e, 0x4d, 0xcc, 0x5b, 0x8d, 0xde, 0x25, 0xaa, 0x0a, 
+    0xfc, 0xca, 0xd1, 0x54, 0x90, 0xbe, 0xef, 0xe6, 0x93, 0x49, 
+    0xbb, 0x33, 0xaf, 0xf9, 0x5b, 0x98, 0xc4, 0xce, 0x3c, 0x22, 
+    0xd0, 0xed, 0x5a, 0x9e, 0x01, 0xc0, 0x6a, 0x76, 0x2a, 0x74, 
+    0x8f, 0x65, 0xdd, 0xec, 0x7b, 0xd8, 0xef, 0x0e, 0xc0, 0x0d, 
+    0xdb, 0x47, 0x8b, 0x6a, 0x0d, 0xc5, 0xe5, 0x53, 0xb2, 0x22, 
+    0x84, 0x71, 0xb2, 0xf6, 0x99, 0x42, 0x14, 0x43, 0x94, 0xad, 
+    0x61, 0xbd, 0x00, 0xef, 0xf7, 0xfd, 0x20, 0xd7, 0x0c, 0x67, 
+    0xf7, 0xca, 0xdf, 0x78, 0xb0, 0x28, 0x95, 0xb9, 0x0e, 0x4c, 
+    0x3c, 0xbb, 0x2b, 0x85, 0xa3, 0x68, 0x78, 0x6f, 0x31, 0xe7, 
+    0x5f, 0x7b, 0xa7, 0x5f, 0xe7, 0x17, 0x35, 0xdd, 0x20, 0x2c, 
+    0x10, 0x3a, 0x98, 0xe3, 0xb8, 0x7a, 0x8e, 0x6f, 0xee, 0x34, 
+    0x14, 0x75, 0x19, 0xf8
+};
+
+
+
+#endif  // __WEIGHT_H__
diff --git a/hwpe/neureka/inc/ecc_check.h b/hwpe/neureka/inc/ecc_check.h
new file mode 100644
index 0000000..3f9783d
--- /dev/null
+++ b/hwpe/neureka/inc/ecc_check.h
@@ -0,0 +1,27 @@
+/*
+ * Copyright 2024 ETH Zurich and University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ *
+ * SPDX-License-Identifier: Apache-2.0
+ */
+
+#ifndef __ECC_CHECK_H__
+#define __ECC_CHECK_H__
+
+#include <stdint.h>
+
+#define ECC_REGS (4)
+extern uint32_t ecc_errs[ECC_REGS];
+
+#endif // __ECC_CHECK_H__
diff --git a/hwpe/neureka/inc/layer_util.h b/hwpe/neureka/inc/layer_util.h
new file mode 100644
index 0000000..e44ede9
--- /dev/null
+++ b/hwpe/neureka/inc/layer_util.h
@@ -0,0 +1,40 @@
+/*
+ * Luka Macan <luka.macan@unibo.it>
+ *
+ * Copyright 2023 ETH Zurich and University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ *
+ * SPDX-License-Identifier: Apache-2.0
+ */
+
+#ifndef __LAYER_UTIL_H__
+#define __LAYER_UTIL_H__
+
+#include "layer_conf.h"
+#include <pmsis.h>
+
+static void layer_info() {
+  printf("Layer info:\n"
+         " - input: (%dx%dx%d)\n"
+         " - output: (%dx%dx%d)\n"
+         " - weight: (%dx%dx%dx%d)\n"
+         " - stride: (%dx%d)\n"
+         " - padding: (%dx%dx%dx%d)\n",
+         INPUT_HEIGHT, INPUT_WIDTH, INPUT_CHANNEL, OUTPUT_HEIGHT, OUTPUT_WIDTH,
+         OUTPUT_CHANNEL, WEIGHT_CHANNEL_OUT, WEIGHT_HEIGHT, WEIGHT_WIDTH,
+         WEIGHT_CHANNEL_IN, STRIDE_HEIGHT, STRIDE_WIDTH, PADDING_TOP,
+         PADDING_BOTTOM, PADDING_LEFT, PADDING_RIGHT);
+}
+
+#endif // __LAYER_UTIL_H__
diff --git a/hwpe/neureka/inc/nnx_layer.h b/hwpe/neureka/inc/nnx_layer.h
new file mode 100644
index 0000000..cbaf4b9
--- /dev/null
+++ b/hwpe/neureka/inc/nnx_layer.h
@@ -0,0 +1,26 @@
+/*
+ * Luka Macan <luka.macan@unibo.it>
+ *
+ * Copyright 2023 ETH Zurich and University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ *
+ * SPDX-License-Identifier: Apache-2.0
+ */
+
+#ifndef __NNX_LAYER_H__
+#define __NNX_LAYER_H__
+
+void execute_nnx_layer(void *unused_args);
+
+#endif // __NNX_LAYER_H__
diff --git a/hwpe/neureka/inc/pmsis.h b/hwpe/neureka/inc/pmsis.h
new file mode 100644
index 0000000..dbb6a91
--- /dev/null
+++ b/hwpe/neureka/inc/pmsis.h
@@ -0,0 +1,11 @@
+// fake pmsis.h
+#include <pulp.h>
+#include <stdint.h>
+// fake data in L2, actually in L1!
+#ifndef PI_L1
+  #define PI_L1 __attribute__((section(".data_l1")))
+#endif
+#ifndef PI_L2
+  #define PI_L2 __attribute__((section(".data_l1")))
+#endif
+
diff --git a/hwpe/neureka/pulp-nnx b/hwpe/neureka/pulp-nnx
new file mode 160000
index 0000000..4becda2
--- /dev/null
+++ b/hwpe/neureka/pulp-nnx
@@ -0,0 +1 @@
+Subproject commit 4becda239309035887db4f7fdf54a63ac8463180
diff --git a/hwpe/neureka/pulp_inject_fault.tcl b/hwpe/neureka/pulp_inject_fault.tcl
new file mode 100644
index 0000000..fa7efb2
--- /dev/null
+++ b/hwpe/neureka/pulp_inject_fault.tcl
@@ -0,0 +1,53 @@
+# Copyright 2023 ETH Zurich and University of Bologna.
+# Solderpad Hardware License, Version 0.51, see LICENSE for details.
+# SPDX-License-Identifier: SHL-0.51
+#
+# Author: Michael Rogenmoser (michaero@iis.ee.ethz.ch)
+
+transcript quietly
+if {! [info exists ::env(VSIM_PATH)]} {error "Define VSIM_PATH"}
+set utils_base_path  [file join $::env(VSIM_PATH) scripts fault_injection_utils]
+set script_base_path [file join $::env(VSIM_PATH) fault_injection_sim scripts]
+
+set verbosity            2
+set log_injections       1
+# Easy way to generate a variable seed
+# set seed                 [clock seconds]
+# Default value
+set seed                 12345
+set print_statistics     1
+
+set inject_start_time  80000000000ps
+set inject_stop_time  150000000000ps
+set injection_clock "pulp_cluster_tb/cluster_i/clk_i"
+set injection_clock_trigger 0
+set fault_period 100
+set rand_initial_injection_phase 1
+# max_num set to 0 means until stop_time
+set max_num_fault_inject 0
+set signal_fault_duration 20ns
+set register_fault_duration 0ns
+
+set allow_multi_bit_upset $::env(MULTI_BIT_UPSET)
+set use_bitwidth_as_weight 0
+set check_core_output_modification 0
+set check_core_next_state_modification 0
+set reg_to_sig_ratio 1
+
+source [file join $utils_base_path pulp_extract_nets.tcl]
+
+set inject_signals_netlist []
+set inject_register_netlist []
+set output_netlist []
+set next_state_netlist []
+set assertion_disable_list []
+
+# for {set idx 0} {$idx < 12} {incr idx} {
+#     set inject_signals_netlist [list {*}$inject_signals_netlist {*}[get_all_core_nets $idx]]
+#     set output_netlist [list {*}$output_netlist {*}[get_core_output_nets $idx]]
+# }
+
+set inject_register_netlist [list {*}$inject_register_netlist {*}[get_memory_slice {0 16} {0 50}]]
+
+source [file join $script_base_path inject_fault.tcl]
+
diff --git a/hwpe/neureka/src/main.c b/hwpe/neureka/src/main.c
new file mode 100644
index 0000000..d0133cf
--- /dev/null
+++ b/hwpe/neureka/src/main.c
@@ -0,0 +1,105 @@
+/*
+ * Copyright (C) 2020-2024 ETH Zurich and University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+/*
+ * Authors:  Francesco Conti <fconti@iis.ee.ethz.ch>
+ *           Gianna Paulin <pauling@iis.ee.ethz.ch>
+ *           Renzo Andri <andrire@iis.ee.ethz.ch>
+ *           Arpan Suravi Prasad <prasadar@iis.ee.ethz.ch>
+ *           Luka Macan <luka.macan@unibo.it>
+ * Main Test Program for N-EUREKA
+ */
+
+#include <pulp.h>
+#include <stdint.h>
+#include <stdio.h>
+
+#include "layer_util.h"
+#include "nnx_layer.h"
+#ifndef NO_ECC
+#include "ecc_check.h"
+#endif
+
+#define OUTPUT_SIZE 512
+
+extern uint8_t output[];
+
+#ifndef NO_ECC
+uint32_t ecc_errs[ECC_REGS];
+#endif
+
+static int check_output() {
+    uint32_t checksum = 0;
+    for (int i = 0; i < OUTPUT_SIZE; i++) {
+        checksum += output[i];
+    }
+    return (checksum != 0x00007330);
+}
+
+int errors = 0;
+#ifndef NO_ECC
+unsigned int intc_data_correctable_cnt = 0;
+unsigned int intc_meta_correctable_cnt = 0;
+unsigned int intc_data_uncorrectable_cnt = 0;
+unsigned int intc_meta_uncorrectable_cnt = 0;
+#endif
+
+int main() {
+
+  unsigned int core_id = get_core_id();
+  unsigned int cluster_id = rt_cluster_id();
+
+  if (core_id == 0) {
+
+    // execute NNX layer
+    execute_nnx_layer(NULL);
+  
+    errors = check_output();
+  
+    *(int *) 0x1A1040A0 = errors;
+    if(errors)
+      printf ("[KO] Terminated test with errors!!!\n");
+    else
+      printf ("[OK] Terminated test with no errors!!!\n");
+
+    #ifndef NO_ECC
+      // Check number of detected errors by ECC modules inside interconnect
+      intc_data_correctable_cnt = hwpe_hci_ecc_get_data_correctable_count(cluster_id);
+      intc_meta_correctable_cnt = hwpe_hci_ecc_get_meta_correctable_count(cluster_id);
+      intc_data_uncorrectable_cnt = hwpe_hci_ecc_get_data_uncorrectable_count(cluster_id);
+      intc_meta_uncorrectable_cnt = hwpe_hci_ecc_get_meta_uncorrectable_count(cluster_id);
+      for (int i = 0; i < 16; i++) {
+        intc_meta_correctable_cnt += tcdm_scrubber_get_mismatch_count(cluster_id, i);
+      }
+
+      printf("Data errors corrected inside Neureka: %d. Data errors uncorrectable inside Neureka: %d\n",
+        ecc_errs[0], ecc_errs[1]);
+      printf("Meta errors corrected inside Neureka: %d. Meta errors uncorrectable inside Neureka: %d\n",
+        ecc_errs[2], ecc_errs[3]);
+
+      printf("Data errors corrected inside intc: %d. Data errors uncorrectable inside intc: %d\n",
+        intc_data_correctable_cnt, intc_data_uncorrectable_cnt);
+      printf("Meta errors corrected inside intc: %d. Meta errors uncorrectable inside intc: %d\n",
+        intc_meta_correctable_cnt, intc_meta_uncorrectable_cnt);
+      #endif
+  }
+  synch_barrier();
+  #ifndef NO_ECC
+    return (errors != 0) && (intc_data_uncorrectable_cnt == 0 && intc_meta_uncorrectable_cnt == 0 && (ecc_errs[1]==0 && ecc_errs[3]==0));
+  #else
+    return errors;
+  #endif
+}
diff --git a/hwpe/neureka/src/nnx_layer.c b/hwpe/neureka/src/nnx_layer.c
new file mode 100644
index 0000000..06a7dd4
--- /dev/null
+++ b/hwpe/neureka/src/nnx_layer.c
@@ -0,0 +1,176 @@
+/*
+ * Luka Macan <luka.macan@unibo.it>
+ *
+ * Copyright 2023 ETH Zurich and University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ *
+ * SPDX-License-Identifier: Apache-2.0
+ */
+
+#include "nnx_layer.h"
+#include "ecc_check.h"
+#include "pulp.h"
+
+#include "neureka.h"
+#include "neureka_pulp_cluster_bsp.h"
+#include "neureka_task.h"
+#include "pulp_nnx_neureka.h"
+
+#define NULL 0
+
+typedef neureka_norm_mode_e nnx_norm_mode_e;
+typedef neureka_quant_t nnx_quant_t;
+typedef neureka_quant_function_e nnx_quant_function_e;
+typedef neureka_norm_t nnx_norm_t;
+typedef neureka_task_t nnx_task_t;
+typedef neureka_dev_t nnx_dev_t;
+typedef neureka_pulp_cluster_conf_t nnx_bsp_conf_t;
+typedef neureka_task_flag_e nnx_task_flag_e;
+
+#define nnxTaskFlagTrue neurekaTaskFlagTrue
+#define nnxTaskFlagFalse neurekaTaskFlagFalse
+
+#define nnx_task_init neureka_task_init
+#define nnx_task_set_op_to_conv neureka_task_set_op_to_conv
+#define nnx_task_set_bits neureka_task_set_bits
+#define nnx_task_set_norm_quant neureka_task_set_norm_quant
+#define nnx_task_set_weight_offset neureka_task_set_weight_offset
+#define nnx_task_set_weight_source neureka_task_set_weight_source
+#define nnx_task_set_activation_prefetch neureka_task_set_activation_prefetch
+#define nnx_task_set_dims neureka_task_set_dims
+#define nnx_task_set_ptrs_conv neureka_task_set_ptrs_conv
+#define nnx_task_set_ptrs_norm_quant neureka_task_set_ptrs_norm_quant
+
+#define nnx_bsp_get_dev neureka_pulp_cluster_get_dev
+
+#define nnx_init neureka_nnx_init
+#define nnx_dispatch_wait neureka_nnx_dispatch_wait
+#define nnx_dispatch neureka_nnx_dispatch
+#define nnx_resolve_wait neureka_nnx_resolve_wait
+#define nnx_read_ecc_regs neureka_nnx_read_ecc_regs
+#define nnx_term neureka_nnx_term
+
+// Generated headers
+#include "layer_conf.h"
+#if HAS_BIAS != 0
+  #include "bias.h"
+#endif
+#include "input.h"
+#include "output.h"
+#if HAS_NORM_QUANT != 0
+  #include "scale.h"
+#endif
+#include "weight.h"
+
+static void task_prepare(nnx_task_t *task) {
+  nnx_task_init(task);
+  nnx_task_set_op_to_conv(task, WEIGHT_HEIGHT, GROUPS > 1);
+  nnx_task_set_bits(task, INPUT_BITS, OUTPUT_BITS, WEIGHT_BITS);
+
+  nnx_task_set_weight_offset(task, weightOffsetModeLayerWise, WEIGHT_OFFSET);
+
+#ifdef NEUREKA_WEIGHT_SOURCE_WMEM
+  nnx_task_set_weight_source(task, neurekaWeightSourceWmem);
+  nnx_task_set_activation_prefetch(task, activationPrefetchOn);
+#else
+  neureka_task_set_weight_source(task, neurekaWeightSourceTcdm);
+  nnx_task_set_activation_prefetch(task, activationPrefetchOff);
+#endif
+#if INPUT_SIGNED == 1
+  neureka_task_set_input_signed(task);
+#else
+  neureka_task_set_input_unsigned(task);
+#endif
+
+  const uint32_t w_in_stride = INPUT_CHANNEL * INPUT_BITS / 8;
+  const uint32_t h_in_stride = INPUT_WIDTH * w_in_stride;
+  const uint32_t w_out_stride = OUTPUT_CHANNEL * OUTPUT_BITS / 8;
+  const uint32_t h_out_stride = OUTPUT_WIDTH * w_out_stride;
+
+#if STRIDE_HEIGHT == 2 && STRIDE_WIDTH == 2
+  nnx_task_set_dims_stride2x2(
+      task, INPUT_HEIGHT, INPUT_WIDTH, INPUT_CHANNEL, h_in_stride, w_in_stride,
+      OUTPUT_HEIGHT, OUTPUT_WIDTH, OUTPUT_CHANNEL, h_out_stride, w_out_stride,
+      WEIGHT_HEIGHT, WEIGHT_WIDTH, PADDING_TOP, PADDING_BOTTOM, PADDING_LEFT,
+      PADDING_RIGHT);
+#else
+  nnx_task_set_dims(task, INPUT_WIDTH, INPUT_CHANNEL, h_in_stride, w_in_stride,
+                    OUTPUT_HEIGHT, OUTPUT_WIDTH, OUTPUT_CHANNEL, h_out_stride,
+                    w_out_stride, PADDING_TOP, PADDING_BOTTOM, PADDING_LEFT,
+                    PADDING_RIGHT);
+#endif
+
+  nnx_task_set_ptrs_conv(task, (uint32_t)input, INPUT_WIDTH, w_in_stride,
+                         PADDING_TOP, PADDING_LEFT, (uint32_t)output,
+                         (uint32_t)weight);
+#if HAS_NORM_QUANT == 1
+#if SCALE_BITS == 8
+  const nnx_norm_mode_e normMode = normMode8Bit;
+#elif SCALE_BITS == 32
+  const nnx_norm_mode_e normMode = normMode32Bit;
+#endif
+
+  const nnx_task_flag_e flag_bias =
+      HAS_BIAS ? nnxTaskFlagTrue : nnxTaskFlagFalse;
+  const uint32_t bias_ptr = (uint32_t)(HAS_BIAS ? bias : NULL);
+
+  nnx_quant_function_e quant_function =
+      HAS_RELU ? quantFunctionRelu : quantFunctionIdentity;
+
+  nnx_task_set_norm_quant(task,
+                          (nnx_quant_t){.shift_amount = OUTSHIFT,
+                                        .function = quant_function,
+                                        .flag_rounding = nnxTaskFlagFalse},
+                          (nnx_norm_t){.mode = normMode,
+                                       .flag_bias = flag_bias,
+                                       .flag_shift = nnxTaskFlagFalse});
+
+  nnx_task_set_ptrs_norm_quant(task, (uint32_t)scale, NULL, bias_ptr);
+#endif // HAS_NORM_QUANT
+}
+
+static void task_execute(nnx_task_t *task) {
+  nnx_dev_t *dev = nnx_bsp_get_dev();
+
+  nnx_bsp_conf_t conf = {.max_stall = 8};
+  nnx_init(dev, &conf);
+
+  nnx_dispatch_wait(dev);
+
+  // printf("CFG:\n");
+  // for (int i=0; i<sizeof(neureka_task_data_t)/4; i++) {
+  //   printf("%08x\n", ((uint32_t *) &task->data)[i]);
+  // }
+#if STRIDE_HEIGHT == 2 && STRIDE_WIDTH == 2
+  nnx_dispatch_stride2x2(dev, task, INPUT_WIDTH, INPUT_CHANNEL, OUTPUT_HEIGHT,
+                         OUTPUT_WIDTH, OUTPUT_CHANNEL, WEIGHT_HEIGHT,
+                         WEIGHT_WIDTH);
+#else
+  nnx_dispatch(dev, task);
+#endif
+
+  nnx_resolve_wait(dev, task);
+#ifndef NO_ECC
+  nnx_read_ecc_regs(dev, (uint32_t)ecc_errs);
+#endif
+
+  nnx_term(dev);
+
+}
+
+void execute_nnx_layer(void *args) {
+  nnx_task_t task;
+  task_prepare(&task);
+  task_execute(&task);
+}
diff --git a/hwpe/redmule/Makefile b/hwpe/redmule/Makefile
new file mode 100644
index 0000000..2d118a6
--- /dev/null
+++ b/hwpe/redmule/Makefile
@@ -0,0 +1,23 @@
+PULP_APP = test
+PULP_APP_SRCS = redmule.c
+PULP_CFLAGS = -O3
+
+ifeq ($(use_dma),1)
+ PULP_CFLAGS += -DUSE_DMA
+endif
+ifeq ($(no_ecc),1)
+ PULP_CFLAGS += -DNO_ECC
+endif
+
+ifeq ($(fault_inject),1)
+  export FAULT_INJECTION=1
+  export FAULT_INJECTION_SCRIPT=$(CURDIR)/pulp_inject_fault.tcl
+endif
+
+ifeq ($(multi_bit_upset),1)
+  export MULTI_BIT_UPSET=1
+else
+  export MULTI_BIT_UPSET=0
+endif
+
+include $(PULP_SDK_HOME)/install/rules/pulp_rt.mk
diff --git a/hwpe/redmule/archi_redmule.h b/hwpe/redmule/archi_redmule.h
new file mode 100644
index 0000000..40eceee
--- /dev/null
+++ b/hwpe/redmule/archi_redmule.h
@@ -0,0 +1,170 @@
+/*
+ * Copyright (C) 2022-2023 ETH Zurich and University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ * SPDX-License-Identifier: Apache-2.0
+ * 
+ * Author: Yvan Tortorella  <yvan.tortorella@unibo.it>
+ *
+ * High-level architecture of RedMulE
+ *
+ */
+
+#ifndef __ARCHI_REDMULE_H__
+#define __ARCHI_REDMULE_H__
+
+/*
+ * |========================================================================|
+ * ||                                                                      ||
+ * ||Control and generic configuration register layout                     ||
+ * |========================================================================|
+ * || # reg |  offset  |  bits   |   bitmask    ||  content                ||
+ * ||-------+----------+---------+--------------++-------------------------||
+ * ||    0  |  0x0000  |  31: 0  |  0xFFFFFFFF  ||  TRIGGER                ||
+ * ||    1  |  0x0004  |  31: 0  |  0xFFFFFFFF  ||  ACQUIRE                ||
+ * ||    2  |  0x0008  |  31: 0  |  0xFFFFFFFF  ||  EVT_ENABLE             ||
+ * ||    3  |  0x000c  |  31: 0  |  0xFFFFFFFF  ||  STATUS                 ||
+ * ||    4  |  0x0010  |  31: 0  |  0xFFFFFFFF  ||  RUNNING_JOB            ||
+ * ||    5  |  0x0014  |  31: 0  |  0xFFFFFFFF  ||  SOFT_CLEAR             ||
+ * |========================================================================|
+ * ||                                                                      ||
+ * ||Job-dependent registers layout                                        ||
+ * |========================================================================|
+ * || # reg |  offset  |  bits   |   bitmask    ||  content                ||
+ * ||-------+----------+---------+--------------++-------------------------||
+ * ||    0  |  0x0040  |  31: 0  |  0xFFFFFFFF  ||  X_ADDR                 ||
+ * ||-------+----------+---------+--------------++-------------------------||
+ * ||    1  |  0x0044  |  31: 0  |  0xFFFFFFFF  ||  W_ADDR                 ||
+ * ||-------+----------+---------+--------------++-------------------------||
+ * ||    2  |  0x0048  |  31: 0  |  0xFFFFFFFF  ||  Z_ADDR                 ||
+ * ||-------+----------+---------+--------------++-------------------------||
+ * ||    3  |  0x004C  |         |              ||  Matrix Config 0 Reg    ||
+ * ||       |          |  31:16  |  0xFFFF0000  ||  K Size (W Columns)     ||
+ * ||       |          |  15: 0  |  0x0000FFFF  ||  M Size (X Rows)        ||
+ * ||-------+----------+---------+--------------++-------------------------||
+ * ||    4  |  0x0050  |         |              ||  Matrix Config 1 Reg    ||
+ * ||       |          |  31:16  |  0xFFFFFFFF  ||  N Size (X Cols/W Rows) ||
+ * ||-------+----------+---------+--------------++-------------------------||
+ * ||    5  |  0x0054  |         |              ||  Matrix Arithmetic Reg  ||
+ * ||       |          |  12:10  |  0x00001C00  ||  Operation selection    ||
+ * ||       |          |   9: 7  |  0x00000380  ||  Input/Output format    ||
+ * |========================================================================|
+ *
+ */
+
+/* PULP Cluster Archi defines */
+#define ARCHI_CLUST_CTRL_BASE ARCHI_CLUSTER_CTRL_ADDR
+#define ARCHI_CLUST_HWPE_BASE ARCHI_HWCE_ADDR
+#define DMA_COMMAND_QUEUE     ARCHI_MCHAN_DEMUX_ADDR
+#define DMA_STATUS_REGISTER   (ARCHI_MCHAN_DEMUX_ADDR + 4)
+#define ARCHI_CL_HWPE_EVT0 12
+#define ARCHI_CL_HWPE_EVT1 13
+#define FC_DMA_EVENT 8
+#define CL_DMA_EVENT 22
+#define CLUST_CTRL_HWPE_EN 0x18
+#define CLUST_CTRL_HWPE_EN_MASK 0x800
+#define __builtin_bitinsert(a,b,c,d) (a | (((b << (32-c)) >> (32-c)) << d))
+
+// RedMulE architecture
+#define ADDR_WIDTH   32
+#define DATA_WIDTH   256
+#define REDMULE_FMT  16
+#define ARRAY_HEIGHT 4
+#define PIPE_REGS    3
+#define ARRAY_WIDTH  12 /* Superior limit is ARRAY_HEIGHT*PIPE_REGS */
+
+// Commands
+#define REDMULE_TRIGGER     0x00
+#define REDMULE_ACQUIRE     0x04
+#define REDMULE_FINISHED    0x08
+#define REDMULE_STATUS      0x0C
+#define REDMULE_RUNNING_JOB 0x10
+#define REDMULE_SOFT_CLEAR  0x14
+
+// Registers
+#define REDMULE_REG_OFFS  0x40
+// #define REDMULE_REG_X_PTR 0x00
+// #define REDMULE_REG_W_PTR 0x04
+// #define REDMULE_REG_Z_PTR 0x08
+// #define REDMULE_MCFG0_PTR 0x0C
+// #define REDMULE_MCFG1_PTR 0x10
+// #define REDMULE_ARITH_PTR 0x14
+#define REDMULE_REG_X_PTR              0x00
+#define REDMULE_REG_W_PTR              0x04
+#define REDMULE_REG_Y_PTR              0x08
+#define REDMULE_REG_Z_PTR              0x0C
+#define REDMULE_REG_X_ITER_PTR         0x10
+#define REDMULE_REG_W_ITER_PTR         0x14
+#define REDMULE_REG_LEFTOVERS_PTR      0x18
+#define REDMULE_REG_LEFT_PARAMS_PTR    0x1C
+#define REDMULE_REG_X_D1_STRIDE_PTR    0x20
+#define REDMULE_REG_W_TOT_LEN_PTR      0x24
+#define REDMULE_REG_TOT_X_READ_PTR     0x28
+#define REDMULE_REG_W_D0_STRIDE_PTR    0x2C
+#define REDMULE_REG_YZ_TOT_LEN_PTR     0x30
+#define REDMULE_REG_YZ_D0_STRIDE_PTR   0x34
+#define REDMULE_REG_YZ_D2_STRIDE_PTR   0x38
+#define REDMULE_REG_X_ROWS_OFFS_PTR    0x3C
+#define REDMULE_REG_X_BUFFER_SLOTS_PTR 0x40
+#define REDMULE_REG_X_TOT_LEN_PTR      0x44
+#define REDMULE_REG_OP_SELECTION       0x48
+
+#define REDMULE_ECC_REG_OFFS           0x90
+#define DATA_CORR_ERR                  0x00
+#define DATA_UNCORR_ERR                0x04
+#define METADATA_CORR_ERR              0x08
+#define METADATA_UNCORR_ERR            0x0c
+
+// OPs definition
+#define MATMUL 0x0
+#define GEMM   0x1
+#define ADDMAX 0x2
+#define ADDMIN 0x3
+#define MULMAX 0x4
+#define MULMIN 0x5
+#define MAXMIN 0x6
+#define MINMAX 0x7
+
+// GEMM formats
+#define Float8     0x0
+#define Float16    0x1
+#define Float8Alt  0x2
+#define Float16Alt 0x3
+
+#define RNE       0x0
+#define RTZ       0x1
+#define OP_FMADD  0x0
+#define OP_ADD    0x2
+#define OP_MUL    0x3
+#define OP_MINMAX 0x7
+
+// FP Formats encoding
+#define FP16    0x2
+#define FP8     0x3
+#define FP16ALT 0x4
+#define FP8ALT  0x5
+
+/* DMA Archi */
+#define DMA_TX  0
+#define DMA_RX  1
+#define DMA_INC 1
+
+#define PLP_DMA_TYPE_BIT    0x00000011
+#define PLP_DMA_INCR_BIT    0x00000012
+#define PLP_DMA_2D_BIT      0x00000013
+#define PLP_DMA_ELE_BIT     0x00000014
+#define PLP_DMA_ILE_BIT     0x00000015
+#define PLP_DMA_BLE_BIT     0x00000016
+#define PLP_DMA_2D_TCDM_BIT 0x0000017
+
+#endif
diff --git a/hwpe/redmule/hal_redmule.h b/hwpe/redmule/hal_redmule.h
new file mode 100644
index 0000000..8fc5000
--- /dev/null
+++ b/hwpe/redmule/hal_redmule.h
@@ -0,0 +1,556 @@
+/*
+ * Copyright (C) 2022-2023 ETH Zurich and University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ * SPDX-License-Identifier: Apache-2.0
+ * 
+ * Author: Yvan Tortorella  <yvan.tortorella@unibo.it>
+ *
+ * RedMulE Hardware Abstraction Layer (HAL)
+ */
+
+#ifndef __HAL_REDMULE_H__
+#define __HAL_REDMULE_H__
+
+#include <pulp.h>
+#include "inc/x_input.h"
+#include "inc/w_input.h"
+#include "inc/y_input.h"
+#include "inc/z_output.h"
+#include "inc/golden.h"
+#include "inc/tensor_dim.h"
+
+/*
+ * 
+ * For control, generic configuration register layout,
+ * and job-dependent register map, look at redmule_archi.h
+ *
+ */
+
+// For all the following functions we use __builtin_pulp_OffsetedWrite and __builtin_pulp_OffsetedRead
+// instead of classic load/store because otherwise the compiler is not able to correctly factorize
+// the HWPE base in case several accesses are done, ending up with twice more code
+
+#define HWPE_WRITE(value, offset) *(int *)(ARCHI_CLUST_HWPE_BASE + offset) = value
+#define HWPE_READ(offset) *(int *)(ARCHI_CLUST_HWPE_BASE + offset)
+
+static inline void redmule_x_add_set (unsigned int value) {
+  HWPE_WRITE(value, REDMULE_REG_OFFS + REDMULE_REG_X_PTR);
+}
+
+static inline void redmule_w_add_set (unsigned int value) {
+  HWPE_WRITE(value, REDMULE_REG_OFFS + REDMULE_REG_W_PTR);
+}
+
+static inline void redmule_y_add_set (unsigned int value) {
+  HWPE_WRITE(value, REDMULE_REG_OFFS + REDMULE_REG_Y_PTR);
+}
+
+static inline void redmule_z_add_set (unsigned int value) {
+  HWPE_WRITE(value, REDMULE_REG_OFFS + REDMULE_REG_Z_PTR);
+}
+
+// static inline void redmule_mcfg_set (uint32_t mcfg0, uint32_t mcfg1) {
+//   HWPE_WRITE(mcfg0, REDMULE_REG_OFFS + REDMULE_MCFG0_PTR);
+//   HWPE_WRITE(mcfg1, REDMULE_REG_OFFS + REDMULE_MCFG1_PTR);
+// }
+// 
+// static inline void redmule_arith_set (uint32_t arith) {
+//   HWPE_WRITE(arith, REDMULE_REG_OFFS + REDMULE_ARITH_PTR);
+// }
+
+static inline void hwpe_trigger_job() {
+  HWPE_WRITE(0, REDMULE_TRIGGER);
+}
+
+static inline int hwpe_acquire_job() {
+  return HWPE_READ(REDMULE_ACQUIRE);
+}
+
+static inline unsigned int hwpe_get_status() {
+  return HWPE_READ(REDMULE_STATUS);
+}
+
+static inline unsigned int hwpe_get_running_job() {
+  return HWPE_READ(REDMULE_RUNNING_JOB);
+}
+
+static inline void hwpe_soft_clear() {
+  HWPE_WRITE(0, REDMULE_SOFT_CLEAR);
+}
+
+static inline void hwpe_cg_enable() {
+  *(volatile int*) (ARCHI_CLUST_CTRL_BASE + CLUST_CTRL_HWPE_EN) |= CLUST_CTRL_HWPE_EN_MASK;
+}
+
+static inline void hwpe_cg_disable() {
+  *(volatile int*) (ARCHI_CLUST_CTRL_BASE + CLUST_CTRL_HWPE_EN) &= ~CLUST_CTRL_HWPE_EN_MASK;
+}
+
+static inline void redmule_evt_wait() {
+  do {
+    eu_evt_maskWaitAndClr (1 << ARCHI_CL_HWPE_EVT0);
+  } while((*(int volatile *)(ARCHI_CLUST_HWPE_BASE + REDMULE_STATUS)) != 0);
+}
+
+static inline int hwpe_wait_acquire() {
+  int job_id = hwpe_acquire_job();
+  while(job_id < 0) {
+    eu_evt_maskWaitAndClr (1 << ARCHI_CL_HWPE_EVT0);
+    job_id = hwpe_acquire_job();
+  }
+  return job_id;
+}
+
+static inline unsigned int redmule_get_data_correctable_count () {
+  return HWPE_READ(REDMULE_ECC_REG_OFFS + DATA_CORR_ERR);
+}
+
+static inline unsigned int redmule_get_data_uncorrectable_count () {
+  return HWPE_READ(REDMULE_ECC_REG_OFFS + DATA_UNCORR_ERR);
+}
+
+static inline unsigned int redmule_get_meta_correctable_count () {
+  return HWPE_READ(REDMULE_ECC_REG_OFFS + METADATA_CORR_ERR);
+}
+
+static inline unsigned int redmule_get_meta_uncorrectable_count () {
+  return HWPE_READ(REDMULE_ECC_REG_OFFS + METADATA_UNCORR_ERR);
+}
+
+/* DMA APIs */
+static inline int mchan_alloc(){
+  return *(volatile int*) DMA_COMMAND_QUEUE;
+}
+
+static inline void mchan_transfer(unsigned int len,
+                                  unsigned int ext_addr,
+                                  unsigned int tcdm_addr) {
+
+  *(volatile int*) DMA_COMMAND_QUEUE = len |
+                                       (DMA_RX << PLP_DMA_TYPE_BIT)  |
+                                       (DMA_INC << PLP_DMA_INCR_BIT) |
+                                       (0 << PLP_DMA_2D_BIT)         |
+                                       (1 << PLP_DMA_ELE_BIT)        |
+                                       (1 << PLP_DMA_ILE_BIT)        |
+                                       (0 << PLP_DMA_BLE_BIT)        |
+                                       (0 << PLP_DMA_2D_TCDM_BIT);
+  *(volatile int*) DMA_COMMAND_QUEUE = tcdm_addr;
+  *(volatile int*) DMA_COMMAND_QUEUE = ext_addr;
+}
+
+static inline void mchan_barrier(int id) {
+  while(((*(volatile int*)(DMA_STATUS_REGISTER)) >> id ) & 0x1 ) {
+    eu_evt_maskWaitAndClr(1 << FC_DMA_EVENT);
+ }
+}
+
+static inline void mchan_free(int id) {
+  *(volatile int*) DMA_STATUS_REGISTER = 0x1 << id;
+}
+
+// void redmule_cfg (unsigned int x,  unsigned int w,  unsigned int z,
+//                   uint16_t m_size, uint16_t n_size, uint16_t k_size,
+//                   uint8_t gemm_op, uint8_t gemm_fmt){
+// 
+//   uint32_t mcfg_reg0 = 0;
+//   uint32_t mcfg_reg1 = 0;
+//   uint32_t arith_reg = 0;
+// 
+//   mcfg_reg0 = (k_size << 16) |
+//               (m_size <<  0);
+//   mcfg_reg1 = n_size << 0;
+// 
+//   arith_reg = (gemm_op << 10) |
+//               (gemm_fmt << 7);
+// 
+//   redmule_x_add_set ((unsigned int) x);
+//   redmule_w_add_set ((unsigned int) w);
+//   redmule_z_add_set ((unsigned int) z);
+//   redmule_mcfg_set  ((unsigned int) mcfg_reg0,
+//                      (unsigned int) mcfg_reg1);
+//   redmule_arith_set ((unsigned int) arith_reg);
+//   
+// }
+
+void redmule_cfg (uint16_t m_size, uint16_t n_size, uint16_t k_size, uint8_t gemm_ops){
+   uint32_t x_iters        = 0;
+   uint32_t w_iters        = 0;
+   uint32_t leftovers      = 0;
+   uint32_t left_params    = 0;
+   uint32_t x_d1_stride    = 0;
+   uint32_t x_rows_offs    = 0;
+   uint32_t w_tot_len      = 0;
+   uint32_t w_d1_len       = 0;
+   uint32_t w_d0_stride    = 0;
+   uint32_t yz_tot_len     = 0;
+   uint32_t yz_d0_stride   = 0;
+   uint32_t yz_d2_stride   = 0;
+   uint32_t tot_x_read     = 0;
+   uint32_t x_buffer_slots = 0;
+   uint32_t op_selection   = 0;
+   uint16_t tot_stores     = 0;
+   uint16_t w_rows         = n_size;
+   uint16_t depth          = DATA_WIDTH/(ARRAY_HEIGHT*FPFORMAT);
+   uint8_t  tile           = ARRAY_HEIGHT*(PIPE_REGS + 1);
+   _Bool    x_rows_sub     = 0;
+   _Bool    x_cols_sub     = 0;
+   _Bool    w_cols_sub     = 0;
+   uint16_t x_rows_iter,
+            x_rows_iter_tmp,
+            w_rows_iter,
+            w_rows_iter_tmp;
+   uint16_t x_cols_iter,
+            x_cols_iter_tmp,
+            w_cols_iter,
+            w_cols_iter_tmp;
+   uint8_t  x_rows_lftovr,
+            x_cols_lftovr,
+            w_rows_lftovr,
+            w_cols_lftovr,
+            slots;
+
+   // Calculating the number of iterations alng the two dimensions of the X matrix
+   x_rows_iter_tmp = m_size/ARRAY_WIDTH;
+   x_cols_iter_tmp = n_size/tile;
+
+   // Calculating the number of iterations alng the two dimensions of the W matrix
+   w_rows_iter_tmp = w_rows;
+   w_cols_iter_tmp = k_size/tile;
+
+   // Calculating the residuals along the input dimensions
+   x_rows_lftovr = m_size - (x_rows_iter_tmp*ARRAY_WIDTH);
+   x_cols_lftovr = n_size - (x_cols_iter_tmp*tile);
+
+   // Calculating the residuals along the weight dimensions
+   w_rows_lftovr = n_size - (ARRAY_HEIGHT*(w_rows/ARRAY_HEIGHT));
+   w_cols_lftovr = k_size - (w_cols_iter_tmp*tile);
+
+   if (w_cols_lftovr != 0)
+     w_cols_iter = w_cols_iter_tmp + 1;
+   else 
+     w_cols_iter = w_cols_iter_tmp;
+
+  if (w_rows_lftovr != 0)
+    w_rows_iter = w_rows_iter_tmp + ARRAY_HEIGHT - w_rows_lftovr;
+  else
+    w_rows_iter = w_rows_iter_tmp;
+
+   if (x_cols_lftovr != 0)
+     x_cols_iter = x_cols_iter_tmp + 1;
+   else 
+     x_cols_iter = x_cols_iter_tmp;
+
+   if (x_rows_lftovr != 0)
+     x_rows_iter = x_rows_iter_tmp + 1;
+   else 
+     x_rows_iter = x_rows_iter_tmp;
+
+   if (x_cols_lftovr%depth != 0)
+     x_buffer_slots = x_cols_lftovr/depth + 1;
+   else
+     x_buffer_slots = x_cols_lftovr/depth;
+
+   // Calculating the number of total stores
+   tot_stores = x_rows_iter*w_cols_iter;
+
+   // Determining if input matrixes are sub-matrixes
+   if (m_size < ARRAY_WIDTH)
+     x_rows_sub = 1;
+   if (n_size < ARRAY_HEIGHT)
+     x_cols_sub = 1;
+   if (k_size < tile)
+    w_cols_sub = 1;
+
+   // Operation selection
+   switch (gemm_ops) {
+     case MATMUL:
+       op_selection |= (RNE << 29 | RNE << 26 | OP_FMADD << 22 | OP_MINMAX << 18 | SRC_FMT << 15 | DST_FMT << 12) | 0;
+     break;
+     
+     case GEMM:
+       op_selection |= (RNE << 29 | RNE << 26 | OP_FMADD << 22 | OP_MINMAX << 18 | SRC_FMT << 15 | DST_FMT << 12) | 1;
+     break;
+     
+     case ADDMAX:
+       op_selection |= (RNE << 29 | RTZ << 26 | OP_ADD << 22 | OP_MINMAX << 18 | SRC_FMT << 15 | DST_FMT << 12) | 1;
+     break;
+     
+     case ADDMIN:
+       op_selection |= (RNE << 29 | RNE << 26 | OP_ADD << 22 | OP_MINMAX << 18 | SRC_FMT << 15 | DST_FMT << 12) | 1;
+     break;
+     
+     case MULMAX:
+       op_selection |= (RNE << 29 | RTZ << 26 | OP_MUL << 22 | OP_MINMAX << 18 | SRC_FMT << 15 | DST_FMT << 12) | 1;
+     break;
+     
+     case MULMIN:
+       op_selection |= (RNE << 29 | RNE << 26 | OP_MUL << 22 | OP_MINMAX << 18 | SRC_FMT << 15 | DST_FMT << 12) | 1;
+     break;
+     
+     case MAXMIN:
+       op_selection |= (RTZ << 29 | RNE << 26 | OP_MINMAX << 22 | OP_MINMAX << 18 | SRC_FMT << 15 | DST_FMT << 12) | 1;
+     break;
+     
+     case MINMAX:
+       op_selection |= (RNE << 29 | RTZ << 26 | OP_MINMAX << 22 | OP_MINMAX << 18 | SRC_FMT << 15 | DST_FMT << 12) | 1;
+     break;
+   }
+
+   // Storing iterations and residuals in registers
+   x_iters      |= x_rows_iter   << 16 | x_cols_iter   << 0;
+   w_iters      |= w_rows_iter   << 16 | w_cols_iter   << 0;
+   leftovers    |= x_rows_lftovr << 24 | x_cols_lftovr << 16 | w_rows_lftovr << 8  | w_cols_lftovr << 0;
+   left_params  |= tot_stores    << 16 | x_rows_sub    << 15 | x_cols_sub    << 14 | w_cols_sub   << 13;
+   x_d1_stride   = ((4*FPFORMAT)/ADDR_WIDTH)*(((DATA_WIDTH/FPFORMAT)*x_cols_iter_tmp) + x_cols_lftovr);
+   x_rows_offs   = ARRAY_WIDTH*x_d1_stride;
+   w_tot_len     = w_rows_iter*w_cols_iter*x_rows_iter;
+   w_d0_stride   = ((4*FPFORMAT)/ADDR_WIDTH)*(((DATA_WIDTH/FPFORMAT)*w_cols_iter_tmp) + w_cols_lftovr);
+   yz_tot_len    = ARRAY_WIDTH*x_rows_iter*w_cols_iter;
+   yz_d0_stride  = w_d0_stride;
+   yz_d2_stride  = ARRAY_WIDTH*w_d0_stride;
+   tot_x_read    = x_rows_iter*x_cols_iter*w_cols_iter;
+
+   // Writing the computations in configuration register
+   HWPE_WRITE(x_iters       , REDMULE_REG_OFFS + REDMULE_REG_X_ITER_PTR         );
+   HWPE_WRITE(w_iters       , REDMULE_REG_OFFS + REDMULE_REG_W_ITER_PTR         );
+   HWPE_WRITE(leftovers     , REDMULE_REG_OFFS + REDMULE_REG_LEFTOVERS_PTR      );
+   HWPE_WRITE(left_params   , REDMULE_REG_OFFS + REDMULE_REG_LEFT_PARAMS_PTR    );
+   HWPE_WRITE(x_d1_stride   , REDMULE_REG_OFFS + REDMULE_REG_X_D1_STRIDE_PTR    );
+   HWPE_WRITE(x_rows_offs   , REDMULE_REG_OFFS + REDMULE_REG_X_ROWS_OFFS_PTR    );
+   HWPE_WRITE(tot_x_read    , REDMULE_REG_OFFS + REDMULE_REG_TOT_X_READ_PTR     );
+   HWPE_WRITE(x_buffer_slots, REDMULE_REG_OFFS + REDMULE_REG_X_BUFFER_SLOTS_PTR );
+   HWPE_WRITE(w_tot_len     , REDMULE_REG_OFFS + REDMULE_REG_W_TOT_LEN_PTR      );
+   HWPE_WRITE(w_d0_stride   , REDMULE_REG_OFFS + REDMULE_REG_W_D0_STRIDE_PTR    );
+   HWPE_WRITE(yz_tot_len    , REDMULE_REG_OFFS + REDMULE_REG_YZ_TOT_LEN_PTR     );
+   HWPE_WRITE(yz_d0_stride  , REDMULE_REG_OFFS + REDMULE_REG_YZ_D0_STRIDE_PTR   );
+   HWPE_WRITE(yz_d2_stride  , REDMULE_REG_OFFS + REDMULE_REG_YZ_D2_STRIDE_PTR   );
+   HWPE_WRITE(op_selection  , REDMULE_REG_OFFS + REDMULE_REG_OP_SELECTION       );
+}
+
+void generate_test_data16(int x_start_addr,
+                          int w_start_addr,
+                          int y_start_addr,
+                          int m_size,
+                          int n_size,
+                          int k_size) {
+  
+  int x_addr     = x_start_addr;
+  int w_addr     = w_start_addr;
+  int y_addr     = y_start_addr;
+  int x_end_addr = x_start_addr + (2*m_size*n_size);
+  int w_end_addr = w_start_addr + (2*n_size*k_size);
+  int y_end_addr = y_start_addr + (2*m_size*k_size);
+
+  // Generating input stimuli from golden model
+  for (x_addr = x_start_addr; x_addr < x_end_addr; x_addr += 2) {
+    int x = x_addr - x_start_addr;
+    *(uint32_t *)(x_addr) = x_inp[x/2];
+  }
+
+  // Generating Weight stimuli from golden model
+  for (w_addr = w_start_addr; w_addr < w_end_addr; w_addr += 2) {
+    int w = w_addr - w_start_addr;
+    *(uint32_t *)(w_addr) = w_inp[w/2];
+  }
+
+  for (y_addr = y_start_addr; y_addr < y_end_addr; y_addr += 2) {
+    int y = y_addr - y_start_addr;
+    *(uint32_t *)(y_addr) = y_inp[y/2];
+  }
+}
+
+int redmule_compare16 (int z_start_addr, int m_size, int k_size) {
+  int err = 0;
+  int z_end_addr = z_start_addr + 2*m_size*k_size;
+  uint16_t z_computed;
+  uint16_t diff, diff_1, diff_2;
+
+  for (int z_addr = z_start_addr; z_addr < z_end_addr; z_addr += 2) {
+    int z = z_addr - z_start_addr;
+    z_computed = *(uint32_t *)(z_addr);
+
+    if ( z_computed != z_oup[z/2] ) {
+      diff_1 = z_computed - z_oup[z/2];
+      if (diff_1 > 3) {
+        diff_2 = z_oup[z/2] - z_computed;
+        if (diff_2 > 3) {
+          err++;
+        }
+      }
+    }
+  }
+
+  return err;
+
+}
+
+int redmule16_compare_int(uint32_t *actual_z, uint32_t *golden_z, int len) {
+  #define ERR 0x0011
+  uint32_t actual_word = 0;
+  uint16_t actual_MSHWord, actual_LSHWord;
+  uint32_t golden_word = 0;
+  uint16_t golden_MSHWord, golden_LSHWord;
+  uint32_t actual = 0;
+  uint32_t golden = 0;
+
+  int errors = 0;
+  int error;
+
+  for (int i=0; i<len; i++) {
+    error = 0;
+    actual_word = *(actual_z+i);
+    golden_word = *(golden_z+i);
+
+    uint16_t diff = 0;
+    
+    // Chechink Least Significant Half-Word
+    actual_LSHWord = (uint16_t)(actual_word & 0x0000FFFF);
+    golden_LSHWord = (uint16_t)(golden_word & 0x0000FFFF);
+
+    diff = (actual_LSHWord > golden_LSHWord) ? (actual_LSHWord - golden_LSHWord) : 0;
+    diff = (actual_LSHWord < golden_LSHWord) ? (golden_LSHWord - actual_LSHWord) : 0;
+
+    if (diff > ERR) {
+      error = 1;
+      #ifdef VERBOSE
+        tfp_printf ("diff: 0x%08x\n", diff);
+        tfp_printf ("LSW: Error!\n");
+      #endif
+    }
+
+    // Checking Most Significant Half-Word
+    actual_MSHWord = (uint16_t)((actual_word >> 16) & 0x0000FFFF);
+    golden_MSHWord = (uint16_t)((golden_word >> 16) & 0x0000FFFF);
+
+    diff = (actual_MSHWord > golden_MSHWord) ? (actual_MSHWord - golden_MSHWord) : 0;
+    diff = (actual_MSHWord < golden_MSHWord) ? (golden_MSHWord - actual_MSHWord) : 0;
+
+    if (diff > ERR) {
+      error = 1;
+      #ifdef VERBOSE
+        tfp_printf ("diff: 0x%08x\n", diff);
+        tfp_printf ("MSW: Error!\n");
+      #endif
+    }
+    
+    errors += error;
+
+    #ifdef DEBUG
+      tfp_printf("  Golden: 0x%08x; Actual: 0x%08x,\n", golden_word, actual_word);
+    #endif
+
+    #ifdef VERBOSE
+    if(error) {
+        if(errors==1) tfp_printf("  golden     <- actual     @ address    @ index\n");
+        tfp_printf("  0x%08x <- 0x%08x @ 0x%08x @ 0x%08x\n", golden_word, actual_word, (actual_z+i), i*4);
+    }
+    #endif
+  }
+  return errors;
+}
+
+int redmule8_compare_int(uint32_t *actual_z, uint32_t *golden_z, int len) {
+  #define ERR 0x0011
+  uint32_t actual_word = 0;
+  uint8_t  actual_Byte0,
+           actual_Byte1,
+           actual_Byte2,
+           actual_Byte3;
+  uint32_t golden_word = 0;
+  uint8_t  golden_Byte0,
+           golden_Byte1,
+           golden_Byte2,
+           golden_Byte3;
+  uint32_t actual = 0;
+  uint32_t golden = 0;
+
+  int errors = 0;
+  int error;
+
+  for (int i=0; i<len; i++) {
+    error = 0;
+    actual_word = *(actual_z+i);
+    golden_word = *(golden_z+i);
+
+    uint8_t diff = 0;
+    
+    // Cheching Byte0
+    actual_Byte0 = (uint8_t)(actual_word & 0x000000FF);
+    golden_Byte0 = (uint8_t)(golden_word & 0x000000FF);
+
+    diff = (actual_Byte0 > golden_Byte0) ? (actual_Byte0 - golden_Byte0) : 0;
+    diff = (actual_Byte0 < golden_Byte0) ? (golden_Byte0 - actual_Byte0) : 0;
+
+    if (diff > ERR) {
+      error = 1;
+      tfp_printf ("diff: 0x%08x\n", diff);
+      tfp_printf ("Byte0: Error!\n");
+    }
+
+    // Cheching Byte1
+    actual_Byte1 = (uint8_t)( (actual_word >> 8 ) & 0x000000FF);
+    golden_Byte1 = (uint8_t)( (golden_word >> 8 ) & 0x000000FF);
+
+    diff = (actual_Byte1 > golden_Byte1) ? (actual_Byte1 - golden_Byte1) : 0;
+    diff = (actual_Byte1 < golden_Byte1) ? (golden_Byte1 - actual_Byte1) : 0;
+
+    if (diff > ERR) {
+      error = 1;
+      tfp_printf ("diff: 0x%08x\n", diff);
+      tfp_printf ("Byte1: Error!\n");
+    }
+
+    // Cheching Byte2
+    actual_Byte2 = (uint8_t)( (actual_word >> 16 ) & 0x000000FF);
+    golden_Byte2 = (uint8_t)( (golden_word >> 16 ) & 0x000000FF);
+
+    diff = (actual_Byte2 > golden_Byte2) ? (actual_Byte2 - golden_Byte2) : 0;
+    diff = (actual_Byte2 < golden_Byte2) ? (golden_Byte2 - actual_Byte2) : 0;
+
+    if (diff > ERR) {
+      error = 1;
+      tfp_printf ("diff: 0x%08x\n", diff);
+      tfp_printf ("Byte2: Error!\n");
+    }
+
+    // Cheching Byte3
+    actual_Byte3 = (uint8_t)( (actual_word >> 24 ) & 0x000000FF);
+    golden_Byte3 = (uint8_t)( (golden_word >> 24 ) & 0x000000FF);
+
+    diff = (actual_Byte3 > golden_Byte3) ? (actual_Byte3 - golden_Byte3) : 0;
+    diff = (actual_Byte3 < golden_Byte3) ? (golden_Byte3 - actual_Byte3) : 0;
+
+    if (diff > ERR) {
+      error = 1;
+      tfp_printf ("diff: 0x%08x\n", diff);
+      tfp_printf ("Byte3: Error!\n");
+    }
+    
+    errors += error;
+
+    #ifdef DEBUG
+      tfp_printf("  Golden: 0x%08x; Actual: 0x%08x,\n", golden_word, actual_word);
+    #endif
+
+    #ifdef VERBOSE
+      if(error) {
+        if(errors==1) tfp_printf("  golden     <- actual     @ address    @ index\n");
+        tfp_printf("  0x%08x <- 0x%08x @ 0x%08x @ 0x%08x\n", golden_word, actual_word, (actual_z+i), i*4);
+      }
+    #endif
+  }
+  return errors;
+}
+
+#endif
diff --git a/hwpe/redmule/inc/golden.h b/hwpe/redmule/inc/golden.h
new file mode 100644
index 0000000..f664e47
--- /dev/null
+++ b/hwpe/redmule/inc/golden.h
@@ -0,0 +1,387 @@
+ /* Header file generated by RedMulE Golden Model */
+uint32_t golden [384] = {
+0x48974845,
+0x48384608,
+0x487b4855,
+0x48804869,
+0x48b046d1,
+0x483f48db,
+0x485f48c9,
+0x483a4881,
+0x472c484b,
+0x492b4762,
+0x48fd4822,
+0x492e488e,
+0x484f483e,
+0x46d749e8,
+0x489d484b,
+0x47e9490b,
+0x47d2484f,
+0x474744be,
+0x46c047c7,
+0x48af4727,
+0x482d46c5,
+0x482e483d,
+0x479f4897,
+0x4749488b,
+0x46a8489a,
+0x488b46f2,
+0x47e84891,
+0x483d4872,
+0x46fd4716,
+0x46a049b5,
+0x47a446e7,
+0x476748a1,
+0x49354939,
+0x48c14703,
+0x48bd4863,
+0x48cf4913,
+0x48b848b6,
+0x49204946,
+0x48e1495e,
+0x48b24938,
+0x4882493a,
+0x49d5483b,
+0x49724911,
+0x49df496b,
+0x488848f2,
+0x48214a46,
+0x490c48c1,
+0x48a349b2,
+0x47b0463a,
+0x476244cb,
+0x46b94765,
+0x4814466a,
+0x47964631,
+0x474b4666,
+0x47044798,
+0x47614838,
+0x459047d3,
+0x48a245ea,
+0x484447f1,
+0x4776484b,
+0x46d847d6,
+0x44d348f3,
+0x478d46fa,
+0x466e481e,
+0x481e4827,
+0x479445a2,
+0x48064727,
+0x48d5475d,
+0x48284708,
+0x480d4862,
+0x48324895,
+0x47f148bd,
+0x46a7482a,
+0x492d47b1,
+0x4884484d,
+0x485f48dc,
+0x480c476d,
+0x46d348e9,
+0x48844728,
+0x480e48a0,
+0x48134862,
+0x485a4675,
+0x473847e8,
+0x48234836,
+0x482146e7,
+0x47b34822,
+0x48554846,
+0x47174863,
+0x47c14872,
+0x488e46d5,
+0x485f47e2,
+0x48b8487c,
+0x4788481e,
+0x467748bd,
+0x47f846c9,
+0x47fc48fe,
+0x47b247a0,
+0x467e4588,
+0x46c74662,
+0x481246e8,
+0x474e4536,
+0x468f46c0,
+0x4679481f,
+0x46e246a1,
+0x45604809,
+0x47eb4630,
+0x475746b5,
+0x477f4848,
+0x46d846a6,
+0x459a4870,
+0x46784670,
+0x468c47d2,
+0x48c44762,
+0x479146e3,
+0x486d46b1,
+0x486747d0,
+0x47f6468d,
+0x475648a5,
+0x48544857,
+0x48384866,
+0x46ec484d,
+0x48f647d2,
+0x4879484a,
+0x483c4848,
+0x4806471d,
+0x473048fa,
+0x47b84768,
+0x46f94865,
+0x491848a8,
+0x486746ca,
+0x48624800,
+0x491048d3,
+0x4849474e,
+0x486b48eb,
+0x48c54966,
+0x483048f4,
+0x477848f9,
+0x499e481e,
+0x48f148cf,
+0x49234982,
+0x47cf487c,
+0x464949ea,
+0x495e4773,
+0x483f48b2,
+0x497548a7,
+0x481e4616,
+0x4866481f,
+0x486448b6,
+0x487347dc,
+0x487f485c,
+0x491f4938,
+0x48b6490d,
+0x48a148f8,
+0x492d4859,
+0x4915489c,
+0x48874899,
+0x4859486c,
+0x471e49ca,
+0x49184867,
+0x482748d3,
+0x4998488b,
+0x481d4704,
+0x488048b8,
+0x49444876,
+0x48f2470c,
+0x489b48b9,
+0x48e54956,
+0x48a548d6,
+0x485648dc,
+0x49ab484e,
+0x490e48e0,
+0x494548dd,
+0x48dd488b,
+0x47ea4a32,
+0x49114835,
+0x48194965,
+0x481e460e,
+0x4673452c,
+0x4717475c,
+0x46d046f6,
+0x46bc4696,
+0x481e4726,
+0x46ea4763,
+0x475846fe,
+0x4627478b,
+0x483f4704,
+0x47b146ad,
+0x48164792,
+0x468446f2,
+0x45a84827,
+0x47a4472f,
+0x462b4797,
+0x48ab483f,
+0x4863468f,
+0x4766485a,
+0x48cb481d,
+0x490347dc,
+0x483048fc,
+0x483e48cc,
+0x486448ab,
+0x47634966,
+0x499d4794,
+0x488b488e,
+0x496048dc,
+0x484c4854,
+0x474c499c,
+0x48bc4826,
+0x48834949,
+0x4905489d,
+0x481e4718,
+0x48f448e3,
+0x490448c1,
+0x48b347e8,
+0x48d44892,
+0x489448ff,
+0x488648d5,
+0x480348fa,
+0x492e47d2,
+0x48b24870,
+0x492b48e5,
+0x4785487b,
+0x471d49e3,
+0x48bf4837,
+0x48c4489b,
+0x4871475c,
+0x4811464a,
+0x471c47af,
+0x48174817,
+0x484e463b,
+0x464f477f,
+0x487c4704,
+0x472547a3,
+0x462a4853,
+0x4860465a,
+0x48804736,
+0x482b47e1,
+0x46c04811,
+0x475d48dc,
+0x48064668,
+0x46f44893,
+0x49594858,
+0x487b463d,
+0x484e480f,
+0x48a648c0,
+0x48944847,
+0x484a48a0,
+0x48f4491e,
+0x48b548fc,
+0x47d248ce,
+0x497f47db,
+0x49394955,
+0x48ce48a7,
+0x48844890,
+0x476349d6,
+0x4922486e,
+0x48c348f4,
+0x491c47ec,
+0x47834698,
+0x47544715,
+0x47524745,
+0x4832472f,
+0x48094817,
+0x48c347f8,
+0x480047e6,
+0x473048b6,
+0x48cb480a,
+0x488e479e,
+0x488e47c2,
+0x47ee472f,
+0x4744489d,
+0x48514755,
+0x47d34846,
+0x48a04838,
+0x47624634,
+0x48064786,
+0x482d47e3,
+0x486c4726,
+0x480347b7,
+0x481448ac,
+0x483948e0,
+0x47504827,
+0x48c546f2,
+0x4886483f,
+0x485648ad,
+0x47a947e8,
+0x47434937,
+0x481f46d0,
+0x4804484c,
+0x481f47fd,
+0x4813456d,
+0x4807474d,
+0x480e4688,
+0x481046e8,
+0x4799469f,
+0x478f4853,
+0x482447f2,
+0x471f47d0,
+0x485f46da,
+0x481c4813,
+0x4863482e,
+0x480b4786,
+0x46b848c9,
+0x46e2475a,
+0x46c54852,
+0x480245af,
+0x46c24466,
+0x4743465d,
+0x47ba46b7,
+0x46c34636,
+0x47844677,
+0x47c2485a,
+0x46ac46dc,
+0x460e47de,
+0x4834465f,
+0x476947f4,
+0x481046fc,
+0x45ea45fd,
+0x45b548d0,
+0x47834704,
+0x46c44830,
+0x47c74759,
+0x45b0453d,
+0x47024741,
+0x47934736,
+0x47ba461b,
+0x46dd470b,
+0x470b4657,
+0x4710470d,
+0x468f486c,
+0x46ba45c3,
+0x483b479d,
+0x477446c9,
+0x46a746a9,
+0x46064833,
+0x46a94690,
+0x46a746f5,
+0x48bb47ac,
+0x4803452c,
+0x4824470f,
+0x48cb47d5,
+0x484a4707,
+0x47974832,
+0x482c4851,
+0x4877487a,
+0x465d4891,
+0x48ce47f4,
+0x48994898,
+0x486a484e,
+0x47f047ac,
+0x4611493e,
+0x489e47e2,
+0x46af488c,
+0x48364665,
+0x46b645e4,
+0x46b946a1,
+0x46dd46c8,
+0x474b4658,
+0x4777467b,
+0x47984769,
+0x475e4785,
+0x4656472a,
+0x488145fb,
+0x472d46fc,
+0x47a3476e,
+0x46ca465d,
+0x45004855,
+0x479a464f,
+0x473846c3,
+0x486c481e,
+0x48014659,
+0x477a4756,
+0x487b47d5,
+0x48084706,
+0x4838484f,
+0x48634870,
+0x480648d3,
+0x47714865,
+0x494c46be,
+0x484c4915,
+0x48624900,
+0x46e8481a,
+0x46a04974,
+0x483d4775,
+0x480e487c,
+};
\ No newline at end of file
diff --git a/hwpe/redmule/inc/tensor_dim.h b/hwpe/redmule/inc/tensor_dim.h
new file mode 100644
index 0000000..21bd0d8
--- /dev/null
+++ b/hwpe/redmule/inc/tensor_dim.h
@@ -0,0 +1,13 @@
+ /* Header file generated by RedMulE Golden Model */
+#ifndef __TENSOR_DIM__
+#define __TENSOR_DIM__
+
+#define M_SIZE  24 
+#define N_SIZE  32 
+#define K_SIZE  32
+#define SRC_FMT FP16
+#define DST_FMT FP16
+#define FPFORMAT 16
+uint8_t gemm_ops = GEMM; 
+
+#endif
diff --git a/hwpe/redmule/inc/w_2D.h b/hwpe/redmule/inc/w_2D.h
new file mode 100644
index 0000000..9409c64
--- /dev/null
+++ b/hwpe/redmule/inc/w_2D.h
@@ -0,0 +1,35 @@
+ /* Header file generated by RedMulE Golden Model */
+uint16_t w_inp_2D [32][32] = {
+0x311a, 0x39e0, 0x387d, 0x3a4a, 0x386f, 0x3ada, 0x392f, 0x3854, 0x3014, 0x2fd2, 0x31c9, 0x2fca, 0x2e55, 0x3bc8, 0x396d, 0x3b1d, 0x39f6, 0x333a, 0x3908, 0x3628, 0x3bab, 0x3b8b, 0x3b4a, 0x322d, 0x3925, 0x317a, 0x3725, 0x31c2, 0x3066, 0x38f3, 0x3a17, 0x3476, 
+0x3bda, 0x3196, 0x3922, 0x3680, 0x396a, 0x3021, 0x3761, 0x374d, 0x2fc2, 0x3967, 0x3b94, 0x33b5, 0x3797, 0x34d6, 0x3655, 0x2176, 0x39bc, 0x3999, 0x3658, 0x3904, 0x3759, 0x2ade, 0x3a5a, 0x3b78, 0x36c7, 0x2d01, 0x3b58, 0x2d9a, 0x373d, 0x3952, 0x38e8, 0x3887, 
+0x37b6, 0x3a88, 0x2f8a, 0x2d79, 0x3413, 0x3421, 0x3976, 0x32b2, 0x3446, 0x2d99, 0x3a56, 0x3322, 0x3b49, 0x39fa, 0x3acd, 0x3af6, 0x304c, 0x3abb, 0x3a83, 0x38b2, 0x3ab9, 0x363e, 0x389f, 0x31bb, 0x38e1, 0x3bc4, 0x3b9b, 0x2984, 0x3a43, 0x3b2f, 0x35d6, 0x3bda, 
+0x2df3, 0x3bf8, 0x2acc, 0x378b, 0x3555, 0x2e59, 0x31d4, 0x34ec, 0x3a46, 0x3bab, 0x3214, 0x3161, 0x3470, 0x3a03, 0x368e, 0x31ad, 0x27cb, 0x2ecb, 0x3422, 0x39f7, 0x3644, 0x3a77, 0x313f, 0x34f2, 0x39b3, 0x3bf2, 0x379a, 0x3456, 0x35fe, 0x3ae7, 0x3964, 0x385f, 
+0x3b16, 0x3999, 0x3833, 0x2eda, 0x3afd, 0x3a4a, 0x3ba2, 0x2bd4, 0x3b38, 0x31a2, 0x32dd, 0x353c, 0x366f, 0x375e, 0x3821, 0x367a, 0x3b44, 0x39e6, 0x3787, 0x339e, 0x39d7, 0x38c6, 0x37d5, 0x342f, 0x3984, 0x319b, 0x33b5, 0x35ab, 0x398a, 0x374e, 0x36b6, 0x3b21, 
+0x3bbb, 0x2ab3, 0x2ad5, 0x33bc, 0x2bef, 0x3780, 0x3738, 0x3a0b, 0x3b09, 0x30ca, 0x384e, 0x3ab3, 0x39bd, 0x3453, 0x3a6d, 0x3957, 0x2c10, 0x30e9, 0x35d4, 0x3aef, 0x3be9, 0x39ad, 0x3a74, 0x3af9, 0x3739, 0x2d4d, 0x39fe, 0x3b72, 0x2c57, 0x398c, 0x381f, 0x3930, 
+0x3820, 0x321b, 0x3964, 0x2964, 0x33a0, 0x2d00, 0x2490, 0x336b, 0x3465, 0x3b2e, 0x3aa0, 0x371f, 0x300e, 0x3a09, 0x3bf1, 0x25cc, 0x3b6f, 0x3384, 0x3a88, 0x3acb, 0x3814, 0x36d0, 0x3081, 0x3a2c, 0x3353, 0x39cb, 0x31ed, 0x3af6, 0x3721, 0x36c7, 0x2ce2, 0x390d, 
+0x3698, 0x3ab2, 0x3b3e, 0x2eb4, 0x3998, 0x39e3, 0x3a77, 0x3632, 0x2c12, 0x3bd5, 0x3ba3, 0x3bba, 0x323c, 0x367b, 0x3557, 0x39c8, 0x37db, 0x3b45, 0x3b6e, 0x3931, 0x3121, 0x3a8d, 0x3a55, 0x3b9b, 0x358a, 0x3925, 0x3491, 0x3912, 0x3b6b, 0x3584, 0x32df, 0x3120, 
+0x32b2, 0x3b0a, 0x2cad, 0x3465, 0x3ad3, 0x3bcd, 0x363b, 0x3afe, 0x354b, 0x3374, 0x39af, 0x3b7f, 0x308c, 0x2e72, 0x3380, 0x3b70, 0x3902, 0x38d8, 0x39f3, 0x3a4b, 0x3853, 0x397b, 0x2ebe, 0x387f, 0x2845, 0x37e2, 0x360f, 0x370b, 0x3acb, 0x35d4, 0x36e6, 0x3262, 
+0x2e88, 0x3a54, 0x2ee3, 0x3575, 0x3afe, 0x2aee, 0x39a0, 0x3aae, 0x3693, 0x3432, 0x3834, 0x3b9b, 0x3bcb, 0x2e3a, 0x356d, 0x374e, 0x3924, 0x383c, 0x311e, 0x3ac5, 0x352d, 0x311e, 0x38ca, 0x34d4, 0x36ca, 0x34ed, 0x3a13, 0x33eb, 0x3639, 0x3828, 0x3b3c, 0x3939, 
+0x3837, 0x3521, 0x2cb5, 0x3629, 0x3924, 0x384c, 0x366a, 0x3bbf, 0x2e9e, 0x3ba8, 0x33ad, 0x38c8, 0x3934, 0x3907, 0x249a, 0x3690, 0x3a09, 0x3215, 0x3898, 0x325d, 0x37d5, 0x3195, 0x361c, 0x3ae4, 0x351f, 0x3452, 0x3bc0, 0x375c, 0x39bf, 0x317a, 0x3aae, 0x283a, 
+0x3476, 0x3b92, 0x3472, 0x383e, 0x280f, 0x39d6, 0x2fd1, 0x31f4, 0x2ffb, 0x3b97, 0x3692, 0x36c0, 0x3989, 0x33cf, 0x3ba6, 0x3239, 0x35d7, 0x33ab, 0x31eb, 0x3b47, 0x389b, 0x3b88, 0x3580, 0x354c, 0x3802, 0x3b9a, 0x3b94, 0x2a92, 0x2db1, 0x38bd, 0x2dfb, 0x3900, 
+0x344f, 0x3739, 0x27a5, 0x3b2e, 0x342b, 0x34bb, 0x30c8, 0x3ae8, 0x3b26, 0x3982, 0x38c0, 0x3408, 0x38c8, 0x36ef, 0x3bf0, 0x3acf, 0x3a3c, 0x3825, 0x31a5, 0x3ada, 0x3b5b, 0x37db, 0x3a01, 0x3663, 0x3a7d, 0x327b, 0x3a1f, 0x3862, 0x38af, 0x3204, 0x372e, 0x3b19, 
+0x3708, 0x3622, 0x2e62, 0x39ab, 0x2d4d, 0x31b4, 0x3552, 0x3bbc, 0x36f2, 0x36eb, 0x38ef, 0x3755, 0x3bbe, 0x2c17, 0x3815, 0x2f53, 0x363f, 0x38c1, 0x3246, 0x386b, 0x34de, 0x34e4, 0x3baa, 0x349e, 0x32ce, 0x3a68, 0x373f, 0x2cce, 0x3b36, 0x28ba, 0x3b50, 0x3232, 
+0x1f34, 0x3928, 0x35cd, 0x3b38, 0x30ce, 0x35a1, 0x3a06, 0x3a32, 0x3a53, 0x3489, 0x3241, 0x372f, 0x390c, 0x3a1b, 0x378a, 0x3713, 0x3769, 0x37a8, 0x3418, 0x3ad4, 0x3a4e, 0x3bf7, 0x37a5, 0x34dc, 0x39b2, 0x351b, 0x3372, 0x349f, 0x2f50, 0x3ab1, 0x3795, 0x2db7, 
+0x3864, 0x3157, 0x3900, 0x323e, 0x389e, 0x3880, 0x3b1f, 0x37a1, 0x396c, 0x2e43, 0x2c2a, 0x3b78, 0x3988, 0x3a14, 0x39c1, 0x3b51, 0x3780, 0x3bf2, 0x2d19, 0x3815, 0x3a5f, 0x3641, 0x2f62, 0x37d5, 0x3564, 0x139a, 0x3ab8, 0x28f7, 0x3785, 0x34e1, 0x3097, 0x3768, 
+0x3971, 0x3ae2, 0x32ae, 0x2fd5, 0x382a, 0x346c, 0x3133, 0x3167, 0x3940, 0x2d12, 0x389a, 0x3bd0, 0x3943, 0x391c, 0x3a75, 0x2a11, 0x391e, 0x372d, 0x3a79, 0x3b72, 0x3373, 0x39b7, 0x35d7, 0x372b, 0x3a6d, 0x38a1, 0x3279, 0x3434, 0x3694, 0x3b45, 0x3abb, 0x392d, 
+0x34a8, 0x3757, 0x32ca, 0x345d, 0x36a5, 0x3854, 0x2dcd, 0x30af, 0x38dd, 0x3067, 0x3411, 0x3997, 0x397a, 0x3a64, 0x38b8, 0x3962, 0x3509, 0x3bb6, 0x3a66, 0x339f, 0x372a, 0x31a8, 0x37da, 0x36ff, 0x33c6, 0x31da, 0x3977, 0x3b72, 0x3841, 0x3567, 0x3433, 0x33b8, 
+0x39fe, 0x3a10, 0x3bf2, 0x35e7, 0x3a4a, 0x3b3e, 0x2ec7, 0x3aa4, 0x3846, 0x3af9, 0x38a9, 0x2c1f, 0x39ab, 0x349f, 0x31d6, 0x39ae, 0x3b79, 0x352d, 0x3516, 0x347c, 0x2f33, 0x35ad, 0x31c4, 0x3b52, 0x354b, 0x3786, 0x3ab7, 0x3896, 0x34ac, 0x352f, 0x37e6, 0x326a, 
+0x2e44, 0x34c7, 0x388d, 0x3bf4, 0x363f, 0x3b3d, 0x33b1, 0x3b8b, 0x3340, 0x37f7, 0x3b07, 0x25bf, 0x398e, 0x3505, 0x3bd7, 0x366d, 0x388a, 0x2cc0, 0x359a, 0x3b9a, 0x3b99, 0x379d, 0x3b6b, 0x39b8, 0x3223, 0x2703, 0x3ba9, 0x2ecb, 0x3759, 0x39d8, 0x37ac, 0x32cf, 
+0x35f2, 0x38a3, 0x399e, 0x3bd2, 0x3780, 0x3af3, 0x3b5e, 0x337b, 0x3a08, 0x35da, 0x3446, 0x3b25, 0x3ad0, 0x3bee, 0x3141, 0x32d8, 0x34ce, 0x2ac9, 0x3800, 0x3a8a, 0x2d53, 0x368a, 0x3561, 0x3998, 0x35a3, 0x3677, 0x3ab2, 0x3269, 0x3236, 0x3b3e, 0x3aba, 0x3bac, 
+0x395d, 0x3820, 0x1df6, 0x3bb5, 0x35b5, 0x3675, 0x3b74, 0x360f, 0x34de, 0x3a0c, 0x3aeb, 0x299d, 0x3207, 0x3bd8, 0x2178, 0x3995, 0x3948, 0x3908, 0x3843, 0x2ea5, 0x3045, 0x3989, 0x345d, 0x39c5, 0x3a89, 0x3863, 0x3be0, 0x397a, 0x38f1, 0x39e2, 0x3b08, 0x352e, 
+0x385f, 0x28f2, 0x3bc3, 0x35e0, 0x380c, 0x3b9c, 0x3afc, 0x390a, 0x3689, 0x34fd, 0x2cf5, 0x308e, 0x342b, 0x3921, 0x3a67, 0x3ad6, 0x2986, 0x32fc, 0x35aa, 0x3507, 0x3608, 0x33fd, 0x3bf3, 0x39e2, 0x3b0f, 0x30b7, 0x3896, 0x3ae4, 0x2145, 0x35b6, 0x2e1d, 0x3ad1, 
+0x333d, 0x3afb, 0x2703, 0x3413, 0x1d7d, 0x3b7f, 0x3ae1, 0x303c, 0x3004, 0x39d3, 0x3554, 0x31a4, 0x354e, 0x3662, 0x39c5, 0x2eb7, 0x2c6e, 0x397f, 0x31d8, 0x1f0c, 0x38e3, 0x35f0, 0x2714, 0x28d1, 0x375e, 0x3a75, 0x3830, 0x3578, 0x397d, 0x3b18, 0x383c, 0x3498, 
+0x39ad, 0x3598, 0x23c4, 0x34ea, 0x3a61, 0x2b00, 0x3707, 0x3ae1, 0x37ae, 0x389d, 0x37fa, 0x3673, 0x3278, 0xf3e, 0x3809, 0x33c6, 0x3bf5, 0x3279, 0x3816, 0x360c, 0x39c8, 0x381f, 0x3741, 0x2d66, 0x38c0, 0x37d3, 0x377a, 0x3621, 0x2faf, 0x392e, 0x2de6, 0x33c5, 
+0x3803, 0x2600, 0x32e9, 0x39b4, 0x38d2, 0x34e8, 0x2fe6, 0x3199, 0x3643, 0x3a77, 0x27cc, 0x39d7, 0x34c6, 0x2ea8, 0x364e, 0x3b07, 0x31c7, 0x30a1, 0x31b1, 0x3b8f, 0x3571, 0x3b75, 0x3989, 0x3805, 0x39fb, 0x3945, 0x352b, 0x31d8, 0x3904, 0x3440, 0x3a57, 0x2cf7, 
+0x3b39, 0x2fcd, 0x2b89, 0x2edd, 0x3682, 0x36a9, 0x32c8, 0x37ac, 0x32a5, 0x3311, 0x394b, 0x3b84, 0x3aec, 0x3601, 0x2765, 0x3b69, 0x396b, 0x3727, 0x3bfe, 0x3907, 0x376f, 0x3674, 0x3973, 0x3671, 0x3491, 0x3993, 0x383f, 0x3335, 0x3989, 0x3550, 0x3077, 0x35f5, 
+0x3a59, 0x3950, 0x380c, 0x37cd, 0x30bf, 0x3607, 0x3afa, 0x3b5d, 0x32b9, 0x386b, 0x35bd, 0x3aca, 0x3ba5, 0x3b2d, 0x3b19, 0x3b8b, 0x345e, 0x2845, 0x34aa, 0x372a, 0x3448, 0x34f5, 0x3ae2, 0x3637, 0x2cb5, 0x354b, 0x3b15, 0x2ca8, 0x2641, 0x3178, 0x2cfe, 0x39b4, 
+0x3bdd, 0x3acb, 0x3a05, 0x38a2, 0x3b4a, 0x34e5, 0x395f, 0x394b, 0x34c4, 0x3aa5, 0x29bb, 0x2d96, 0x339d, 0x387c, 0x382e, 0x385a, 0x396b, 0x3aa9, 0x2f1e, 0x33a7, 0x3b90, 0x3b7b, 0x3b5f, 0x39d3, 0x3b18, 0x354f, 0x2cdb, 0x3a6f, 0x3434, 0x34ff, 0x3a5b, 0x3b84, 
+0x3a33, 0x384b, 0x2e67, 0x3b85, 0x3853, 0x380c, 0x346a, 0x3aaa, 0x3492, 0x33e8, 0x3bf2, 0x38ae, 0x3a29, 0x3830, 0x3221, 0x35b1, 0x3a48, 0x2c68, 0x2ced, 0x3a7e, 0x3539, 0x3922, 0x374c, 0x3aaa, 0x2dae, 0x395d, 0x3b3d, 0x3890, 0x2cfe, 0x2dd6, 0x3bad, 0x33c5, 
+0x2c07, 0x3a2c, 0x37a8, 0x390f, 0x2fc8, 0x35ae, 0x388c, 0x30ee, 0x3674, 0x391d, 0x3bfc, 0x36bf, 0x322d, 0x3a78, 0x35c0, 0x3492, 0x3ac8, 0x3504, 0x3315, 0x381d, 0x3a7a, 0x3a08, 0x343c, 0x3bda, 0x341b, 0x39f0, 0x3b9e, 0x395d, 0x3c00, 0x38ab, 0x3bcf, 0x3564, 
+0x33c4, 0x3b0d, 0x3623, 0x33b9, 0x3b92, 0x1e71, 0x2c57, 0x36d0, 0x314b, 0x3a16, 0x3372, 0x341b, 0x3aaa, 0x3444, 0x396b, 0x2dd7, 0x3b30, 0x3559, 0x3b5b, 0x3a29, 0x2d19, 0x38b7, 0x3b01, 0x3afa, 0x398a, 0x3839, 0x3ac9, 0x2e31, 0x3924, 0x39f2, 0x3a7f, 0x3285 
+};
\ No newline at end of file
diff --git a/hwpe/redmule/inc/w_input.h b/hwpe/redmule/inc/w_input.h
new file mode 100644
index 0000000..dc4d3be
--- /dev/null
+++ b/hwpe/redmule/inc/w_input.h
@@ -0,0 +1,35 @@
+ /* Header file generated by RedMulE Golden Model */
+uint16_t w_inp [1024] = {
+0x311a, 0x39e0, 0x387d, 0x3a4a, 0x386f, 0x3ada, 0x392f, 0x3854, 0x3014, 0x2fd2, 0x31c9, 0x2fca, 0x2e55, 0x3bc8, 0x396d, 0x3b1d, 0x39f6, 0x333a, 0x3908, 0x3628, 0x3bab, 0x3b8b, 0x3b4a, 0x322d, 0x3925, 0x317a, 0x3725, 0x31c2, 0x3066, 0x38f3, 0x3a17, 0x3476, 
+0x3bda, 0x3196, 0x3922, 0x3680, 0x396a, 0x3021, 0x3761, 0x374d, 0x2fc2, 0x3967, 0x3b94, 0x33b5, 0x3797, 0x34d6, 0x3655, 0x2176, 0x39bc, 0x3999, 0x3658, 0x3904, 0x3759, 0x2ade, 0x3a5a, 0x3b78, 0x36c7, 0x2d01, 0x3b58, 0x2d9a, 0x373d, 0x3952, 0x38e8, 0x3887, 
+0x37b6, 0x3a88, 0x2f8a, 0x2d79, 0x3413, 0x3421, 0x3976, 0x32b2, 0x3446, 0x2d99, 0x3a56, 0x3322, 0x3b49, 0x39fa, 0x3acd, 0x3af6, 0x304c, 0x3abb, 0x3a83, 0x38b2, 0x3ab9, 0x363e, 0x389f, 0x31bb, 0x38e1, 0x3bc4, 0x3b9b, 0x2984, 0x3a43, 0x3b2f, 0x35d6, 0x3bda, 
+0x2df3, 0x3bf8, 0x2acc, 0x378b, 0x3555, 0x2e59, 0x31d4, 0x34ec, 0x3a46, 0x3bab, 0x3214, 0x3161, 0x3470, 0x3a03, 0x368e, 0x31ad, 0x27cb, 0x2ecb, 0x3422, 0x39f7, 0x3644, 0x3a77, 0x313f, 0x34f2, 0x39b3, 0x3bf2, 0x379a, 0x3456, 0x35fe, 0x3ae7, 0x3964, 0x385f, 
+0x3b16, 0x3999, 0x3833, 0x2eda, 0x3afd, 0x3a4a, 0x3ba2, 0x2bd4, 0x3b38, 0x31a2, 0x32dd, 0x353c, 0x366f, 0x375e, 0x3821, 0x367a, 0x3b44, 0x39e6, 0x3787, 0x339e, 0x39d7, 0x38c6, 0x37d5, 0x342f, 0x3984, 0x319b, 0x33b5, 0x35ab, 0x398a, 0x374e, 0x36b6, 0x3b21, 
+0x3bbb, 0x2ab3, 0x2ad5, 0x33bc, 0x2bef, 0x3780, 0x3738, 0x3a0b, 0x3b09, 0x30ca, 0x384e, 0x3ab3, 0x39bd, 0x3453, 0x3a6d, 0x3957, 0x2c10, 0x30e9, 0x35d4, 0x3aef, 0x3be9, 0x39ad, 0x3a74, 0x3af9, 0x3739, 0x2d4d, 0x39fe, 0x3b72, 0x2c57, 0x398c, 0x381f, 0x3930, 
+0x3820, 0x321b, 0x3964, 0x2964, 0x33a0, 0x2d00, 0x2490, 0x336b, 0x3465, 0x3b2e, 0x3aa0, 0x371f, 0x300e, 0x3a09, 0x3bf1, 0x25cc, 0x3b6f, 0x3384, 0x3a88, 0x3acb, 0x3814, 0x36d0, 0x3081, 0x3a2c, 0x3353, 0x39cb, 0x31ed, 0x3af6, 0x3721, 0x36c7, 0x2ce2, 0x390d, 
+0x3698, 0x3ab2, 0x3b3e, 0x2eb4, 0x3998, 0x39e3, 0x3a77, 0x3632, 0x2c12, 0x3bd5, 0x3ba3, 0x3bba, 0x323c, 0x367b, 0x3557, 0x39c8, 0x37db, 0x3b45, 0x3b6e, 0x3931, 0x3121, 0x3a8d, 0x3a55, 0x3b9b, 0x358a, 0x3925, 0x3491, 0x3912, 0x3b6b, 0x3584, 0x32df, 0x3120, 
+0x32b2, 0x3b0a, 0x2cad, 0x3465, 0x3ad3, 0x3bcd, 0x363b, 0x3afe, 0x354b, 0x3374, 0x39af, 0x3b7f, 0x308c, 0x2e72, 0x3380, 0x3b70, 0x3902, 0x38d8, 0x39f3, 0x3a4b, 0x3853, 0x397b, 0x2ebe, 0x387f, 0x2845, 0x37e2, 0x360f, 0x370b, 0x3acb, 0x35d4, 0x36e6, 0x3262, 
+0x2e88, 0x3a54, 0x2ee3, 0x3575, 0x3afe, 0x2aee, 0x39a0, 0x3aae, 0x3693, 0x3432, 0x3834, 0x3b9b, 0x3bcb, 0x2e3a, 0x356d, 0x374e, 0x3924, 0x383c, 0x311e, 0x3ac5, 0x352d, 0x311e, 0x38ca, 0x34d4, 0x36ca, 0x34ed, 0x3a13, 0x33eb, 0x3639, 0x3828, 0x3b3c, 0x3939, 
+0x3837, 0x3521, 0x2cb5, 0x3629, 0x3924, 0x384c, 0x366a, 0x3bbf, 0x2e9e, 0x3ba8, 0x33ad, 0x38c8, 0x3934, 0x3907, 0x249a, 0x3690, 0x3a09, 0x3215, 0x3898, 0x325d, 0x37d5, 0x3195, 0x361c, 0x3ae4, 0x351f, 0x3452, 0x3bc0, 0x375c, 0x39bf, 0x317a, 0x3aae, 0x283a, 
+0x3476, 0x3b92, 0x3472, 0x383e, 0x280f, 0x39d6, 0x2fd1, 0x31f4, 0x2ffb, 0x3b97, 0x3692, 0x36c0, 0x3989, 0x33cf, 0x3ba6, 0x3239, 0x35d7, 0x33ab, 0x31eb, 0x3b47, 0x389b, 0x3b88, 0x3580, 0x354c, 0x3802, 0x3b9a, 0x3b94, 0x2a92, 0x2db1, 0x38bd, 0x2dfb, 0x3900, 
+0x344f, 0x3739, 0x27a5, 0x3b2e, 0x342b, 0x34bb, 0x30c8, 0x3ae8, 0x3b26, 0x3982, 0x38c0, 0x3408, 0x38c8, 0x36ef, 0x3bf0, 0x3acf, 0x3a3c, 0x3825, 0x31a5, 0x3ada, 0x3b5b, 0x37db, 0x3a01, 0x3663, 0x3a7d, 0x327b, 0x3a1f, 0x3862, 0x38af, 0x3204, 0x372e, 0x3b19, 
+0x3708, 0x3622, 0x2e62, 0x39ab, 0x2d4d, 0x31b4, 0x3552, 0x3bbc, 0x36f2, 0x36eb, 0x38ef, 0x3755, 0x3bbe, 0x2c17, 0x3815, 0x2f53, 0x363f, 0x38c1, 0x3246, 0x386b, 0x34de, 0x34e4, 0x3baa, 0x349e, 0x32ce, 0x3a68, 0x373f, 0x2cce, 0x3b36, 0x28ba, 0x3b50, 0x3232, 
+0x1f34, 0x3928, 0x35cd, 0x3b38, 0x30ce, 0x35a1, 0x3a06, 0x3a32, 0x3a53, 0x3489, 0x3241, 0x372f, 0x390c, 0x3a1b, 0x378a, 0x3713, 0x3769, 0x37a8, 0x3418, 0x3ad4, 0x3a4e, 0x3bf7, 0x37a5, 0x34dc, 0x39b2, 0x351b, 0x3372, 0x349f, 0x2f50, 0x3ab1, 0x3795, 0x2db7, 
+0x3864, 0x3157, 0x3900, 0x323e, 0x389e, 0x3880, 0x3b1f, 0x37a1, 0x396c, 0x2e43, 0x2c2a, 0x3b78, 0x3988, 0x3a14, 0x39c1, 0x3b51, 0x3780, 0x3bf2, 0x2d19, 0x3815, 0x3a5f, 0x3641, 0x2f62, 0x37d5, 0x3564, 0x139a, 0x3ab8, 0x28f7, 0x3785, 0x34e1, 0x3097, 0x3768, 
+0x3971, 0x3ae2, 0x32ae, 0x2fd5, 0x382a, 0x346c, 0x3133, 0x3167, 0x3940, 0x2d12, 0x389a, 0x3bd0, 0x3943, 0x391c, 0x3a75, 0x2a11, 0x391e, 0x372d, 0x3a79, 0x3b72, 0x3373, 0x39b7, 0x35d7, 0x372b, 0x3a6d, 0x38a1, 0x3279, 0x3434, 0x3694, 0x3b45, 0x3abb, 0x392d, 
+0x34a8, 0x3757, 0x32ca, 0x345d, 0x36a5, 0x3854, 0x2dcd, 0x30af, 0x38dd, 0x3067, 0x3411, 0x3997, 0x397a, 0x3a64, 0x38b8, 0x3962, 0x3509, 0x3bb6, 0x3a66, 0x339f, 0x372a, 0x31a8, 0x37da, 0x36ff, 0x33c6, 0x31da, 0x3977, 0x3b72, 0x3841, 0x3567, 0x3433, 0x33b8, 
+0x39fe, 0x3a10, 0x3bf2, 0x35e7, 0x3a4a, 0x3b3e, 0x2ec7, 0x3aa4, 0x3846, 0x3af9, 0x38a9, 0x2c1f, 0x39ab, 0x349f, 0x31d6, 0x39ae, 0x3b79, 0x352d, 0x3516, 0x347c, 0x2f33, 0x35ad, 0x31c4, 0x3b52, 0x354b, 0x3786, 0x3ab7, 0x3896, 0x34ac, 0x352f, 0x37e6, 0x326a, 
+0x2e44, 0x34c7, 0x388d, 0x3bf4, 0x363f, 0x3b3d, 0x33b1, 0x3b8b, 0x3340, 0x37f7, 0x3b07, 0x25bf, 0x398e, 0x3505, 0x3bd7, 0x366d, 0x388a, 0x2cc0, 0x359a, 0x3b9a, 0x3b99, 0x379d, 0x3b6b, 0x39b8, 0x3223, 0x2703, 0x3ba9, 0x2ecb, 0x3759, 0x39d8, 0x37ac, 0x32cf, 
+0x35f2, 0x38a3, 0x399e, 0x3bd2, 0x3780, 0x3af3, 0x3b5e, 0x337b, 0x3a08, 0x35da, 0x3446, 0x3b25, 0x3ad0, 0x3bee, 0x3141, 0x32d8, 0x34ce, 0x2ac9, 0x3800, 0x3a8a, 0x2d53, 0x368a, 0x3561, 0x3998, 0x35a3, 0x3677, 0x3ab2, 0x3269, 0x3236, 0x3b3e, 0x3aba, 0x3bac, 
+0x395d, 0x3820, 0x1df6, 0x3bb5, 0x35b5, 0x3675, 0x3b74, 0x360f, 0x34de, 0x3a0c, 0x3aeb, 0x299d, 0x3207, 0x3bd8, 0x2178, 0x3995, 0x3948, 0x3908, 0x3843, 0x2ea5, 0x3045, 0x3989, 0x345d, 0x39c5, 0x3a89, 0x3863, 0x3be0, 0x397a, 0x38f1, 0x39e2, 0x3b08, 0x352e, 
+0x385f, 0x28f2, 0x3bc3, 0x35e0, 0x380c, 0x3b9c, 0x3afc, 0x390a, 0x3689, 0x34fd, 0x2cf5, 0x308e, 0x342b, 0x3921, 0x3a67, 0x3ad6, 0x2986, 0x32fc, 0x35aa, 0x3507, 0x3608, 0x33fd, 0x3bf3, 0x39e2, 0x3b0f, 0x30b7, 0x3896, 0x3ae4, 0x2145, 0x35b6, 0x2e1d, 0x3ad1, 
+0x333d, 0x3afb, 0x2703, 0x3413, 0x1d7d, 0x3b7f, 0x3ae1, 0x303c, 0x3004, 0x39d3, 0x3554, 0x31a4, 0x354e, 0x3662, 0x39c5, 0x2eb7, 0x2c6e, 0x397f, 0x31d8, 0x1f0c, 0x38e3, 0x35f0, 0x2714, 0x28d1, 0x375e, 0x3a75, 0x3830, 0x3578, 0x397d, 0x3b18, 0x383c, 0x3498, 
+0x39ad, 0x3598, 0x23c4, 0x34ea, 0x3a61, 0x2b00, 0x3707, 0x3ae1, 0x37ae, 0x389d, 0x37fa, 0x3673, 0x3278, 0xf3e, 0x3809, 0x33c6, 0x3bf5, 0x3279, 0x3816, 0x360c, 0x39c8, 0x381f, 0x3741, 0x2d66, 0x38c0, 0x37d3, 0x377a, 0x3621, 0x2faf, 0x392e, 0x2de6, 0x33c5, 
+0x3803, 0x2600, 0x32e9, 0x39b4, 0x38d2, 0x34e8, 0x2fe6, 0x3199, 0x3643, 0x3a77, 0x27cc, 0x39d7, 0x34c6, 0x2ea8, 0x364e, 0x3b07, 0x31c7, 0x30a1, 0x31b1, 0x3b8f, 0x3571, 0x3b75, 0x3989, 0x3805, 0x39fb, 0x3945, 0x352b, 0x31d8, 0x3904, 0x3440, 0x3a57, 0x2cf7, 
+0x3b39, 0x2fcd, 0x2b89, 0x2edd, 0x3682, 0x36a9, 0x32c8, 0x37ac, 0x32a5, 0x3311, 0x394b, 0x3b84, 0x3aec, 0x3601, 0x2765, 0x3b69, 0x396b, 0x3727, 0x3bfe, 0x3907, 0x376f, 0x3674, 0x3973, 0x3671, 0x3491, 0x3993, 0x383f, 0x3335, 0x3989, 0x3550, 0x3077, 0x35f5, 
+0x3a59, 0x3950, 0x380c, 0x37cd, 0x30bf, 0x3607, 0x3afa, 0x3b5d, 0x32b9, 0x386b, 0x35bd, 0x3aca, 0x3ba5, 0x3b2d, 0x3b19, 0x3b8b, 0x345e, 0x2845, 0x34aa, 0x372a, 0x3448, 0x34f5, 0x3ae2, 0x3637, 0x2cb5, 0x354b, 0x3b15, 0x2ca8, 0x2641, 0x3178, 0x2cfe, 0x39b4, 
+0x3bdd, 0x3acb, 0x3a05, 0x38a2, 0x3b4a, 0x34e5, 0x395f, 0x394b, 0x34c4, 0x3aa5, 0x29bb, 0x2d96, 0x339d, 0x387c, 0x382e, 0x385a, 0x396b, 0x3aa9, 0x2f1e, 0x33a7, 0x3b90, 0x3b7b, 0x3b5f, 0x39d3, 0x3b18, 0x354f, 0x2cdb, 0x3a6f, 0x3434, 0x34ff, 0x3a5b, 0x3b84, 
+0x3a33, 0x384b, 0x2e67, 0x3b85, 0x3853, 0x380c, 0x346a, 0x3aaa, 0x3492, 0x33e8, 0x3bf2, 0x38ae, 0x3a29, 0x3830, 0x3221, 0x35b1, 0x3a48, 0x2c68, 0x2ced, 0x3a7e, 0x3539, 0x3922, 0x374c, 0x3aaa, 0x2dae, 0x395d, 0x3b3d, 0x3890, 0x2cfe, 0x2dd6, 0x3bad, 0x33c5, 
+0x2c07, 0x3a2c, 0x37a8, 0x390f, 0x2fc8, 0x35ae, 0x388c, 0x30ee, 0x3674, 0x391d, 0x3bfc, 0x36bf, 0x322d, 0x3a78, 0x35c0, 0x3492, 0x3ac8, 0x3504, 0x3315, 0x381d, 0x3a7a, 0x3a08, 0x343c, 0x3bda, 0x341b, 0x39f0, 0x3b9e, 0x395d, 0x3c00, 0x38ab, 0x3bcf, 0x3564, 
+0x33c4, 0x3b0d, 0x3623, 0x33b9, 0x3b92, 0x1e71, 0x2c57, 0x36d0, 0x314b, 0x3a16, 0x3372, 0x341b, 0x3aaa, 0x3444, 0x396b, 0x2dd7, 0x3b30, 0x3559, 0x3b5b, 0x3a29, 0x2d19, 0x38b7, 0x3b01, 0x3afa, 0x398a, 0x3839, 0x3ac9, 0x2e31, 0x3924, 0x39f2, 0x3a7f, 0x3285 
+};
\ No newline at end of file
diff --git a/hwpe/redmule/inc/x_2D.h b/hwpe/redmule/inc/x_2D.h
new file mode 100644
index 0000000..0b589f8
--- /dev/null
+++ b/hwpe/redmule/inc/x_2D.h
@@ -0,0 +1,27 @@
+ /* Header file generated by RedMulE Golden Model */
+uint16_t x_inp_2D [24][32] = {
+0x2153, 0x3bb5, 0x3896, 0x365f, 0x2483, 0x3518, 0x2dd1, 0x3bca, 0x397b, 0x29b1, 0x3705, 0x36c8, 0x398b, 0x3661, 0x2f05, 0x365a, 0x3bf9, 0x34df, 0x363b, 0x38d9, 0x39c6, 0x3abb, 0x3952, 0x38f2, 0x392d, 0x3b3e, 0x2afb, 0x3a9d, 0x353b, 0x3b73, 0x3a01, 0x3679, 
+0x3934, 0x397d, 0x2904, 0x3822, 0x3462, 0x3b44, 0x39e9, 0x28be, 0x331e, 0x3a1d, 0x39e5, 0x34da, 0x3a19, 0x3906, 0x1d35, 0x3871, 0x31e7, 0x3b29, 0x325d, 0x3797, 0x2b2f, 0x38b4, 0x232f, 0x38aa, 0x3aca, 0x316f, 0x3811, 0x3950, 0x32ea, 0x3bc7, 0x382c, 0x38a2, 
+0x29ce, 0x3afa, 0x3a39, 0x2ccc, 0x39fd, 0x3b3d, 0x384a, 0x3a35, 0x3802, 0x366a, 0x37ec, 0x3598, 0x3bf8, 0x3a85, 0x3a1b, 0x386e, 0x3b4c, 0x39de, 0x38c2, 0x2f93, 0x3b4c, 0x39c4, 0x3b9e, 0x3844, 0x346d, 0x3bff, 0x32ce, 0x296d, 0x3130, 0x3b3d, 0x3b44, 0x369d, 
+0x3b13, 0x31ed, 0x330a, 0x3831, 0x34e7, 0x37b3, 0x331a, 0x3918, 0x32d3, 0x3995, 0x3991, 0x3919, 0x3a26, 0x385b, 0x2b76, 0x3a3b, 0x37f2, 0x26a7, 0x3225, 0x3b64, 0x28f0, 0x3456, 0x3822, 0x341e, 0x381a, 0x38d8, 0x2c11, 0x33be, 0x33ac, 0x353f, 0x3476, 0x3abc, 
+0x36ec, 0x3a1d, 0x39d3, 0x3821, 0x36ac, 0x3bce, 0x3ad2, 0x3616, 0x36a1, 0x2cb3, 0x38d2, 0x314f, 0x385c, 0x3b63, 0x3bb6, 0x2951, 0x372d, 0x2c42, 0x3823, 0x3883, 0x3872, 0x31ee, 0x36c5, 0x399a, 0x31b0, 0x3887, 0x3884, 0x3865, 0x3896, 0x36c3, 0x32e3, 0x346c, 
+0x3935, 0x3b50, 0x2b6d, 0x38cd, 0x388f, 0x3389, 0x395d, 0x31cd, 0x2efd, 0x3154, 0x2f35, 0x3444, 0x3293, 0x3b6b, 0x1bec, 0x3b69, 0x3bf3, 0x3611, 0x3508, 0x3742, 0x3a50, 0x3ab7, 0x3457, 0x38d3, 0x3344, 0x38e8, 0x33c0, 0x3668, 0x3bee, 0x3b21, 0x3727, 0x3121, 
+0x316c, 0x3288, 0x2d50, 0x2e74, 0x35d5, 0x37e2, 0x303d, 0x36af, 0x341f, 0x3436, 0x2df7, 0x399d, 0x30f4, 0x3aaf, 0x34e4, 0x2c2a, 0x3116, 0x34d3, 0x36ac, 0x35e3, 0x3760, 0x36e1, 0x3ad2, 0x3547, 0x38f4, 0x369c, 0x3ba9, 0x34f0, 0x3a39, 0x3b19, 0x36e6, 0x395d, 
+0x3be8, 0x3293, 0x3bfc, 0x3435, 0x2eb3, 0x3360, 0x3919, 0x3bed, 0x396a, 0x37fc, 0x3242, 0x384b, 0x38cb, 0x3b2c, 0x3b28, 0x28cf, 0x3828, 0x3855, 0x3ba9, 0x2fa7, 0x340b, 0x32f1, 0x3ada, 0x36fa, 0x31f5, 0x3436, 0x29d0, 0x33e6, 0x3232, 0x3bec, 0x3904, 0x2797, 
+0x3b81, 0x3bac, 0x38d2, 0x343d, 0x31af, 0x3b1e, 0x33fc, 0x3864, 0x3624, 0x3905, 0x2945, 0x3b52, 0x2d08, 0x3a17, 0x3b84, 0x3804, 0x3a24, 0x38a3, 0x3562, 0x3ae6, 0x3bba, 0x3a45, 0x3679, 0x31fa, 0x3994, 0x2c3d, 0x383f, 0x399d, 0x34f7, 0x360e, 0x35f3, 0x38f0, 
+0x38d4, 0x399a, 0x3a48, 0x3987, 0x3b54, 0x382c, 0x3210, 0x35ef, 0x36ca, 0x31b4, 0x3625, 0x371f, 0x37bd, 0x3680, 0x3a3a, 0x3ac0, 0x3bbf, 0x3bf5, 0x39f2, 0x29c2, 0x363e, 0x3a4e, 0x3596, 0x3b1b, 0x3459, 0x3669, 0x3aa1, 0x39c3, 0x3376, 0x390d, 0x2456, 0x39b5, 
+0x3a66, 0x3ad8, 0x3b51, 0x36aa, 0x32be, 0x3ac8, 0x392b, 0x3740, 0x3a48, 0x38f5, 0x3b2d, 0x3a5f, 0x2ff3, 0x366f, 0x39d3, 0x35e5, 0x3822, 0x38db, 0x3b8a, 0x34be, 0x2d33, 0x36dd, 0x3578, 0x3bdf, 0x2c7e, 0x39cf, 0x32ff, 0x35c9, 0x3970, 0x3bcb, 0x351e, 0x3956, 
+0x2c42, 0x3308, 0x377a, 0x361c, 0x39a0, 0x36c9, 0x2dcb, 0x3bf2, 0x3b5f, 0x33ee, 0x24c1, 0x2ce9, 0x3927, 0x305d, 0x3702, 0x3119, 0x35f9, 0x3855, 0x3374, 0x349b, 0x3bcf, 0x2dea, 0x34f0, 0x363f, 0x37da, 0x3a74, 0x35fc, 0x35fa, 0x316b, 0x3804, 0x37a7, 0x3986, 
+0x3073, 0x3aed, 0x31c7, 0x3844, 0x34a4, 0x387d, 0x3a20, 0x3037, 0x3a00, 0x3b70, 0x377f, 0x3686, 0x3b7e, 0x38b3, 0x32e3, 0x3323, 0x391e, 0x3228, 0x3930, 0x3997, 0x3a5e, 0x398b, 0x3512, 0x35b0, 0x365c, 0x325d, 0x3b61, 0x38b8, 0x39a4, 0x3423, 0x3bd7, 0x38af, 
+0x2d3d, 0x382d, 0x38ac, 0x26ca, 0x395e, 0x21a8, 0x3520, 0x386f, 0x3b95, 0x32c0, 0x3b84, 0x3a51, 0x3b4b, 0x31d2, 0x3747, 0x3b96, 0x3b40, 0x3535, 0x38d1, 0x3899, 0x3b00, 0x3827, 0x3ae3, 0x38c8, 0x3a07, 0x338d, 0x2e96, 0x3a46, 0x394a, 0x39de, 0x2951, 0x3a02, 
+0x3838, 0x2d45, 0x28c0, 0x3958, 0x3070, 0x2aa2, 0x3510, 0x38ce, 0x271c, 0x3440, 0x3954, 0x30bc, 0x3b35, 0x2f1d, 0x3afb, 0x2dae, 0x356f, 0x2e13, 0x3981, 0x326d, 0x3a28, 0x3a36, 0x3a95, 0x38cb, 0x38db, 0x3150, 0x2c9e, 0x34c5, 0x3adb, 0x3bdf, 0x38f2, 0x3994, 
+0x36f8, 0x31c0, 0x3a4f, 0x3825, 0x394b, 0x3a8b, 0x38ac, 0x3167, 0x2e2d, 0x3a93, 0x34f3, 0x37bd, 0x3b63, 0x2f2f, 0x3ae0, 0x3ad8, 0x34a8, 0x2e1c, 0x3890, 0x3705, 0x3b69, 0x3bc1, 0x28af, 0x3b36, 0x348b, 0x3111, 0x3a8d, 0x389c, 0x3916, 0x36dc, 0x3bae, 0x3874, 
+0x3593, 0x3638, 0x3018, 0x3a56, 0x38a3, 0x2ad4, 0x3a25, 0x38d7, 0x3864, 0x31c1, 0x28d1, 0x39c8, 0x37d6, 0x2c7f, 0x3ba5, 0x34b8, 0x3bef, 0x3b83, 0x3ab5, 0x3062, 0x38bc, 0x399c, 0x2ce4, 0x2f2c, 0x39bf, 0x2ed1, 0x385f, 0x37e0, 0x35ee, 0x397d, 0x3b0c, 0x3049, 
+0x39d5, 0x322e, 0x3936, 0x3747, 0x2e15, 0x3b41, 0x3874, 0x3bd0, 0x2c04, 0x3800, 0x375b, 0x3b2d, 0x38d8, 0x3a51, 0x3406, 0x38da, 0x38ba, 0x3497, 0x382e, 0x35fc, 0x39d4, 0x3775, 0x3b1e, 0x3813, 0x3649, 0x31af, 0x37bb, 0x334a, 0x3a6e, 0x3284, 0x26e0, 0x2e01, 
+0x2ebb, 0x344b, 0x3821, 0x381a, 0x385a, 0x2534, 0x3635, 0x2a92, 0x3b8c, 0x31f0, 0x3947, 0x3ac7, 0x3743, 0x3924, 0x39e4, 0x358f, 0x2b62, 0x392c, 0x3955, 0x3341, 0x3676, 0x38ac, 0x3957, 0x335b, 0x2ca2, 0x39ff, 0x37cb, 0x341f, 0x3ac9, 0x3b6c, 0x2f14, 0x34c3, 
+0x3018, 0x3169, 0x355b, 0x3624, 0x31ed, 0x379e, 0x3268, 0x309b, 0x35db, 0x3872, 0x3bdb, 0x34c7, 0x3408, 0x3359, 0x3920, 0x331f, 0x3866, 0x3af0, 0x2a1a, 0x39e0, 0x3b14, 0x34fa, 0x2d18, 0x3963, 0x35e8, 0x2539, 0x38f5, 0x37b3, 0x378f, 0x31b5, 0x3a6c, 0x3685, 
+0x3a06, 0x318a, 0x2934, 0x33c1, 0x3be8, 0x375b, 0x3860, 0x3543, 0x3702, 0x3951, 0x3677, 0x37ff, 0x2e27, 0x2e3a, 0x340f, 0x3817, 0x2f04, 0x357e, 0x3a1d, 0x2dd6, 0x252a, 0x3945, 0x162a, 0x3b19, 0x3a53, 0x35d2, 0x3a5d, 0x3474, 0x38e9, 0x374b, 0x387c, 0x1f1a, 
+0x38ac, 0x3291, 0x3393, 0x3b53, 0x3169, 0x3bca, 0x2f1a, 0x3551, 0x38a3, 0x28e3, 0x369d, 0x34a1, 0x38a8, 0x34c3, 0x3841, 0x390d, 0x3b13, 0x3282, 0x3a29, 0x3a78, 0x2df3, 0x3a37, 0x35f4, 0x35a6, 0x38e8, 0x3328, 0x3beb, 0x390b, 0x32dc, 0x34dc, 0x396d, 0x3a78, 
+0x39ba, 0x3a06, 0x2cdd, 0x3bc3, 0x2d43, 0x2992, 0x3663, 0x3a68, 0x2c3e, 0x394e, 0x2c9f, 0x380e, 0x37f5, 0x3557, 0x2873, 0x390f, 0x39e7, 0x3939, 0x3669, 0x385c, 0x3a68, 0x32c4, 0x2b04, 0x2d6d, 0x39d3, 0x3895, 0x331d, 0x3b59, 0x3463, 0x2b6a, 0x31de, 0x3296, 
+0x3aae, 0x3bcd, 0x345a, 0x3897, 0x374b, 0x3bd4, 0x38a2, 0x357f, 0x3402, 0x3a0c, 0x3507, 0x3865, 0x3a54, 0x3878, 0x3859, 0x383e, 0x32b5, 0x34ea, 0x328d, 0x38b6, 0x3464, 0x2f5b, 0x35ff, 0x3817, 0x2f24, 0x3533, 0x3b21, 0x37ba, 0x3837, 0x2e34, 0x3bad, 0x34bc 
+};
\ No newline at end of file
diff --git a/hwpe/redmule/inc/x_input.h b/hwpe/redmule/inc/x_input.h
new file mode 100644
index 0000000..1e38d23
--- /dev/null
+++ b/hwpe/redmule/inc/x_input.h
@@ -0,0 +1,27 @@
+ /* Header file generated by RedMulE Golden Model */
+uint16_t x_inp [768] = {
+0x2153, 0x3bb5, 0x3896, 0x365f, 0x2483, 0x3518, 0x2dd1, 0x3bca, 0x397b, 0x29b1, 0x3705, 0x36c8, 0x398b, 0x3661, 0x2f05, 0x365a, 0x3bf9, 0x34df, 0x363b, 0x38d9, 0x39c6, 0x3abb, 0x3952, 0x38f2, 0x392d, 0x3b3e, 0x2afb, 0x3a9d, 0x353b, 0x3b73, 0x3a01, 0x3679, 
+0x3934, 0x397d, 0x2904, 0x3822, 0x3462, 0x3b44, 0x39e9, 0x28be, 0x331e, 0x3a1d, 0x39e5, 0x34da, 0x3a19, 0x3906, 0x1d35, 0x3871, 0x31e7, 0x3b29, 0x325d, 0x3797, 0x2b2f, 0x38b4, 0x232f, 0x38aa, 0x3aca, 0x316f, 0x3811, 0x3950, 0x32ea, 0x3bc7, 0x382c, 0x38a2, 
+0x29ce, 0x3afa, 0x3a39, 0x2ccc, 0x39fd, 0x3b3d, 0x384a, 0x3a35, 0x3802, 0x366a, 0x37ec, 0x3598, 0x3bf8, 0x3a85, 0x3a1b, 0x386e, 0x3b4c, 0x39de, 0x38c2, 0x2f93, 0x3b4c, 0x39c4, 0x3b9e, 0x3844, 0x346d, 0x3bff, 0x32ce, 0x296d, 0x3130, 0x3b3d, 0x3b44, 0x369d, 
+0x3b13, 0x31ed, 0x330a, 0x3831, 0x34e7, 0x37b3, 0x331a, 0x3918, 0x32d3, 0x3995, 0x3991, 0x3919, 0x3a26, 0x385b, 0x2b76, 0x3a3b, 0x37f2, 0x26a7, 0x3225, 0x3b64, 0x28f0, 0x3456, 0x3822, 0x341e, 0x381a, 0x38d8, 0x2c11, 0x33be, 0x33ac, 0x353f, 0x3476, 0x3abc, 
+0x36ec, 0x3a1d, 0x39d3, 0x3821, 0x36ac, 0x3bce, 0x3ad2, 0x3616, 0x36a1, 0x2cb3, 0x38d2, 0x314f, 0x385c, 0x3b63, 0x3bb6, 0x2951, 0x372d, 0x2c42, 0x3823, 0x3883, 0x3872, 0x31ee, 0x36c5, 0x399a, 0x31b0, 0x3887, 0x3884, 0x3865, 0x3896, 0x36c3, 0x32e3, 0x346c, 
+0x3935, 0x3b50, 0x2b6d, 0x38cd, 0x388f, 0x3389, 0x395d, 0x31cd, 0x2efd, 0x3154, 0x2f35, 0x3444, 0x3293, 0x3b6b, 0x1bec, 0x3b69, 0x3bf3, 0x3611, 0x3508, 0x3742, 0x3a50, 0x3ab7, 0x3457, 0x38d3, 0x3344, 0x38e8, 0x33c0, 0x3668, 0x3bee, 0x3b21, 0x3727, 0x3121, 
+0x316c, 0x3288, 0x2d50, 0x2e74, 0x35d5, 0x37e2, 0x303d, 0x36af, 0x341f, 0x3436, 0x2df7, 0x399d, 0x30f4, 0x3aaf, 0x34e4, 0x2c2a, 0x3116, 0x34d3, 0x36ac, 0x35e3, 0x3760, 0x36e1, 0x3ad2, 0x3547, 0x38f4, 0x369c, 0x3ba9, 0x34f0, 0x3a39, 0x3b19, 0x36e6, 0x395d, 
+0x3be8, 0x3293, 0x3bfc, 0x3435, 0x2eb3, 0x3360, 0x3919, 0x3bed, 0x396a, 0x37fc, 0x3242, 0x384b, 0x38cb, 0x3b2c, 0x3b28, 0x28cf, 0x3828, 0x3855, 0x3ba9, 0x2fa7, 0x340b, 0x32f1, 0x3ada, 0x36fa, 0x31f5, 0x3436, 0x29d0, 0x33e6, 0x3232, 0x3bec, 0x3904, 0x2797, 
+0x3b81, 0x3bac, 0x38d2, 0x343d, 0x31af, 0x3b1e, 0x33fc, 0x3864, 0x3624, 0x3905, 0x2945, 0x3b52, 0x2d08, 0x3a17, 0x3b84, 0x3804, 0x3a24, 0x38a3, 0x3562, 0x3ae6, 0x3bba, 0x3a45, 0x3679, 0x31fa, 0x3994, 0x2c3d, 0x383f, 0x399d, 0x34f7, 0x360e, 0x35f3, 0x38f0, 
+0x38d4, 0x399a, 0x3a48, 0x3987, 0x3b54, 0x382c, 0x3210, 0x35ef, 0x36ca, 0x31b4, 0x3625, 0x371f, 0x37bd, 0x3680, 0x3a3a, 0x3ac0, 0x3bbf, 0x3bf5, 0x39f2, 0x29c2, 0x363e, 0x3a4e, 0x3596, 0x3b1b, 0x3459, 0x3669, 0x3aa1, 0x39c3, 0x3376, 0x390d, 0x2456, 0x39b5, 
+0x3a66, 0x3ad8, 0x3b51, 0x36aa, 0x32be, 0x3ac8, 0x392b, 0x3740, 0x3a48, 0x38f5, 0x3b2d, 0x3a5f, 0x2ff3, 0x366f, 0x39d3, 0x35e5, 0x3822, 0x38db, 0x3b8a, 0x34be, 0x2d33, 0x36dd, 0x3578, 0x3bdf, 0x2c7e, 0x39cf, 0x32ff, 0x35c9, 0x3970, 0x3bcb, 0x351e, 0x3956, 
+0x2c42, 0x3308, 0x377a, 0x361c, 0x39a0, 0x36c9, 0x2dcb, 0x3bf2, 0x3b5f, 0x33ee, 0x24c1, 0x2ce9, 0x3927, 0x305d, 0x3702, 0x3119, 0x35f9, 0x3855, 0x3374, 0x349b, 0x3bcf, 0x2dea, 0x34f0, 0x363f, 0x37da, 0x3a74, 0x35fc, 0x35fa, 0x316b, 0x3804, 0x37a7, 0x3986, 
+0x3073, 0x3aed, 0x31c7, 0x3844, 0x34a4, 0x387d, 0x3a20, 0x3037, 0x3a00, 0x3b70, 0x377f, 0x3686, 0x3b7e, 0x38b3, 0x32e3, 0x3323, 0x391e, 0x3228, 0x3930, 0x3997, 0x3a5e, 0x398b, 0x3512, 0x35b0, 0x365c, 0x325d, 0x3b61, 0x38b8, 0x39a4, 0x3423, 0x3bd7, 0x38af, 
+0x2d3d, 0x382d, 0x38ac, 0x26ca, 0x395e, 0x21a8, 0x3520, 0x386f, 0x3b95, 0x32c0, 0x3b84, 0x3a51, 0x3b4b, 0x31d2, 0x3747, 0x3b96, 0x3b40, 0x3535, 0x38d1, 0x3899, 0x3b00, 0x3827, 0x3ae3, 0x38c8, 0x3a07, 0x338d, 0x2e96, 0x3a46, 0x394a, 0x39de, 0x2951, 0x3a02, 
+0x3838, 0x2d45, 0x28c0, 0x3958, 0x3070, 0x2aa2, 0x3510, 0x38ce, 0x271c, 0x3440, 0x3954, 0x30bc, 0x3b35, 0x2f1d, 0x3afb, 0x2dae, 0x356f, 0x2e13, 0x3981, 0x326d, 0x3a28, 0x3a36, 0x3a95, 0x38cb, 0x38db, 0x3150, 0x2c9e, 0x34c5, 0x3adb, 0x3bdf, 0x38f2, 0x3994, 
+0x36f8, 0x31c0, 0x3a4f, 0x3825, 0x394b, 0x3a8b, 0x38ac, 0x3167, 0x2e2d, 0x3a93, 0x34f3, 0x37bd, 0x3b63, 0x2f2f, 0x3ae0, 0x3ad8, 0x34a8, 0x2e1c, 0x3890, 0x3705, 0x3b69, 0x3bc1, 0x28af, 0x3b36, 0x348b, 0x3111, 0x3a8d, 0x389c, 0x3916, 0x36dc, 0x3bae, 0x3874, 
+0x3593, 0x3638, 0x3018, 0x3a56, 0x38a3, 0x2ad4, 0x3a25, 0x38d7, 0x3864, 0x31c1, 0x28d1, 0x39c8, 0x37d6, 0x2c7f, 0x3ba5, 0x34b8, 0x3bef, 0x3b83, 0x3ab5, 0x3062, 0x38bc, 0x399c, 0x2ce4, 0x2f2c, 0x39bf, 0x2ed1, 0x385f, 0x37e0, 0x35ee, 0x397d, 0x3b0c, 0x3049, 
+0x39d5, 0x322e, 0x3936, 0x3747, 0x2e15, 0x3b41, 0x3874, 0x3bd0, 0x2c04, 0x3800, 0x375b, 0x3b2d, 0x38d8, 0x3a51, 0x3406, 0x38da, 0x38ba, 0x3497, 0x382e, 0x35fc, 0x39d4, 0x3775, 0x3b1e, 0x3813, 0x3649, 0x31af, 0x37bb, 0x334a, 0x3a6e, 0x3284, 0x26e0, 0x2e01, 
+0x2ebb, 0x344b, 0x3821, 0x381a, 0x385a, 0x2534, 0x3635, 0x2a92, 0x3b8c, 0x31f0, 0x3947, 0x3ac7, 0x3743, 0x3924, 0x39e4, 0x358f, 0x2b62, 0x392c, 0x3955, 0x3341, 0x3676, 0x38ac, 0x3957, 0x335b, 0x2ca2, 0x39ff, 0x37cb, 0x341f, 0x3ac9, 0x3b6c, 0x2f14, 0x34c3, 
+0x3018, 0x3169, 0x355b, 0x3624, 0x31ed, 0x379e, 0x3268, 0x309b, 0x35db, 0x3872, 0x3bdb, 0x34c7, 0x3408, 0x3359, 0x3920, 0x331f, 0x3866, 0x3af0, 0x2a1a, 0x39e0, 0x3b14, 0x34fa, 0x2d18, 0x3963, 0x35e8, 0x2539, 0x38f5, 0x37b3, 0x378f, 0x31b5, 0x3a6c, 0x3685, 
+0x3a06, 0x318a, 0x2934, 0x33c1, 0x3be8, 0x375b, 0x3860, 0x3543, 0x3702, 0x3951, 0x3677, 0x37ff, 0x2e27, 0x2e3a, 0x340f, 0x3817, 0x2f04, 0x357e, 0x3a1d, 0x2dd6, 0x252a, 0x3945, 0x162a, 0x3b19, 0x3a53, 0x35d2, 0x3a5d, 0x3474, 0x38e9, 0x374b, 0x387c, 0x1f1a, 
+0x38ac, 0x3291, 0x3393, 0x3b53, 0x3169, 0x3bca, 0x2f1a, 0x3551, 0x38a3, 0x28e3, 0x369d, 0x34a1, 0x38a8, 0x34c3, 0x3841, 0x390d, 0x3b13, 0x3282, 0x3a29, 0x3a78, 0x2df3, 0x3a37, 0x35f4, 0x35a6, 0x38e8, 0x3328, 0x3beb, 0x390b, 0x32dc, 0x34dc, 0x396d, 0x3a78, 
+0x39ba, 0x3a06, 0x2cdd, 0x3bc3, 0x2d43, 0x2992, 0x3663, 0x3a68, 0x2c3e, 0x394e, 0x2c9f, 0x380e, 0x37f5, 0x3557, 0x2873, 0x390f, 0x39e7, 0x3939, 0x3669, 0x385c, 0x3a68, 0x32c4, 0x2b04, 0x2d6d, 0x39d3, 0x3895, 0x331d, 0x3b59, 0x3463, 0x2b6a, 0x31de, 0x3296, 
+0x3aae, 0x3bcd, 0x345a, 0x3897, 0x374b, 0x3bd4, 0x38a2, 0x357f, 0x3402, 0x3a0c, 0x3507, 0x3865, 0x3a54, 0x3878, 0x3859, 0x383e, 0x32b5, 0x34ea, 0x328d, 0x38b6, 0x3464, 0x2f5b, 0x35ff, 0x3817, 0x2f24, 0x3533, 0x3b21, 0x37ba, 0x3837, 0x2e34, 0x3bad, 0x34bc 
+};
\ No newline at end of file
diff --git a/hwpe/redmule/inc/y_2D.h b/hwpe/redmule/inc/y_2D.h
new file mode 100644
index 0000000..9484a10
--- /dev/null
+++ b/hwpe/redmule/inc/y_2D.h
@@ -0,0 +1,27 @@
+ /* Header file generated by RedMulE Golden Model */
+uint16_t y_inp_2D [32][32] = {
+0x3150, 0x2dc1, 0x3033, 0x31f5, 0x3bb6, 0x3bff, 0x39f9, 0x3662, 0x3720, 0x351d, 0x384b, 0x3093, 0x3b9d, 0x35ad, 0x3695, 0x3466, 0x2300, 0x3445, 0x33ae, 0x3586, 0x38a3, 0x3bdb, 0x33a2, 0x379b, 0x3a0e, 0x38b0, 0x39ba, 0x379b, 0x39d3, 0x3a51, 0x3b30, 0x3794, 
+0x3b76, 0x3042, 0x38cc, 0x2dfc, 0x3b1a, 0x37fb, 0x38f7, 0x3824, 0x386f, 0x38c7, 0x36ee, 0x3a9c, 0x38d3, 0x2c67, 0x3a80, 0x2f30, 0x3328, 0x3721, 0x3790, 0x34e5, 0x3a6c, 0x3643, 0x3934, 0x3034, 0x38d4, 0x362e, 0x3b4b, 0x3408, 0x30c2, 0x370e, 0x3b31, 0x3b16, 
+0x3b6b, 0x39d4, 0x339c, 0x381e, 0x313e, 0x3671, 0x3ae2, 0x3479, 0x3940, 0x342d, 0x3925, 0x370a, 0x35d8, 0x2dad, 0x3888, 0x24b9, 0x375d, 0x34bd, 0x3243, 0x2ebb, 0x3970, 0x3a21, 0x3a07, 0x3877, 0x3888, 0x3569, 0x372d, 0x2ac1, 0x331e, 0x384d, 0x3996, 0x34a4, 
+0x35c1, 0x33a9, 0x21ed, 0x3a42, 0x388d, 0x34e4, 0x33c3, 0x34f9, 0x3a7b, 0x33fb, 0x2cdd, 0x3b0e, 0x333b, 0x3973, 0x34fc, 0x3771, 0x32ea, 0x2de4, 0x31a8, 0x3946, 0x3657, 0x3a4e, 0x36f6, 0x2829, 0x3ba2, 0x3bdc, 0x3bb3, 0x306c, 0x398d, 0x3a1f, 0x3991, 0x3846, 
+0x3547, 0x3292, 0x2e85, 0x31ed, 0x3979, 0x3a90, 0x28a4, 0x3bed, 0x36d8, 0x340e, 0x3b6a, 0x3ab6, 0x3824, 0x382b, 0x3ac3, 0x3811, 0x36d7, 0x3519, 0x3a92, 0x3a42, 0x29d1, 0x383a, 0x3a9b, 0x300e, 0x2cd3, 0x39cd, 0x3874, 0x3a07, 0x2eb1, 0x3b86, 0x3ad8, 0x3a5d, 
+0x3712, 0x284a, 0x38c1, 0x3bec, 0x39c0, 0x32cd, 0x3ad8, 0x3bce, 0x3817, 0x3896, 0x3aa7, 0x3870, 0x3996, 0x32cc, 0x3a4c, 0x3757, 0x3814, 0x3b65, 0x3acb, 0x376e, 0x34c0, 0x3609, 0x3bf0, 0x3b24, 0x3b29, 0x3848, 0x34b7, 0x398a, 0x220c, 0x3498, 0x3a8c, 0x3883, 
+0x38c4, 0x3af6, 0x3a42, 0x2dd6, 0x3147, 0x3717, 0x3a8e, 0x3af9, 0x3296, 0x38ef, 0x34fa, 0x3555, 0x3b29, 0x38de, 0x315e, 0x3773, 0x3b67, 0x3116, 0x38ec, 0x357c, 0x35d0, 0x2518, 0x3958, 0x2a03, 0x37d9, 0x3699, 0x3a1e, 0x3230, 0x3b13, 0x36d4, 0x3b2a, 0x39ad, 
+0x3b10, 0x351a, 0x3b97, 0x3326, 0x2b54, 0x3b7d, 0x386f, 0x373e, 0x37fa, 0x389b, 0x3b90, 0x3292, 0x3975, 0x38f3, 0x37f1, 0x3590, 0x3810, 0x2fd7, 0x3bf7, 0x3a5a, 0x3a1c, 0x34dd, 0x354c, 0x32f8, 0x3095, 0x321e, 0x39e0, 0x395c, 0x3717, 0x357f, 0x394a, 0x34b1, 
+0x3ba4, 0x380c, 0x3604, 0x2f50, 0x348d, 0x3828, 0x3a9f, 0x39ce, 0x32ca, 0x3906, 0x3ab2, 0x2ca5, 0x38c9, 0x362a, 0x34b2, 0x29dc, 0x3a36, 0x3052, 0x31b7, 0x3589, 0x387c, 0x3401, 0x3b22, 0x3ad6, 0x3ae8, 0x3238, 0x3494, 0x3502, 0x3717, 0x3a6c, 0x3229, 0x368c, 
+0x3056, 0x3a56, 0x3498, 0x39eb, 0x2864, 0x342d, 0x39e0, 0x34a1, 0x2b99, 0x3a04, 0x38ff, 0x328c, 0x34d9, 0x387d, 0x3a3c, 0x32e5, 0x39eb, 0x3984, 0x34dd, 0x38a7, 0x373f, 0x39b4, 0x3235, 0x2f58, 0x2f39, 0x3800, 0x3758, 0x3939, 0x39fc, 0x3a4b, 0x38bf, 0x30ee, 
+0x345e, 0x39c8, 0x3a6d, 0x3262, 0x3b81, 0x31dc, 0x3a15, 0x3bd0, 0x36af, 0x36de, 0x37d5, 0x39d7, 0x3ad3, 0x3ac1, 0x3109, 0x35ea, 0x31c6, 0x398d, 0x3987, 0x3a4a, 0x34d2, 0x2ed2, 0x35e6, 0x352c, 0x39eb, 0x3bd6, 0x3a5b, 0x39d1, 0x34aa, 0x3ade, 0x394b, 0x38a1, 
+0x2bed, 0x38de, 0x3811, 0x3813, 0x391a, 0x374b, 0x3829, 0x3725, 0x38f0, 0x3583, 0x3966, 0x3a7d, 0x375a, 0x38fe, 0x3696, 0x361c, 0x39a8, 0x35f0, 0x38e1, 0x3003, 0x3595, 0x316e, 0x3862, 0x3af8, 0x3af2, 0x34c8, 0x381d, 0x37d8, 0x3893, 0x3a9c, 0x3989, 0x308c, 
+0x30cc, 0x2538, 0x399d, 0x3919, 0x399e, 0x21cc, 0x38e9, 0x30f8, 0x3a20, 0x3b3c, 0x3990, 0x259c, 0x3143, 0x3080, 0x3967, 0x3afb, 0x3a1b, 0x3779, 0x2eeb, 0x39f3, 0x379a, 0x369c, 0x3985, 0x3a1b, 0x3ba6, 0x3a53, 0x28d5, 0x3881, 0x31d9, 0x3a34, 0x3bd9, 0x393a, 
+0x3601, 0x2c6e, 0x3636, 0x3298, 0x39bb, 0x3a08, 0x38db, 0x35ad, 0x3a09, 0x36a6, 0x3bc7, 0x3bac, 0x34ae, 0x3291, 0x290b, 0x3250, 0x2648, 0x333d, 0x2bf3, 0x34b1, 0x30e0, 0x351f, 0x3a74, 0x38dc, 0x3883, 0x2841, 0x35e1, 0x390d, 0x3a50, 0x3abd, 0x386d, 0x3bb7, 
+0x3b94, 0x36b7, 0x3a49, 0x332f, 0x3a1d, 0x354b, 0x3bab, 0x3346, 0x3417, 0x351e, 0x3b6d, 0x391a, 0x2db3, 0x3b1c, 0x3a4a, 0x37b7, 0x36cf, 0x3a56, 0x39c4, 0x3be9, 0x34f0, 0x39be, 0x3691, 0x1ba5, 0x3888, 0x3040, 0x3ae1, 0x3b9b, 0x398f, 0x3a49, 0x3a16, 0x38c0, 
+0x386c, 0x39ab, 0x37fa, 0x382c, 0x3a6f, 0x393f, 0x340d, 0x38ef, 0x39d1, 0x3845, 0x398f, 0x363e, 0x3687, 0x3052, 0x3a2b, 0x392c, 0x2f5c, 0x3412, 0x3a1f, 0x3b2f, 0x3bcc, 0x3a63, 0x3a89, 0x36e9, 0x3921, 0x3b80, 0x2dc0, 0x3a03, 0x3beb, 0x38d3, 0x36cb, 0x39a3, 
+0x3978, 0x3a88, 0x3ba4, 0x3561, 0x28c5, 0x33a0, 0x37be, 0x2c39, 0x30ee, 0x3782, 0x2c07, 0x354e, 0x3491, 0x3a92, 0x331a, 0x3b15, 0x32e1, 0x3839, 0x3afb, 0x36c2, 0x2fd0, 0x29ad, 0x3b2e, 0x39c1, 0x2a8c, 0x341a, 0x2f90, 0x395a, 0x3969, 0x37ea, 0x3a5c, 0x3b6d, 
+0x3971, 0x3a93, 0x304e, 0x3623, 0x3a22, 0x31ee, 0x29df, 0x2c93, 0x3a01, 0x3a62, 0x366c, 0x371d, 0x3af3, 0x2e08, 0x3ac0, 0x3642, 0x3a28, 0x368d, 0x2d3d, 0x36d9, 0x32c3, 0x373f, 0x36fe, 0x3487, 0x2c81, 0x3623, 0x3b59, 0x3a91, 0x350a, 0x34f4, 0x3b09, 0x2c25, 
+0x3b13, 0x325a, 0x379e, 0x3a7d, 0x34b1, 0x39d5, 0x2ba8, 0x322b, 0x3b5e, 0x37ab, 0x2e24, 0x3ba9, 0x3a3d, 0x34f7, 0x3ba1, 0x3877, 0x3071, 0x39fb, 0x3bbd, 0x3633, 0x3b36, 0x2daa, 0x3b9b, 0x3aa0, 0x395c, 0x3b8f, 0x38d5, 0x3ab0, 0x3a8f, 0x36c2, 0x3b1f, 0x3489, 
+0x2acc, 0x3845, 0x3715, 0x37d8, 0x3992, 0x3bff, 0x350e, 0x3ad7, 0x39b0, 0x35ac, 0x3287, 0x385f, 0x3bd4, 0x37a3, 0x3438, 0x39a5, 0x3bcf, 0x38c3, 0x34f6, 0x3ae3, 0x3b57, 0x39af, 0x35eb, 0x3bed, 0x34d4, 0x2a95, 0x3b13, 0x384e, 0x3a3b, 0x33da, 0x3bce, 0x3b99, 
+0x3559, 0x3335, 0x3a2e, 0x3123, 0x38db, 0x33d0, 0x3638, 0x3b17, 0x3a72, 0x3afc, 0x3936, 0x3838, 0x2b69, 0x3895, 0x3a1a, 0x3192, 0x39d5, 0x37a5, 0x2eb0, 0x2e8b, 0x329a, 0x3b90, 0x390a, 0x3a1e, 0x3847, 0x375d, 0x3873, 0x35e2, 0x3771, 0x30f5, 0x3231, 0x3bd7, 
+0x2bbc, 0x3ace, 0x31ad, 0x3a6b, 0x28a4, 0x3b48, 0x3ba3, 0x3a84, 0x3353, 0x39f6, 0x381f, 0x2dd6, 0x314c, 0x34af, 0x3929, 0x3921, 0x383b, 0x34b0, 0x3923, 0x32c9, 0x3ae7, 0x318f, 0x3480, 0x2ad8, 0x3042, 0x3a4c, 0x349d, 0x2c12, 0x3abb, 0x3a57, 0x3b0d, 0x3111, 
+0x3359, 0x3a84, 0x38f2, 0x368d, 0x2f4b, 0x3ba0, 0x395c, 0x3026, 0x3a15, 0x2a04, 0x326e, 0x3522, 0x31a2, 0x382f, 0x2ada, 0x3b7c, 0x2f80, 0x3af5, 0x2d35, 0x38fa, 0x39ab, 0x2c6d, 0x2e7a, 0x39f6, 0x31a4, 0x3a53, 0x358c, 0x3951, 0x3a4e, 0x3916, 0x2a3f, 0x3ae9, 
+0x3b03, 0x39f8, 0x39fe, 0x3a61, 0x39fb, 0x3704, 0x360d, 0x39a7, 0x37a9, 0x348f, 0x3a30, 0x3af5, 0x366f, 0x3b29, 0x3a6a, 0x33d5, 0x370a, 0x39cd, 0x3444, 0x3bea, 0x3b2b, 0x312e, 0x3b8e, 0x32cf, 0x3b79, 0x3302, 0x3bba, 0x3962, 0x3413, 0x37a1, 0x39e0, 0x3805 
+};
\ No newline at end of file
diff --git a/hwpe/redmule/inc/y_input.h b/hwpe/redmule/inc/y_input.h
new file mode 100644
index 0000000..45a2375
--- /dev/null
+++ b/hwpe/redmule/inc/y_input.h
@@ -0,0 +1,27 @@
+ /* Header file generated by RedMulE Golden Model */
+uint16_t y_inp [768] = {
+0x3150, 0x2dc1, 0x3033, 0x31f5, 0x3bb6, 0x3bff, 0x39f9, 0x3662, 0x3720, 0x351d, 0x384b, 0x3093, 0x3b9d, 0x35ad, 0x3695, 0x3466, 0x2300, 0x3445, 0x33ae, 0x3586, 0x38a3, 0x3bdb, 0x33a2, 0x379b, 0x3a0e, 0x38b0, 0x39ba, 0x379b, 0x39d3, 0x3a51, 0x3b30, 0x3794, 
+0x3b76, 0x3042, 0x38cc, 0x2dfc, 0x3b1a, 0x37fb, 0x38f7, 0x3824, 0x386f, 0x38c7, 0x36ee, 0x3a9c, 0x38d3, 0x2c67, 0x3a80, 0x2f30, 0x3328, 0x3721, 0x3790, 0x34e5, 0x3a6c, 0x3643, 0x3934, 0x3034, 0x38d4, 0x362e, 0x3b4b, 0x3408, 0x30c2, 0x370e, 0x3b31, 0x3b16, 
+0x3b6b, 0x39d4, 0x339c, 0x381e, 0x313e, 0x3671, 0x3ae2, 0x3479, 0x3940, 0x342d, 0x3925, 0x370a, 0x35d8, 0x2dad, 0x3888, 0x24b9, 0x375d, 0x34bd, 0x3243, 0x2ebb, 0x3970, 0x3a21, 0x3a07, 0x3877, 0x3888, 0x3569, 0x372d, 0x2ac1, 0x331e, 0x384d, 0x3996, 0x34a4, 
+0x35c1, 0x33a9, 0x21ed, 0x3a42, 0x388d, 0x34e4, 0x33c3, 0x34f9, 0x3a7b, 0x33fb, 0x2cdd, 0x3b0e, 0x333b, 0x3973, 0x34fc, 0x3771, 0x32ea, 0x2de4, 0x31a8, 0x3946, 0x3657, 0x3a4e, 0x36f6, 0x2829, 0x3ba2, 0x3bdc, 0x3bb3, 0x306c, 0x398d, 0x3a1f, 0x3991, 0x3846, 
+0x3547, 0x3292, 0x2e85, 0x31ed, 0x3979, 0x3a90, 0x28a4, 0x3bed, 0x36d8, 0x340e, 0x3b6a, 0x3ab6, 0x3824, 0x382b, 0x3ac3, 0x3811, 0x36d7, 0x3519, 0x3a92, 0x3a42, 0x29d1, 0x383a, 0x3a9b, 0x300e, 0x2cd3, 0x39cd, 0x3874, 0x3a07, 0x2eb1, 0x3b86, 0x3ad8, 0x3a5d, 
+0x3712, 0x284a, 0x38c1, 0x3bec, 0x39c0, 0x32cd, 0x3ad8, 0x3bce, 0x3817, 0x3896, 0x3aa7, 0x3870, 0x3996, 0x32cc, 0x3a4c, 0x3757, 0x3814, 0x3b65, 0x3acb, 0x376e, 0x34c0, 0x3609, 0x3bf0, 0x3b24, 0x3b29, 0x3848, 0x34b7, 0x398a, 0x220c, 0x3498, 0x3a8c, 0x3883, 
+0x38c4, 0x3af6, 0x3a42, 0x2dd6, 0x3147, 0x3717, 0x3a8e, 0x3af9, 0x3296, 0x38ef, 0x34fa, 0x3555, 0x3b29, 0x38de, 0x315e, 0x3773, 0x3b67, 0x3116, 0x38ec, 0x357c, 0x35d0, 0x2518, 0x3958, 0x2a03, 0x37d9, 0x3699, 0x3a1e, 0x3230, 0x3b13, 0x36d4, 0x3b2a, 0x39ad, 
+0x3b10, 0x351a, 0x3b97, 0x3326, 0x2b54, 0x3b7d, 0x386f, 0x373e, 0x37fa, 0x389b, 0x3b90, 0x3292, 0x3975, 0x38f3, 0x37f1, 0x3590, 0x3810, 0x2fd7, 0x3bf7, 0x3a5a, 0x3a1c, 0x34dd, 0x354c, 0x32f8, 0x3095, 0x321e, 0x39e0, 0x395c, 0x3717, 0x357f, 0x394a, 0x34b1, 
+0x3ba4, 0x380c, 0x3604, 0x2f50, 0x348d, 0x3828, 0x3a9f, 0x39ce, 0x32ca, 0x3906, 0x3ab2, 0x2ca5, 0x38c9, 0x362a, 0x34b2, 0x29dc, 0x3a36, 0x3052, 0x31b7, 0x3589, 0x387c, 0x3401, 0x3b22, 0x3ad6, 0x3ae8, 0x3238, 0x3494, 0x3502, 0x3717, 0x3a6c, 0x3229, 0x368c, 
+0x3056, 0x3a56, 0x3498, 0x39eb, 0x2864, 0x342d, 0x39e0, 0x34a1, 0x2b99, 0x3a04, 0x38ff, 0x328c, 0x34d9, 0x387d, 0x3a3c, 0x32e5, 0x39eb, 0x3984, 0x34dd, 0x38a7, 0x373f, 0x39b4, 0x3235, 0x2f58, 0x2f39, 0x3800, 0x3758, 0x3939, 0x39fc, 0x3a4b, 0x38bf, 0x30ee, 
+0x345e, 0x39c8, 0x3a6d, 0x3262, 0x3b81, 0x31dc, 0x3a15, 0x3bd0, 0x36af, 0x36de, 0x37d5, 0x39d7, 0x3ad3, 0x3ac1, 0x3109, 0x35ea, 0x31c6, 0x398d, 0x3987, 0x3a4a, 0x34d2, 0x2ed2, 0x35e6, 0x352c, 0x39eb, 0x3bd6, 0x3a5b, 0x39d1, 0x34aa, 0x3ade, 0x394b, 0x38a1, 
+0x2bed, 0x38de, 0x3811, 0x3813, 0x391a, 0x374b, 0x3829, 0x3725, 0x38f0, 0x3583, 0x3966, 0x3a7d, 0x375a, 0x38fe, 0x3696, 0x361c, 0x39a8, 0x35f0, 0x38e1, 0x3003, 0x3595, 0x316e, 0x3862, 0x3af8, 0x3af2, 0x34c8, 0x381d, 0x37d8, 0x3893, 0x3a9c, 0x3989, 0x308c, 
+0x30cc, 0x2538, 0x399d, 0x3919, 0x399e, 0x21cc, 0x38e9, 0x30f8, 0x3a20, 0x3b3c, 0x3990, 0x259c, 0x3143, 0x3080, 0x3967, 0x3afb, 0x3a1b, 0x3779, 0x2eeb, 0x39f3, 0x379a, 0x369c, 0x3985, 0x3a1b, 0x3ba6, 0x3a53, 0x28d5, 0x3881, 0x31d9, 0x3a34, 0x3bd9, 0x393a, 
+0x3601, 0x2c6e, 0x3636, 0x3298, 0x39bb, 0x3a08, 0x38db, 0x35ad, 0x3a09, 0x36a6, 0x3bc7, 0x3bac, 0x34ae, 0x3291, 0x290b, 0x3250, 0x2648, 0x333d, 0x2bf3, 0x34b1, 0x30e0, 0x351f, 0x3a74, 0x38dc, 0x3883, 0x2841, 0x35e1, 0x390d, 0x3a50, 0x3abd, 0x386d, 0x3bb7, 
+0x3b94, 0x36b7, 0x3a49, 0x332f, 0x3a1d, 0x354b, 0x3bab, 0x3346, 0x3417, 0x351e, 0x3b6d, 0x391a, 0x2db3, 0x3b1c, 0x3a4a, 0x37b7, 0x36cf, 0x3a56, 0x39c4, 0x3be9, 0x34f0, 0x39be, 0x3691, 0x1ba5, 0x3888, 0x3040, 0x3ae1, 0x3b9b, 0x398f, 0x3a49, 0x3a16, 0x38c0, 
+0x386c, 0x39ab, 0x37fa, 0x382c, 0x3a6f, 0x393f, 0x340d, 0x38ef, 0x39d1, 0x3845, 0x398f, 0x363e, 0x3687, 0x3052, 0x3a2b, 0x392c, 0x2f5c, 0x3412, 0x3a1f, 0x3b2f, 0x3bcc, 0x3a63, 0x3a89, 0x36e9, 0x3921, 0x3b80, 0x2dc0, 0x3a03, 0x3beb, 0x38d3, 0x36cb, 0x39a3, 
+0x3978, 0x3a88, 0x3ba4, 0x3561, 0x28c5, 0x33a0, 0x37be, 0x2c39, 0x30ee, 0x3782, 0x2c07, 0x354e, 0x3491, 0x3a92, 0x331a, 0x3b15, 0x32e1, 0x3839, 0x3afb, 0x36c2, 0x2fd0, 0x29ad, 0x3b2e, 0x39c1, 0x2a8c, 0x341a, 0x2f90, 0x395a, 0x3969, 0x37ea, 0x3a5c, 0x3b6d, 
+0x3971, 0x3a93, 0x304e, 0x3623, 0x3a22, 0x31ee, 0x29df, 0x2c93, 0x3a01, 0x3a62, 0x366c, 0x371d, 0x3af3, 0x2e08, 0x3ac0, 0x3642, 0x3a28, 0x368d, 0x2d3d, 0x36d9, 0x32c3, 0x373f, 0x36fe, 0x3487, 0x2c81, 0x3623, 0x3b59, 0x3a91, 0x350a, 0x34f4, 0x3b09, 0x2c25, 
+0x3b13, 0x325a, 0x379e, 0x3a7d, 0x34b1, 0x39d5, 0x2ba8, 0x322b, 0x3b5e, 0x37ab, 0x2e24, 0x3ba9, 0x3a3d, 0x34f7, 0x3ba1, 0x3877, 0x3071, 0x39fb, 0x3bbd, 0x3633, 0x3b36, 0x2daa, 0x3b9b, 0x3aa0, 0x395c, 0x3b8f, 0x38d5, 0x3ab0, 0x3a8f, 0x36c2, 0x3b1f, 0x3489, 
+0x2acc, 0x3845, 0x3715, 0x37d8, 0x3992, 0x3bff, 0x350e, 0x3ad7, 0x39b0, 0x35ac, 0x3287, 0x385f, 0x3bd4, 0x37a3, 0x3438, 0x39a5, 0x3bcf, 0x38c3, 0x34f6, 0x3ae3, 0x3b57, 0x39af, 0x35eb, 0x3bed, 0x34d4, 0x2a95, 0x3b13, 0x384e, 0x3a3b, 0x33da, 0x3bce, 0x3b99, 
+0x3559, 0x3335, 0x3a2e, 0x3123, 0x38db, 0x33d0, 0x3638, 0x3b17, 0x3a72, 0x3afc, 0x3936, 0x3838, 0x2b69, 0x3895, 0x3a1a, 0x3192, 0x39d5, 0x37a5, 0x2eb0, 0x2e8b, 0x329a, 0x3b90, 0x390a, 0x3a1e, 0x3847, 0x375d, 0x3873, 0x35e2, 0x3771, 0x30f5, 0x3231, 0x3bd7, 
+0x2bbc, 0x3ace, 0x31ad, 0x3a6b, 0x28a4, 0x3b48, 0x3ba3, 0x3a84, 0x3353, 0x39f6, 0x381f, 0x2dd6, 0x314c, 0x34af, 0x3929, 0x3921, 0x383b, 0x34b0, 0x3923, 0x32c9, 0x3ae7, 0x318f, 0x3480, 0x2ad8, 0x3042, 0x3a4c, 0x349d, 0x2c12, 0x3abb, 0x3a57, 0x3b0d, 0x3111, 
+0x3359, 0x3a84, 0x38f2, 0x368d, 0x2f4b, 0x3ba0, 0x395c, 0x3026, 0x3a15, 0x2a04, 0x326e, 0x3522, 0x31a2, 0x382f, 0x2ada, 0x3b7c, 0x2f80, 0x3af5, 0x2d35, 0x38fa, 0x39ab, 0x2c6d, 0x2e7a, 0x39f6, 0x31a4, 0x3a53, 0x358c, 0x3951, 0x3a4e, 0x3916, 0x2a3f, 0x3ae9, 
+0x3b03, 0x39f8, 0x39fe, 0x3a61, 0x39fb, 0x3704, 0x360d, 0x39a7, 0x37a9, 0x348f, 0x3a30, 0x3af5, 0x366f, 0x3b29, 0x3a6a, 0x33d5, 0x370a, 0x39cd, 0x3444, 0x3bea, 0x3b2b, 0x312e, 0x3b8e, 0x32cf, 0x3b79, 0x3302, 0x3bba, 0x3962, 0x3413, 0x37a1, 0x39e0, 0x3805 
+};
\ No newline at end of file
diff --git a/hwpe/redmule/inc/z_2D.h b/hwpe/redmule/inc/z_2D.h
new file mode 100644
index 0000000..aff808a
--- /dev/null
+++ b/hwpe/redmule/inc/z_2D.h
@@ -0,0 +1,27 @@
+ /* Header file generated by RedMulE Golden Model */
+uint16_t z_oup_2D [24][32] = {
+0x4845, 0x4897, 0x4608, 0x4838, 0x4855, 0x487b, 0x4869, 0x4880, 0x46d1, 0x48b0, 0x48db, 0x483f, 0x48c9, 0x485f, 0x4881, 0x483a, 0x484b, 0x472c, 0x4762, 0x492b, 0x4822, 0x48fd, 0x488e, 0x492e, 0x483e, 0x484f, 0x49e8, 0x46d7, 0x484b, 0x489d, 0x490b, 0x47e9, 
+0x484f, 0x47d2, 0x44be, 0x4747, 0x47c7, 0x46c0, 0x4727, 0x48af, 0x46c5, 0x482d, 0x483d, 0x482e, 0x4897, 0x479f, 0x488b, 0x4749, 0x489a, 0x46a8, 0x46f2, 0x488b, 0x4891, 0x47e8, 0x4872, 0x483d, 0x4716, 0x46fd, 0x49b5, 0x46a0, 0x46e7, 0x47a4, 0x48a1, 0x4767, 
+0x4939, 0x4935, 0x4703, 0x48c1, 0x4863, 0x48bd, 0x4913, 0x48cf, 0x48b6, 0x48b8, 0x4946, 0x4920, 0x495e, 0x48e1, 0x4938, 0x48b2, 0x493a, 0x4882, 0x483b, 0x49d5, 0x4911, 0x4972, 0x496b, 0x49df, 0x48f2, 0x4888, 0x4a46, 0x4821, 0x48c1, 0x490c, 0x49b2, 0x48a3, 
+0x463a, 0x47b0, 0x44cb, 0x4762, 0x4765, 0x46b9, 0x466a, 0x4814, 0x4631, 0x4796, 0x4666, 0x474b, 0x4798, 0x4704, 0x4838, 0x4761, 0x47d3, 0x4590, 0x45ea, 0x48a2, 0x47f1, 0x4844, 0x484b, 0x4776, 0x47d6, 0x46d8, 0x48f3, 0x44d3, 0x46fa, 0x478d, 0x481e, 0x466e, 
+0x4827, 0x481e, 0x45a2, 0x4794, 0x4727, 0x4806, 0x475d, 0x48d5, 0x4708, 0x4828, 0x4862, 0x480d, 0x4895, 0x4832, 0x48bd, 0x47f1, 0x482a, 0x46a7, 0x47b1, 0x492d, 0x484d, 0x4884, 0x48dc, 0x485f, 0x476d, 0x480c, 0x48e9, 0x46d3, 0x4728, 0x4884, 0x48a0, 0x480e, 
+0x4862, 0x4813, 0x4675, 0x485a, 0x47e8, 0x4738, 0x4836, 0x4823, 0x46e7, 0x4821, 0x4822, 0x47b3, 0x4846, 0x4855, 0x4863, 0x4717, 0x4872, 0x47c1, 0x46d5, 0x488e, 0x47e2, 0x485f, 0x487c, 0x48b8, 0x481e, 0x4788, 0x48bd, 0x4677, 0x46c9, 0x47f8, 0x48fe, 0x47fc, 
+0x47a0, 0x47b2, 0x4588, 0x467e, 0x4662, 0x46c7, 0x46e8, 0x4812, 0x4536, 0x474e, 0x46c0, 0x468f, 0x481f, 0x4679, 0x46a1, 0x46e2, 0x4809, 0x4560, 0x4630, 0x47eb, 0x46b5, 0x4757, 0x4848, 0x477f, 0x46a6, 0x46d8, 0x4870, 0x459a, 0x4670, 0x4678, 0x47d2, 0x468c, 
+0x4762, 0x48c4, 0x46e3, 0x4791, 0x46b1, 0x486d, 0x47d0, 0x4867, 0x468d, 0x47f6, 0x48a5, 0x4756, 0x4857, 0x4854, 0x4866, 0x4838, 0x484d, 0x46ec, 0x47d2, 0x48f6, 0x484a, 0x4879, 0x4848, 0x483c, 0x471d, 0x4806, 0x48fa, 0x4730, 0x4768, 0x47b8, 0x4865, 0x46f9, 
+0x48a8, 0x4918, 0x46ca, 0x4867, 0x4800, 0x4862, 0x48d3, 0x4910, 0x474e, 0x4849, 0x48eb, 0x486b, 0x4966, 0x48c5, 0x48f4, 0x4830, 0x48f9, 0x4778, 0x481e, 0x499e, 0x48cf, 0x48f1, 0x4982, 0x4923, 0x487c, 0x47cf, 0x49ea, 0x4649, 0x4773, 0x495e, 0x48b2, 0x483f, 
+0x48a7, 0x4975, 0x4616, 0x481e, 0x481f, 0x4866, 0x48b6, 0x4864, 0x47dc, 0x4873, 0x485c, 0x487f, 0x4938, 0x491f, 0x490d, 0x48b6, 0x48f8, 0x48a1, 0x4859, 0x492d, 0x489c, 0x4915, 0x4899, 0x4887, 0x486c, 0x4859, 0x49ca, 0x471e, 0x4867, 0x4918, 0x48d3, 0x4827, 
+0x488b, 0x4998, 0x4704, 0x481d, 0x48b8, 0x4880, 0x4876, 0x4944, 0x470c, 0x48f2, 0x48b9, 0x489b, 0x4956, 0x48e5, 0x48d6, 0x48a5, 0x48dc, 0x4856, 0x484e, 0x49ab, 0x48e0, 0x490e, 0x48dd, 0x4945, 0x488b, 0x48dd, 0x4a32, 0x47ea, 0x4835, 0x4911, 0x4965, 0x4819, 
+0x460e, 0x481e, 0x452c, 0x4673, 0x475c, 0x4717, 0x46f6, 0x46d0, 0x4696, 0x46bc, 0x4726, 0x481e, 0x4763, 0x46ea, 0x46fe, 0x4758, 0x478b, 0x4627, 0x4704, 0x483f, 0x46ad, 0x47b1, 0x4792, 0x4816, 0x46f2, 0x4684, 0x4827, 0x45a8, 0x472f, 0x47a4, 0x4797, 0x462b, 
+0x483f, 0x48ab, 0x468f, 0x4863, 0x485a, 0x4766, 0x481d, 0x48cb, 0x47dc, 0x4903, 0x48fc, 0x4830, 0x48cc, 0x483e, 0x48ab, 0x4864, 0x4966, 0x4763, 0x4794, 0x499d, 0x488e, 0x488b, 0x48dc, 0x4960, 0x4854, 0x484c, 0x499c, 0x474c, 0x4826, 0x48bc, 0x4949, 0x4883, 
+0x489d, 0x4905, 0x4718, 0x481e, 0x48e3, 0x48f4, 0x48c1, 0x4904, 0x47e8, 0x48b3, 0x4892, 0x48d4, 0x48ff, 0x4894, 0x48d5, 0x4886, 0x48fa, 0x4803, 0x47d2, 0x492e, 0x4870, 0x48b2, 0x48e5, 0x492b, 0x487b, 0x4785, 0x49e3, 0x471d, 0x4837, 0x48bf, 0x489b, 0x48c4, 
+0x475c, 0x4871, 0x464a, 0x4811, 0x47af, 0x471c, 0x4817, 0x4817, 0x463b, 0x484e, 0x477f, 0x464f, 0x4704, 0x487c, 0x47a3, 0x4725, 0x4853, 0x462a, 0x465a, 0x4860, 0x4736, 0x4880, 0x47e1, 0x482b, 0x4811, 0x46c0, 0x48dc, 0x475d, 0x4668, 0x4806, 0x4893, 0x46f4, 
+0x4858, 0x4959, 0x463d, 0x487b, 0x480f, 0x484e, 0x48c0, 0x48a6, 0x4847, 0x4894, 0x48a0, 0x484a, 0x491e, 0x48f4, 0x48fc, 0x48b5, 0x48ce, 0x47d2, 0x47db, 0x497f, 0x4955, 0x4939, 0x48a7, 0x48ce, 0x4890, 0x4884, 0x49d6, 0x4763, 0x486e, 0x4922, 0x48f4, 0x48c3, 
+0x47ec, 0x491c, 0x4698, 0x4783, 0x4715, 0x4754, 0x4745, 0x4752, 0x472f, 0x4832, 0x4817, 0x4809, 0x47f8, 0x48c3, 0x47e6, 0x4800, 0x48b6, 0x4730, 0x480a, 0x48cb, 0x479e, 0x488e, 0x47c2, 0x488e, 0x472f, 0x47ee, 0x489d, 0x4744, 0x4755, 0x4851, 0x4846, 0x47d3, 
+0x4838, 0x48a0, 0x4634, 0x4762, 0x4786, 0x4806, 0x47e3, 0x482d, 0x4726, 0x486c, 0x47b7, 0x4803, 0x48ac, 0x4814, 0x48e0, 0x4839, 0x4827, 0x4750, 0x46f2, 0x48c5, 0x483f, 0x4886, 0x48ad, 0x4856, 0x47e8, 0x47a9, 0x4937, 0x4743, 0x46d0, 0x481f, 0x484c, 0x4804, 
+0x47fd, 0x481f, 0x456d, 0x4813, 0x474d, 0x4807, 0x4688, 0x480e, 0x46e8, 0x4810, 0x469f, 0x4799, 0x4853, 0x478f, 0x47f2, 0x4824, 0x47d0, 0x471f, 0x46da, 0x485f, 0x4813, 0x481c, 0x482e, 0x4863, 0x4786, 0x480b, 0x48c9, 0x46b8, 0x475a, 0x46e2, 0x4852, 0x46c5, 
+0x45af, 0x4802, 0x4466, 0x46c2, 0x465d, 0x4743, 0x46b7, 0x47ba, 0x4636, 0x46c3, 0x4677, 0x4784, 0x485a, 0x47c2, 0x46dc, 0x46ac, 0x47de, 0x460e, 0x465f, 0x4834, 0x47f4, 0x4769, 0x46fc, 0x4810, 0x45fd, 0x45ea, 0x48d0, 0x45b5, 0x4704, 0x4783, 0x4830, 0x46c4, 
+0x4759, 0x47c7, 0x453d, 0x45b0, 0x4741, 0x4702, 0x4736, 0x4793, 0x461b, 0x47ba, 0x470b, 0x46dd, 0x4657, 0x470b, 0x470d, 0x4710, 0x486c, 0x468f, 0x45c3, 0x46ba, 0x479d, 0x483b, 0x46c9, 0x4774, 0x46a9, 0x46a7, 0x4833, 0x4606, 0x4690, 0x46a9, 0x46f5, 0x46a7, 
+0x47ac, 0x48bb, 0x452c, 0x4803, 0x470f, 0x4824, 0x47d5, 0x48cb, 0x4707, 0x484a, 0x4832, 0x4797, 0x4851, 0x482c, 0x487a, 0x4877, 0x4891, 0x465d, 0x47f4, 0x48ce, 0x4898, 0x4899, 0x484e, 0x486a, 0x47ac, 0x47f0, 0x493e, 0x4611, 0x47e2, 0x489e, 0x488c, 0x46af, 
+0x4665, 0x4836, 0x45e4, 0x46b6, 0x46a1, 0x46b9, 0x46c8, 0x46dd, 0x4658, 0x474b, 0x467b, 0x4777, 0x4769, 0x4798, 0x4785, 0x475e, 0x472a, 0x4656, 0x45fb, 0x4881, 0x46fc, 0x472d, 0x476e, 0x47a3, 0x465d, 0x46ca, 0x4855, 0x4500, 0x464f, 0x479a, 0x46c3, 0x4738, 
+0x481e, 0x486c, 0x4659, 0x4801, 0x4756, 0x477a, 0x47d5, 0x487b, 0x4706, 0x4808, 0x484f, 0x4838, 0x4870, 0x4863, 0x48d3, 0x4806, 0x4865, 0x4771, 0x46be, 0x494c, 0x4915, 0x484c, 0x4900, 0x4862, 0x481a, 0x46e8, 0x4974, 0x46a0, 0x4775, 0x483d, 0x487c, 0x480e 
+};
\ No newline at end of file
diff --git a/hwpe/redmule/inc/z_output.h b/hwpe/redmule/inc/z_output.h
new file mode 100644
index 0000000..96c7e5f
--- /dev/null
+++ b/hwpe/redmule/inc/z_output.h
@@ -0,0 +1,27 @@
+ /* Header file generated by RedMulE Golden Model */
+uint16_t z_oup [768] = {
+0x4845, 0x4897, 0x4608, 0x4838, 0x4855, 0x487b, 0x4869, 0x4880, 0x46d1, 0x48b0, 0x48db, 0x483f, 0x48c9, 0x485f, 0x4881, 0x483a, 0x484b, 0x472c, 0x4762, 0x492b, 0x4822, 0x48fd, 0x488e, 0x492e, 0x483e, 0x484f, 0x49e8, 0x46d7, 0x484b, 0x489d, 0x490b, 0x47e9, 
+0x484f, 0x47d2, 0x44be, 0x4747, 0x47c7, 0x46c0, 0x4727, 0x48af, 0x46c5, 0x482d, 0x483d, 0x482e, 0x4897, 0x479f, 0x488b, 0x4749, 0x489a, 0x46a8, 0x46f2, 0x488b, 0x4891, 0x47e8, 0x4872, 0x483d, 0x4716, 0x46fd, 0x49b5, 0x46a0, 0x46e7, 0x47a4, 0x48a1, 0x4767, 
+0x4939, 0x4935, 0x4703, 0x48c1, 0x4863, 0x48bd, 0x4913, 0x48cf, 0x48b6, 0x48b8, 0x4946, 0x4920, 0x495e, 0x48e1, 0x4938, 0x48b2, 0x493a, 0x4882, 0x483b, 0x49d5, 0x4911, 0x4972, 0x496b, 0x49df, 0x48f2, 0x4888, 0x4a46, 0x4821, 0x48c1, 0x490c, 0x49b2, 0x48a3, 
+0x463a, 0x47b0, 0x44cb, 0x4762, 0x4765, 0x46b9, 0x466a, 0x4814, 0x4631, 0x4796, 0x4666, 0x474b, 0x4798, 0x4704, 0x4838, 0x4761, 0x47d3, 0x4590, 0x45ea, 0x48a2, 0x47f1, 0x4844, 0x484b, 0x4776, 0x47d6, 0x46d8, 0x48f3, 0x44d3, 0x46fa, 0x478d, 0x481e, 0x466e, 
+0x4827, 0x481e, 0x45a2, 0x4794, 0x4727, 0x4806, 0x475d, 0x48d5, 0x4708, 0x4828, 0x4862, 0x480d, 0x4895, 0x4832, 0x48bd, 0x47f1, 0x482a, 0x46a7, 0x47b1, 0x492d, 0x484d, 0x4884, 0x48dc, 0x485f, 0x476d, 0x480c, 0x48e9, 0x46d3, 0x4728, 0x4884, 0x48a0, 0x480e, 
+0x4862, 0x4813, 0x4675, 0x485a, 0x47e8, 0x4738, 0x4836, 0x4823, 0x46e7, 0x4821, 0x4822, 0x47b3, 0x4846, 0x4855, 0x4863, 0x4717, 0x4872, 0x47c1, 0x46d5, 0x488e, 0x47e2, 0x485f, 0x487c, 0x48b8, 0x481e, 0x4788, 0x48bd, 0x4677, 0x46c9, 0x47f8, 0x48fe, 0x47fc, 
+0x47a0, 0x47b2, 0x4588, 0x467e, 0x4662, 0x46c7, 0x46e8, 0x4812, 0x4536, 0x474e, 0x46c0, 0x468f, 0x481f, 0x4679, 0x46a1, 0x46e2, 0x4809, 0x4560, 0x4630, 0x47eb, 0x46b5, 0x4757, 0x4848, 0x477f, 0x46a6, 0x46d8, 0x4870, 0x459a, 0x4670, 0x4678, 0x47d2, 0x468c, 
+0x4762, 0x48c4, 0x46e3, 0x4791, 0x46b1, 0x486d, 0x47d0, 0x4867, 0x468d, 0x47f6, 0x48a5, 0x4756, 0x4857, 0x4854, 0x4866, 0x4838, 0x484d, 0x46ec, 0x47d2, 0x48f6, 0x484a, 0x4879, 0x4848, 0x483c, 0x471d, 0x4806, 0x48fa, 0x4730, 0x4768, 0x47b8, 0x4865, 0x46f9, 
+0x48a8, 0x4918, 0x46ca, 0x4867, 0x4800, 0x4862, 0x48d3, 0x4910, 0x474e, 0x4849, 0x48eb, 0x486b, 0x4966, 0x48c5, 0x48f4, 0x4830, 0x48f9, 0x4778, 0x481e, 0x499e, 0x48cf, 0x48f1, 0x4982, 0x4923, 0x487c, 0x47cf, 0x49ea, 0x4649, 0x4773, 0x495e, 0x48b2, 0x483f, 
+0x48a7, 0x4975, 0x4616, 0x481e, 0x481f, 0x4866, 0x48b6, 0x4864, 0x47dc, 0x4873, 0x485c, 0x487f, 0x4938, 0x491f, 0x490d, 0x48b6, 0x48f8, 0x48a1, 0x4859, 0x492d, 0x489c, 0x4915, 0x4899, 0x4887, 0x486c, 0x4859, 0x49ca, 0x471e, 0x4867, 0x4918, 0x48d3, 0x4827, 
+0x488b, 0x4998, 0x4704, 0x481d, 0x48b8, 0x4880, 0x4876, 0x4944, 0x470c, 0x48f2, 0x48b9, 0x489b, 0x4956, 0x48e5, 0x48d6, 0x48a5, 0x48dc, 0x4856, 0x484e, 0x49ab, 0x48e0, 0x490e, 0x48dd, 0x4945, 0x488b, 0x48dd, 0x4a32, 0x47ea, 0x4835, 0x4911, 0x4965, 0x4819, 
+0x460e, 0x481e, 0x452c, 0x4673, 0x475c, 0x4717, 0x46f6, 0x46d0, 0x4696, 0x46bc, 0x4726, 0x481e, 0x4763, 0x46ea, 0x46fe, 0x4758, 0x478b, 0x4627, 0x4704, 0x483f, 0x46ad, 0x47b1, 0x4792, 0x4816, 0x46f2, 0x4684, 0x4827, 0x45a8, 0x472f, 0x47a4, 0x4797, 0x462b, 
+0x483f, 0x48ab, 0x468f, 0x4863, 0x485a, 0x4766, 0x481d, 0x48cb, 0x47dc, 0x4903, 0x48fc, 0x4830, 0x48cc, 0x483e, 0x48ab, 0x4864, 0x4966, 0x4763, 0x4794, 0x499d, 0x488e, 0x488b, 0x48dc, 0x4960, 0x4854, 0x484c, 0x499c, 0x474c, 0x4826, 0x48bc, 0x4949, 0x4883, 
+0x489d, 0x4905, 0x4718, 0x481e, 0x48e3, 0x48f4, 0x48c1, 0x4904, 0x47e8, 0x48b3, 0x4892, 0x48d4, 0x48ff, 0x4894, 0x48d5, 0x4886, 0x48fa, 0x4803, 0x47d2, 0x492e, 0x4870, 0x48b2, 0x48e5, 0x492b, 0x487b, 0x4785, 0x49e3, 0x471d, 0x4837, 0x48bf, 0x489b, 0x48c4, 
+0x475c, 0x4871, 0x464a, 0x4811, 0x47af, 0x471c, 0x4817, 0x4817, 0x463b, 0x484e, 0x477f, 0x464f, 0x4704, 0x487c, 0x47a3, 0x4725, 0x4853, 0x462a, 0x465a, 0x4860, 0x4736, 0x4880, 0x47e1, 0x482b, 0x4811, 0x46c0, 0x48dc, 0x475d, 0x4668, 0x4806, 0x4893, 0x46f4, 
+0x4858, 0x4959, 0x463d, 0x487b, 0x480f, 0x484e, 0x48c0, 0x48a6, 0x4847, 0x4894, 0x48a0, 0x484a, 0x491e, 0x48f4, 0x48fc, 0x48b5, 0x48ce, 0x47d2, 0x47db, 0x497f, 0x4955, 0x4939, 0x48a7, 0x48ce, 0x4890, 0x4884, 0x49d6, 0x4763, 0x486e, 0x4922, 0x48f4, 0x48c3, 
+0x47ec, 0x491c, 0x4698, 0x4783, 0x4715, 0x4754, 0x4745, 0x4752, 0x472f, 0x4832, 0x4817, 0x4809, 0x47f8, 0x48c3, 0x47e6, 0x4800, 0x48b6, 0x4730, 0x480a, 0x48cb, 0x479e, 0x488e, 0x47c2, 0x488e, 0x472f, 0x47ee, 0x489d, 0x4744, 0x4755, 0x4851, 0x4846, 0x47d3, 
+0x4838, 0x48a0, 0x4634, 0x4762, 0x4786, 0x4806, 0x47e3, 0x482d, 0x4726, 0x486c, 0x47b7, 0x4803, 0x48ac, 0x4814, 0x48e0, 0x4839, 0x4827, 0x4750, 0x46f2, 0x48c5, 0x483f, 0x4886, 0x48ad, 0x4856, 0x47e8, 0x47a9, 0x4937, 0x4743, 0x46d0, 0x481f, 0x484c, 0x4804, 
+0x47fd, 0x481f, 0x456d, 0x4813, 0x474d, 0x4807, 0x4688, 0x480e, 0x46e8, 0x4810, 0x469f, 0x4799, 0x4853, 0x478f, 0x47f2, 0x4824, 0x47d0, 0x471f, 0x46da, 0x485f, 0x4813, 0x481c, 0x482e, 0x4863, 0x4786, 0x480b, 0x48c9, 0x46b8, 0x475a, 0x46e2, 0x4852, 0x46c5, 
+0x45af, 0x4802, 0x4466, 0x46c2, 0x465d, 0x4743, 0x46b7, 0x47ba, 0x4636, 0x46c3, 0x4677, 0x4784, 0x485a, 0x47c2, 0x46dc, 0x46ac, 0x47de, 0x460e, 0x465f, 0x4834, 0x47f4, 0x4769, 0x46fc, 0x4810, 0x45fd, 0x45ea, 0x48d0, 0x45b5, 0x4704, 0x4783, 0x4830, 0x46c4, 
+0x4759, 0x47c7, 0x453d, 0x45b0, 0x4741, 0x4702, 0x4736, 0x4793, 0x461b, 0x47ba, 0x470b, 0x46dd, 0x4657, 0x470b, 0x470d, 0x4710, 0x486c, 0x468f, 0x45c3, 0x46ba, 0x479d, 0x483b, 0x46c9, 0x4774, 0x46a9, 0x46a7, 0x4833, 0x4606, 0x4690, 0x46a9, 0x46f5, 0x46a7, 
+0x47ac, 0x48bb, 0x452c, 0x4803, 0x470f, 0x4824, 0x47d5, 0x48cb, 0x4707, 0x484a, 0x4832, 0x4797, 0x4851, 0x482c, 0x487a, 0x4877, 0x4891, 0x465d, 0x47f4, 0x48ce, 0x4898, 0x4899, 0x484e, 0x486a, 0x47ac, 0x47f0, 0x493e, 0x4611, 0x47e2, 0x489e, 0x488c, 0x46af, 
+0x4665, 0x4836, 0x45e4, 0x46b6, 0x46a1, 0x46b9, 0x46c8, 0x46dd, 0x4658, 0x474b, 0x467b, 0x4777, 0x4769, 0x4798, 0x4785, 0x475e, 0x472a, 0x4656, 0x45fb, 0x4881, 0x46fc, 0x472d, 0x476e, 0x47a3, 0x465d, 0x46ca, 0x4855, 0x4500, 0x464f, 0x479a, 0x46c3, 0x4738, 
+0x481e, 0x486c, 0x4659, 0x4801, 0x4756, 0x477a, 0x47d5, 0x487b, 0x4706, 0x4808, 0x484f, 0x4838, 0x4870, 0x4863, 0x48d3, 0x4806, 0x4865, 0x4771, 0x46be, 0x494c, 0x4915, 0x484c, 0x4900, 0x4862, 0x481a, 0x46e8, 0x4974, 0x46a0, 0x4775, 0x483d, 0x487c, 0x480e 
+};
\ No newline at end of file
diff --git a/hwpe/redmule/pulp_inject_fault.tcl b/hwpe/redmule/pulp_inject_fault.tcl
new file mode 100644
index 0000000..61ccadf
--- /dev/null
+++ b/hwpe/redmule/pulp_inject_fault.tcl
@@ -0,0 +1,53 @@
+# Copyright 2023 ETH Zurich and University of Bologna.
+# Solderpad Hardware License, Version 0.51, see LICENSE for details.
+# SPDX-License-Identifier: SHL-0.51
+#
+# Author: Michael Rogenmoser (michaero@iis.ee.ethz.ch)
+
+transcript quietly
+if {! [info exists ::env(VSIM_PATH)]} {error "Define VSIM_PATH"}
+set utils_base_path  [file join $::env(VSIM_PATH) scripts fault_injection_utils]
+set script_base_path [file join $::env(VSIM_PATH) fault_injection_sim scripts]
+
+set verbosity            2
+set log_injections       1
+# Easy way to generate a variable seed
+# set seed                 [clock seconds]
+# Default value
+set seed                 12345
+set print_statistics     1
+
+set inject_start_time 550000000000ps
+set inject_stop_time 750000000000ps
+set injection_clock "pulp_cluster_tb/cluster_i/clk_i"
+set injection_clock_trigger 0
+set fault_period 150
+set rand_initial_injection_phase 0
+# max_num set to 0 means until stop_time
+set max_num_fault_inject 0
+set signal_fault_duration 20ns
+set register_fault_duration 0ns
+
+set allow_multi_bit_upset $::env(MULTI_BIT_UPSET)
+set use_bitwidth_as_weight 0
+set check_core_output_modification 0
+set check_core_next_state_modification 0
+set reg_to_sig_ratio 1
+
+source [file join $utils_base_path pulp_extract_nets.tcl]
+
+set inject_signals_netlist []
+set inject_register_netlist []
+set output_netlist []
+set next_state_netlist []
+set assertion_disable_list []
+
+# for {set idx 0} {$idx < 12} {incr idx} {
+#     set inject_signals_netlist [list {*}$inject_signals_netlist {*}[get_all_core_nets $idx]]
+#     set output_netlist [list {*}$output_netlist {*}[get_core_output_nets $idx]]
+# }
+
+set inject_register_netlist [list {*}$inject_register_netlist {*}[get_memory_slice {0 16} {256 336}]]
+
+source [file join $script_base_path inject_fault.tcl]
+
diff --git a/hwpe/redmule/redmule.c b/hwpe/redmule/redmule.c
new file mode 100644
index 0000000..61a4b2b
--- /dev/null
+++ b/hwpe/redmule/redmule.c
@@ -0,0 +1,149 @@
+/*
+ * Copyright (C) 2022-2023 ETH Zurich and University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ * SPDX-License-Identifier: Apache-2.0
+ * 
+ * Author: Yvan Tortorella  <yvan.tortorella@unibo.it>
+ *
+ * RedMulE SW test
+ */
+
+#include <stdint.h>
+#include "stdio.h"
+#include "archi_redmule.h"
+#include "hal_redmule.h"
+#include "pulp.h"
+
+int main() {
+
+  volatile int errors = 0;
+  unsigned int cluster_id = rt_cluster_id();
+  #ifndef NO_ECC
+    unsigned int intc_data_correctable_cnt, redmule_data_correctable_cnt = 0;
+    unsigned int intc_meta_correctable_cnt = 0;
+    unsigned int intc_data_uncorrectable_cnt, redmule_data_uncorrectable_cnt = 0;
+    unsigned int intc_meta_uncorrectable_cnt = 0;
+  #endif
+
+  if(get_core_id() == 0){
+
+    uint16_t m_size = M_SIZE;
+    uint16_t n_size = N_SIZE;
+    uint16_t k_size = K_SIZE;
+
+    uint8_t *x_ext = x_inp;
+    uint8_t *w_ext = w_inp;
+    uint8_t *y_ext = y_inp;
+    uint8_t *z_ext = z_oup;
+
+    uint8_t volatile *x = (uint8_t volatile *) pi_l1_malloc(0, (2*m_size*n_size));
+    uint8_t volatile *w = (uint8_t volatile *) pi_l1_malloc(0, (2*n_size*k_size));
+    uint8_t volatile *y = (uint8_t volatile *) pi_l1_malloc(0, (2*m_size*k_size));
+
+    #ifdef USE_DMA
+      volatile unsigned int dma_id = 0;
+      dma_id = mchan_alloc();
+      mchan_transfer((unsigned int) 2*(2*m_size*n_size),
+                     (unsigned int) x_ext,
+                     (unsigned int) x    );
+      mchan_barrier(dma_id);
+      mchan_free(dma_id);
+    
+      dma_id = mchan_alloc();
+      mchan_transfer((unsigned int) 2*(2*n_size*k_size),
+                     (unsigned int) w_ext,
+                     (unsigned int) w    );
+      mchan_barrier(dma_id);
+      mchan_free(dma_id);
+    
+      dma_id = mchan_alloc();
+      mchan_transfer((unsigned int) 2*(2*m_size*k_size),
+                     (unsigned int) y_ext,
+                     (unsigned int) y    );
+      mchan_barrier(dma_id);
+    #else
+      generate_test_data16((int) x, (int) w, (int) y, (int) m_size, (int) n_size, (int) k_size);
+    #endif
+
+    int gold_sum = 0, check_sum = 0;
+    int i,j;
+
+    int offload_id_tmp, offload_id;
+
+    // Enable RedMulE
+    hwpe_cg_enable();
+    asm volatile("": : :"memory");
+
+    hwpe_soft_clear();
+    asm volatile("": : :"memory");
+
+    // redmule_cfg ((unsigned int) x,
+    //              (unsigned int) w,
+    //              (unsigned int) y,
+    //              m_size, n_size, k_size,
+    //              (uint8_t) GEMM,
+    //              (uint8_t) Float16);
+    redmule_x_add_set ((unsigned int) x);
+    redmule_w_add_set ((unsigned int) w);
+    redmule_y_add_set ((unsigned int) y);
+    redmule_z_add_set ((unsigned int) y);
+    redmule_cfg (m_size, n_size, k_size, gemm_ops);
+
+    // Start RedMulE operation
+    hwpe_trigger_job();
+
+    // Wait for end of computation
+    redmule_evt_wait();
+
+    #ifndef NO_ECC
+      // Check number of detected errors by ECC modules inside RedMulE
+      redmule_data_correctable_cnt = redmule_get_data_correctable_count();
+      redmule_data_uncorrectable_cnt = redmule_get_data_uncorrectable_count();
+    #endif
+
+    // Disable RedMulE
+    hwpe_cg_disable();
+
+    errors = redmule_compare16((int) y, (int) m_size, (int) k_size);
+
+    *(int *) 0x1A1040A0 = errors;
+
+    printf ("Terminated test with %d errors. See you!\n", errors);
+
+    #ifndef NO_ECC
+      // Check number of detected errors by ECC modules inside interconnect
+      intc_data_correctable_cnt = hwpe_hci_ecc_get_data_correctable_count(cluster_id);
+      intc_meta_correctable_cnt = hwpe_hci_ecc_get_meta_correctable_count(cluster_id);
+      intc_data_uncorrectable_cnt = hwpe_hci_ecc_get_data_uncorrectable_count(cluster_id);
+      intc_meta_uncorrectable_cnt = hwpe_hci_ecc_get_meta_uncorrectable_count(cluster_id);
+      for (int i = 0; i < 16; i++) {
+        intc_meta_correctable_cnt += tcdm_scrubber_get_mismatch_count(cluster_id, i);
+      }
+
+      printf ("Data errors corrected inside RedMulE: %d. Data errors uncorrectable inside RedMulE: %d \n",
+        redmule_data_correctable_cnt, redmule_data_uncorrectable_cnt);
+      printf("Data errors corrected inside intc: %d. Data errors uncorrectable inside intc: %d\n",
+        intc_data_correctable_cnt, intc_data_uncorrectable_cnt);
+      printf("Meta errors corrected inside intc: %d. Meta errors uncorrectable inside intc: %d\n",
+        intc_meta_correctable_cnt, intc_meta_uncorrectable_cnt);
+    #endif
+
+  }
+  synch_barrier();
+  #ifndef NO_ECC
+    return (errors != 0) && (redmule_data_uncorrectable_cnt==0 && intc_data_uncorrectable_cnt == 0 && intc_meta_uncorrectable_cnt == 0);
+  #else
+    return errors;
+  #endif
+}
diff --git a/hwpe/redmule_256iter/Makefile b/hwpe/redmule_256iter/Makefile
new file mode 100644
index 0000000..88346b6
--- /dev/null
+++ b/hwpe/redmule_256iter/Makefile
@@ -0,0 +1,20 @@
+PULP_APP = test
+PULP_APP_SRCS = redmule.c
+PULP_CFLAGS = -O3
+
+ifeq ($(use_dma),1)
+ PULP_CFLAGS += -DUSE_DMA
+endif
+
+ifeq ($(fault_inject),1)
+  export FAULT_INJECTION=1
+  export FAULT_INJECTION_SCRIPT=$(CURDIR)/pulp_inject_fault.tcl
+endif
+
+ifeq ($(multi_bit_upset),1)
+  export MULTI_BIT_UPSET=1
+else
+  export MULTI_BIT_UPSET=0
+endif
+
+include $(PULP_SDK_HOME)/install/rules/pulp_rt.mk
diff --git a/hwpe/redmule_256iter/archi_redmule.h b/hwpe/redmule_256iter/archi_redmule.h
new file mode 100644
index 0000000..40eceee
--- /dev/null
+++ b/hwpe/redmule_256iter/archi_redmule.h
@@ -0,0 +1,170 @@
+/*
+ * Copyright (C) 2022-2023 ETH Zurich and University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ * SPDX-License-Identifier: Apache-2.0
+ * 
+ * Author: Yvan Tortorella  <yvan.tortorella@unibo.it>
+ *
+ * High-level architecture of RedMulE
+ *
+ */
+
+#ifndef __ARCHI_REDMULE_H__
+#define __ARCHI_REDMULE_H__
+
+/*
+ * |========================================================================|
+ * ||                                                                      ||
+ * ||Control and generic configuration register layout                     ||
+ * |========================================================================|
+ * || # reg |  offset  |  bits   |   bitmask    ||  content                ||
+ * ||-------+----------+---------+--------------++-------------------------||
+ * ||    0  |  0x0000  |  31: 0  |  0xFFFFFFFF  ||  TRIGGER                ||
+ * ||    1  |  0x0004  |  31: 0  |  0xFFFFFFFF  ||  ACQUIRE                ||
+ * ||    2  |  0x0008  |  31: 0  |  0xFFFFFFFF  ||  EVT_ENABLE             ||
+ * ||    3  |  0x000c  |  31: 0  |  0xFFFFFFFF  ||  STATUS                 ||
+ * ||    4  |  0x0010  |  31: 0  |  0xFFFFFFFF  ||  RUNNING_JOB            ||
+ * ||    5  |  0x0014  |  31: 0  |  0xFFFFFFFF  ||  SOFT_CLEAR             ||
+ * |========================================================================|
+ * ||                                                                      ||
+ * ||Job-dependent registers layout                                        ||
+ * |========================================================================|
+ * || # reg |  offset  |  bits   |   bitmask    ||  content                ||
+ * ||-------+----------+---------+--------------++-------------------------||
+ * ||    0  |  0x0040  |  31: 0  |  0xFFFFFFFF  ||  X_ADDR                 ||
+ * ||-------+----------+---------+--------------++-------------------------||
+ * ||    1  |  0x0044  |  31: 0  |  0xFFFFFFFF  ||  W_ADDR                 ||
+ * ||-------+----------+---------+--------------++-------------------------||
+ * ||    2  |  0x0048  |  31: 0  |  0xFFFFFFFF  ||  Z_ADDR                 ||
+ * ||-------+----------+---------+--------------++-------------------------||
+ * ||    3  |  0x004C  |         |              ||  Matrix Config 0 Reg    ||
+ * ||       |          |  31:16  |  0xFFFF0000  ||  K Size (W Columns)     ||
+ * ||       |          |  15: 0  |  0x0000FFFF  ||  M Size (X Rows)        ||
+ * ||-------+----------+---------+--------------++-------------------------||
+ * ||    4  |  0x0050  |         |              ||  Matrix Config 1 Reg    ||
+ * ||       |          |  31:16  |  0xFFFFFFFF  ||  N Size (X Cols/W Rows) ||
+ * ||-------+----------+---------+--------------++-------------------------||
+ * ||    5  |  0x0054  |         |              ||  Matrix Arithmetic Reg  ||
+ * ||       |          |  12:10  |  0x00001C00  ||  Operation selection    ||
+ * ||       |          |   9: 7  |  0x00000380  ||  Input/Output format    ||
+ * |========================================================================|
+ *
+ */
+
+/* PULP Cluster Archi defines */
+#define ARCHI_CLUST_CTRL_BASE ARCHI_CLUSTER_CTRL_ADDR
+#define ARCHI_CLUST_HWPE_BASE ARCHI_HWCE_ADDR
+#define DMA_COMMAND_QUEUE     ARCHI_MCHAN_DEMUX_ADDR
+#define DMA_STATUS_REGISTER   (ARCHI_MCHAN_DEMUX_ADDR + 4)
+#define ARCHI_CL_HWPE_EVT0 12
+#define ARCHI_CL_HWPE_EVT1 13
+#define FC_DMA_EVENT 8
+#define CL_DMA_EVENT 22
+#define CLUST_CTRL_HWPE_EN 0x18
+#define CLUST_CTRL_HWPE_EN_MASK 0x800
+#define __builtin_bitinsert(a,b,c,d) (a | (((b << (32-c)) >> (32-c)) << d))
+
+// RedMulE architecture
+#define ADDR_WIDTH   32
+#define DATA_WIDTH   256
+#define REDMULE_FMT  16
+#define ARRAY_HEIGHT 4
+#define PIPE_REGS    3
+#define ARRAY_WIDTH  12 /* Superior limit is ARRAY_HEIGHT*PIPE_REGS */
+
+// Commands
+#define REDMULE_TRIGGER     0x00
+#define REDMULE_ACQUIRE     0x04
+#define REDMULE_FINISHED    0x08
+#define REDMULE_STATUS      0x0C
+#define REDMULE_RUNNING_JOB 0x10
+#define REDMULE_SOFT_CLEAR  0x14
+
+// Registers
+#define REDMULE_REG_OFFS  0x40
+// #define REDMULE_REG_X_PTR 0x00
+// #define REDMULE_REG_W_PTR 0x04
+// #define REDMULE_REG_Z_PTR 0x08
+// #define REDMULE_MCFG0_PTR 0x0C
+// #define REDMULE_MCFG1_PTR 0x10
+// #define REDMULE_ARITH_PTR 0x14
+#define REDMULE_REG_X_PTR              0x00
+#define REDMULE_REG_W_PTR              0x04
+#define REDMULE_REG_Y_PTR              0x08
+#define REDMULE_REG_Z_PTR              0x0C
+#define REDMULE_REG_X_ITER_PTR         0x10
+#define REDMULE_REG_W_ITER_PTR         0x14
+#define REDMULE_REG_LEFTOVERS_PTR      0x18
+#define REDMULE_REG_LEFT_PARAMS_PTR    0x1C
+#define REDMULE_REG_X_D1_STRIDE_PTR    0x20
+#define REDMULE_REG_W_TOT_LEN_PTR      0x24
+#define REDMULE_REG_TOT_X_READ_PTR     0x28
+#define REDMULE_REG_W_D0_STRIDE_PTR    0x2C
+#define REDMULE_REG_YZ_TOT_LEN_PTR     0x30
+#define REDMULE_REG_YZ_D0_STRIDE_PTR   0x34
+#define REDMULE_REG_YZ_D2_STRIDE_PTR   0x38
+#define REDMULE_REG_X_ROWS_OFFS_PTR    0x3C
+#define REDMULE_REG_X_BUFFER_SLOTS_PTR 0x40
+#define REDMULE_REG_X_TOT_LEN_PTR      0x44
+#define REDMULE_REG_OP_SELECTION       0x48
+
+#define REDMULE_ECC_REG_OFFS           0x90
+#define DATA_CORR_ERR                  0x00
+#define DATA_UNCORR_ERR                0x04
+#define METADATA_CORR_ERR              0x08
+#define METADATA_UNCORR_ERR            0x0c
+
+// OPs definition
+#define MATMUL 0x0
+#define GEMM   0x1
+#define ADDMAX 0x2
+#define ADDMIN 0x3
+#define MULMAX 0x4
+#define MULMIN 0x5
+#define MAXMIN 0x6
+#define MINMAX 0x7
+
+// GEMM formats
+#define Float8     0x0
+#define Float16    0x1
+#define Float8Alt  0x2
+#define Float16Alt 0x3
+
+#define RNE       0x0
+#define RTZ       0x1
+#define OP_FMADD  0x0
+#define OP_ADD    0x2
+#define OP_MUL    0x3
+#define OP_MINMAX 0x7
+
+// FP Formats encoding
+#define FP16    0x2
+#define FP8     0x3
+#define FP16ALT 0x4
+#define FP8ALT  0x5
+
+/* DMA Archi */
+#define DMA_TX  0
+#define DMA_RX  1
+#define DMA_INC 1
+
+#define PLP_DMA_TYPE_BIT    0x00000011
+#define PLP_DMA_INCR_BIT    0x00000012
+#define PLP_DMA_2D_BIT      0x00000013
+#define PLP_DMA_ELE_BIT     0x00000014
+#define PLP_DMA_ILE_BIT     0x00000015
+#define PLP_DMA_BLE_BIT     0x00000016
+#define PLP_DMA_2D_TCDM_BIT 0x0000017
+
+#endif
diff --git a/hwpe/redmule_256iter/hal_redmule.h b/hwpe/redmule_256iter/hal_redmule.h
new file mode 100644
index 0000000..8fc5000
--- /dev/null
+++ b/hwpe/redmule_256iter/hal_redmule.h
@@ -0,0 +1,556 @@
+/*
+ * Copyright (C) 2022-2023 ETH Zurich and University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ * SPDX-License-Identifier: Apache-2.0
+ * 
+ * Author: Yvan Tortorella  <yvan.tortorella@unibo.it>
+ *
+ * RedMulE Hardware Abstraction Layer (HAL)
+ */
+
+#ifndef __HAL_REDMULE_H__
+#define __HAL_REDMULE_H__
+
+#include <pulp.h>
+#include "inc/x_input.h"
+#include "inc/w_input.h"
+#include "inc/y_input.h"
+#include "inc/z_output.h"
+#include "inc/golden.h"
+#include "inc/tensor_dim.h"
+
+/*
+ * 
+ * For control, generic configuration register layout,
+ * and job-dependent register map, look at redmule_archi.h
+ *
+ */
+
+// For all the following functions we use __builtin_pulp_OffsetedWrite and __builtin_pulp_OffsetedRead
+// instead of classic load/store because otherwise the compiler is not able to correctly factorize
+// the HWPE base in case several accesses are done, ending up with twice more code
+
+#define HWPE_WRITE(value, offset) *(int *)(ARCHI_CLUST_HWPE_BASE + offset) = value
+#define HWPE_READ(offset) *(int *)(ARCHI_CLUST_HWPE_BASE + offset)
+
+static inline void redmule_x_add_set (unsigned int value) {
+  HWPE_WRITE(value, REDMULE_REG_OFFS + REDMULE_REG_X_PTR);
+}
+
+static inline void redmule_w_add_set (unsigned int value) {
+  HWPE_WRITE(value, REDMULE_REG_OFFS + REDMULE_REG_W_PTR);
+}
+
+static inline void redmule_y_add_set (unsigned int value) {
+  HWPE_WRITE(value, REDMULE_REG_OFFS + REDMULE_REG_Y_PTR);
+}
+
+static inline void redmule_z_add_set (unsigned int value) {
+  HWPE_WRITE(value, REDMULE_REG_OFFS + REDMULE_REG_Z_PTR);
+}
+
+// static inline void redmule_mcfg_set (uint32_t mcfg0, uint32_t mcfg1) {
+//   HWPE_WRITE(mcfg0, REDMULE_REG_OFFS + REDMULE_MCFG0_PTR);
+//   HWPE_WRITE(mcfg1, REDMULE_REG_OFFS + REDMULE_MCFG1_PTR);
+// }
+// 
+// static inline void redmule_arith_set (uint32_t arith) {
+//   HWPE_WRITE(arith, REDMULE_REG_OFFS + REDMULE_ARITH_PTR);
+// }
+
+static inline void hwpe_trigger_job() {
+  HWPE_WRITE(0, REDMULE_TRIGGER);
+}
+
+static inline int hwpe_acquire_job() {
+  return HWPE_READ(REDMULE_ACQUIRE);
+}
+
+static inline unsigned int hwpe_get_status() {
+  return HWPE_READ(REDMULE_STATUS);
+}
+
+static inline unsigned int hwpe_get_running_job() {
+  return HWPE_READ(REDMULE_RUNNING_JOB);
+}
+
+static inline void hwpe_soft_clear() {
+  HWPE_WRITE(0, REDMULE_SOFT_CLEAR);
+}
+
+static inline void hwpe_cg_enable() {
+  *(volatile int*) (ARCHI_CLUST_CTRL_BASE + CLUST_CTRL_HWPE_EN) |= CLUST_CTRL_HWPE_EN_MASK;
+}
+
+static inline void hwpe_cg_disable() {
+  *(volatile int*) (ARCHI_CLUST_CTRL_BASE + CLUST_CTRL_HWPE_EN) &= ~CLUST_CTRL_HWPE_EN_MASK;
+}
+
+static inline void redmule_evt_wait() {
+  do {
+    eu_evt_maskWaitAndClr (1 << ARCHI_CL_HWPE_EVT0);
+  } while((*(int volatile *)(ARCHI_CLUST_HWPE_BASE + REDMULE_STATUS)) != 0);
+}
+
+static inline int hwpe_wait_acquire() {
+  int job_id = hwpe_acquire_job();
+  while(job_id < 0) {
+    eu_evt_maskWaitAndClr (1 << ARCHI_CL_HWPE_EVT0);
+    job_id = hwpe_acquire_job();
+  }
+  return job_id;
+}
+
+static inline unsigned int redmule_get_data_correctable_count () {
+  return HWPE_READ(REDMULE_ECC_REG_OFFS + DATA_CORR_ERR);
+}
+
+static inline unsigned int redmule_get_data_uncorrectable_count () {
+  return HWPE_READ(REDMULE_ECC_REG_OFFS + DATA_UNCORR_ERR);
+}
+
+static inline unsigned int redmule_get_meta_correctable_count () {
+  return HWPE_READ(REDMULE_ECC_REG_OFFS + METADATA_CORR_ERR);
+}
+
+static inline unsigned int redmule_get_meta_uncorrectable_count () {
+  return HWPE_READ(REDMULE_ECC_REG_OFFS + METADATA_UNCORR_ERR);
+}
+
+/* DMA APIs */
+static inline int mchan_alloc(){
+  return *(volatile int*) DMA_COMMAND_QUEUE;
+}
+
+static inline void mchan_transfer(unsigned int len,
+                                  unsigned int ext_addr,
+                                  unsigned int tcdm_addr) {
+
+  *(volatile int*) DMA_COMMAND_QUEUE = len |
+                                       (DMA_RX << PLP_DMA_TYPE_BIT)  |
+                                       (DMA_INC << PLP_DMA_INCR_BIT) |
+                                       (0 << PLP_DMA_2D_BIT)         |
+                                       (1 << PLP_DMA_ELE_BIT)        |
+                                       (1 << PLP_DMA_ILE_BIT)        |
+                                       (0 << PLP_DMA_BLE_BIT)        |
+                                       (0 << PLP_DMA_2D_TCDM_BIT);
+  *(volatile int*) DMA_COMMAND_QUEUE = tcdm_addr;
+  *(volatile int*) DMA_COMMAND_QUEUE = ext_addr;
+}
+
+static inline void mchan_barrier(int id) {
+  while(((*(volatile int*)(DMA_STATUS_REGISTER)) >> id ) & 0x1 ) {
+    eu_evt_maskWaitAndClr(1 << FC_DMA_EVENT);
+ }
+}
+
+static inline void mchan_free(int id) {
+  *(volatile int*) DMA_STATUS_REGISTER = 0x1 << id;
+}
+
+// void redmule_cfg (unsigned int x,  unsigned int w,  unsigned int z,
+//                   uint16_t m_size, uint16_t n_size, uint16_t k_size,
+//                   uint8_t gemm_op, uint8_t gemm_fmt){
+// 
+//   uint32_t mcfg_reg0 = 0;
+//   uint32_t mcfg_reg1 = 0;
+//   uint32_t arith_reg = 0;
+// 
+//   mcfg_reg0 = (k_size << 16) |
+//               (m_size <<  0);
+//   mcfg_reg1 = n_size << 0;
+// 
+//   arith_reg = (gemm_op << 10) |
+//               (gemm_fmt << 7);
+// 
+//   redmule_x_add_set ((unsigned int) x);
+//   redmule_w_add_set ((unsigned int) w);
+//   redmule_z_add_set ((unsigned int) z);
+//   redmule_mcfg_set  ((unsigned int) mcfg_reg0,
+//                      (unsigned int) mcfg_reg1);
+//   redmule_arith_set ((unsigned int) arith_reg);
+//   
+// }
+
+void redmule_cfg (uint16_t m_size, uint16_t n_size, uint16_t k_size, uint8_t gemm_ops){
+   uint32_t x_iters        = 0;
+   uint32_t w_iters        = 0;
+   uint32_t leftovers      = 0;
+   uint32_t left_params    = 0;
+   uint32_t x_d1_stride    = 0;
+   uint32_t x_rows_offs    = 0;
+   uint32_t w_tot_len      = 0;
+   uint32_t w_d1_len       = 0;
+   uint32_t w_d0_stride    = 0;
+   uint32_t yz_tot_len     = 0;
+   uint32_t yz_d0_stride   = 0;
+   uint32_t yz_d2_stride   = 0;
+   uint32_t tot_x_read     = 0;
+   uint32_t x_buffer_slots = 0;
+   uint32_t op_selection   = 0;
+   uint16_t tot_stores     = 0;
+   uint16_t w_rows         = n_size;
+   uint16_t depth          = DATA_WIDTH/(ARRAY_HEIGHT*FPFORMAT);
+   uint8_t  tile           = ARRAY_HEIGHT*(PIPE_REGS + 1);
+   _Bool    x_rows_sub     = 0;
+   _Bool    x_cols_sub     = 0;
+   _Bool    w_cols_sub     = 0;
+   uint16_t x_rows_iter,
+            x_rows_iter_tmp,
+            w_rows_iter,
+            w_rows_iter_tmp;
+   uint16_t x_cols_iter,
+            x_cols_iter_tmp,
+            w_cols_iter,
+            w_cols_iter_tmp;
+   uint8_t  x_rows_lftovr,
+            x_cols_lftovr,
+            w_rows_lftovr,
+            w_cols_lftovr,
+            slots;
+
+   // Calculating the number of iterations alng the two dimensions of the X matrix
+   x_rows_iter_tmp = m_size/ARRAY_WIDTH;
+   x_cols_iter_tmp = n_size/tile;
+
+   // Calculating the number of iterations alng the two dimensions of the W matrix
+   w_rows_iter_tmp = w_rows;
+   w_cols_iter_tmp = k_size/tile;
+
+   // Calculating the residuals along the input dimensions
+   x_rows_lftovr = m_size - (x_rows_iter_tmp*ARRAY_WIDTH);
+   x_cols_lftovr = n_size - (x_cols_iter_tmp*tile);
+
+   // Calculating the residuals along the weight dimensions
+   w_rows_lftovr = n_size - (ARRAY_HEIGHT*(w_rows/ARRAY_HEIGHT));
+   w_cols_lftovr = k_size - (w_cols_iter_tmp*tile);
+
+   if (w_cols_lftovr != 0)
+     w_cols_iter = w_cols_iter_tmp + 1;
+   else 
+     w_cols_iter = w_cols_iter_tmp;
+
+  if (w_rows_lftovr != 0)
+    w_rows_iter = w_rows_iter_tmp + ARRAY_HEIGHT - w_rows_lftovr;
+  else
+    w_rows_iter = w_rows_iter_tmp;
+
+   if (x_cols_lftovr != 0)
+     x_cols_iter = x_cols_iter_tmp + 1;
+   else 
+     x_cols_iter = x_cols_iter_tmp;
+
+   if (x_rows_lftovr != 0)
+     x_rows_iter = x_rows_iter_tmp + 1;
+   else 
+     x_rows_iter = x_rows_iter_tmp;
+
+   if (x_cols_lftovr%depth != 0)
+     x_buffer_slots = x_cols_lftovr/depth + 1;
+   else
+     x_buffer_slots = x_cols_lftovr/depth;
+
+   // Calculating the number of total stores
+   tot_stores = x_rows_iter*w_cols_iter;
+
+   // Determining if input matrixes are sub-matrixes
+   if (m_size < ARRAY_WIDTH)
+     x_rows_sub = 1;
+   if (n_size < ARRAY_HEIGHT)
+     x_cols_sub = 1;
+   if (k_size < tile)
+    w_cols_sub = 1;
+
+   // Operation selection
+   switch (gemm_ops) {
+     case MATMUL:
+       op_selection |= (RNE << 29 | RNE << 26 | OP_FMADD << 22 | OP_MINMAX << 18 | SRC_FMT << 15 | DST_FMT << 12) | 0;
+     break;
+     
+     case GEMM:
+       op_selection |= (RNE << 29 | RNE << 26 | OP_FMADD << 22 | OP_MINMAX << 18 | SRC_FMT << 15 | DST_FMT << 12) | 1;
+     break;
+     
+     case ADDMAX:
+       op_selection |= (RNE << 29 | RTZ << 26 | OP_ADD << 22 | OP_MINMAX << 18 | SRC_FMT << 15 | DST_FMT << 12) | 1;
+     break;
+     
+     case ADDMIN:
+       op_selection |= (RNE << 29 | RNE << 26 | OP_ADD << 22 | OP_MINMAX << 18 | SRC_FMT << 15 | DST_FMT << 12) | 1;
+     break;
+     
+     case MULMAX:
+       op_selection |= (RNE << 29 | RTZ << 26 | OP_MUL << 22 | OP_MINMAX << 18 | SRC_FMT << 15 | DST_FMT << 12) | 1;
+     break;
+     
+     case MULMIN:
+       op_selection |= (RNE << 29 | RNE << 26 | OP_MUL << 22 | OP_MINMAX << 18 | SRC_FMT << 15 | DST_FMT << 12) | 1;
+     break;
+     
+     case MAXMIN:
+       op_selection |= (RTZ << 29 | RNE << 26 | OP_MINMAX << 22 | OP_MINMAX << 18 | SRC_FMT << 15 | DST_FMT << 12) | 1;
+     break;
+     
+     case MINMAX:
+       op_selection |= (RNE << 29 | RTZ << 26 | OP_MINMAX << 22 | OP_MINMAX << 18 | SRC_FMT << 15 | DST_FMT << 12) | 1;
+     break;
+   }
+
+   // Storing iterations and residuals in registers
+   x_iters      |= x_rows_iter   << 16 | x_cols_iter   << 0;
+   w_iters      |= w_rows_iter   << 16 | w_cols_iter   << 0;
+   leftovers    |= x_rows_lftovr << 24 | x_cols_lftovr << 16 | w_rows_lftovr << 8  | w_cols_lftovr << 0;
+   left_params  |= tot_stores    << 16 | x_rows_sub    << 15 | x_cols_sub    << 14 | w_cols_sub   << 13;
+   x_d1_stride   = ((4*FPFORMAT)/ADDR_WIDTH)*(((DATA_WIDTH/FPFORMAT)*x_cols_iter_tmp) + x_cols_lftovr);
+   x_rows_offs   = ARRAY_WIDTH*x_d1_stride;
+   w_tot_len     = w_rows_iter*w_cols_iter*x_rows_iter;
+   w_d0_stride   = ((4*FPFORMAT)/ADDR_WIDTH)*(((DATA_WIDTH/FPFORMAT)*w_cols_iter_tmp) + w_cols_lftovr);
+   yz_tot_len    = ARRAY_WIDTH*x_rows_iter*w_cols_iter;
+   yz_d0_stride  = w_d0_stride;
+   yz_d2_stride  = ARRAY_WIDTH*w_d0_stride;
+   tot_x_read    = x_rows_iter*x_cols_iter*w_cols_iter;
+
+   // Writing the computations in configuration register
+   HWPE_WRITE(x_iters       , REDMULE_REG_OFFS + REDMULE_REG_X_ITER_PTR         );
+   HWPE_WRITE(w_iters       , REDMULE_REG_OFFS + REDMULE_REG_W_ITER_PTR         );
+   HWPE_WRITE(leftovers     , REDMULE_REG_OFFS + REDMULE_REG_LEFTOVERS_PTR      );
+   HWPE_WRITE(left_params   , REDMULE_REG_OFFS + REDMULE_REG_LEFT_PARAMS_PTR    );
+   HWPE_WRITE(x_d1_stride   , REDMULE_REG_OFFS + REDMULE_REG_X_D1_STRIDE_PTR    );
+   HWPE_WRITE(x_rows_offs   , REDMULE_REG_OFFS + REDMULE_REG_X_ROWS_OFFS_PTR    );
+   HWPE_WRITE(tot_x_read    , REDMULE_REG_OFFS + REDMULE_REG_TOT_X_READ_PTR     );
+   HWPE_WRITE(x_buffer_slots, REDMULE_REG_OFFS + REDMULE_REG_X_BUFFER_SLOTS_PTR );
+   HWPE_WRITE(w_tot_len     , REDMULE_REG_OFFS + REDMULE_REG_W_TOT_LEN_PTR      );
+   HWPE_WRITE(w_d0_stride   , REDMULE_REG_OFFS + REDMULE_REG_W_D0_STRIDE_PTR    );
+   HWPE_WRITE(yz_tot_len    , REDMULE_REG_OFFS + REDMULE_REG_YZ_TOT_LEN_PTR     );
+   HWPE_WRITE(yz_d0_stride  , REDMULE_REG_OFFS + REDMULE_REG_YZ_D0_STRIDE_PTR   );
+   HWPE_WRITE(yz_d2_stride  , REDMULE_REG_OFFS + REDMULE_REG_YZ_D2_STRIDE_PTR   );
+   HWPE_WRITE(op_selection  , REDMULE_REG_OFFS + REDMULE_REG_OP_SELECTION       );
+}
+
+void generate_test_data16(int x_start_addr,
+                          int w_start_addr,
+                          int y_start_addr,
+                          int m_size,
+                          int n_size,
+                          int k_size) {
+  
+  int x_addr     = x_start_addr;
+  int w_addr     = w_start_addr;
+  int y_addr     = y_start_addr;
+  int x_end_addr = x_start_addr + (2*m_size*n_size);
+  int w_end_addr = w_start_addr + (2*n_size*k_size);
+  int y_end_addr = y_start_addr + (2*m_size*k_size);
+
+  // Generating input stimuli from golden model
+  for (x_addr = x_start_addr; x_addr < x_end_addr; x_addr += 2) {
+    int x = x_addr - x_start_addr;
+    *(uint32_t *)(x_addr) = x_inp[x/2];
+  }
+
+  // Generating Weight stimuli from golden model
+  for (w_addr = w_start_addr; w_addr < w_end_addr; w_addr += 2) {
+    int w = w_addr - w_start_addr;
+    *(uint32_t *)(w_addr) = w_inp[w/2];
+  }
+
+  for (y_addr = y_start_addr; y_addr < y_end_addr; y_addr += 2) {
+    int y = y_addr - y_start_addr;
+    *(uint32_t *)(y_addr) = y_inp[y/2];
+  }
+}
+
+int redmule_compare16 (int z_start_addr, int m_size, int k_size) {
+  int err = 0;
+  int z_end_addr = z_start_addr + 2*m_size*k_size;
+  uint16_t z_computed;
+  uint16_t diff, diff_1, diff_2;
+
+  for (int z_addr = z_start_addr; z_addr < z_end_addr; z_addr += 2) {
+    int z = z_addr - z_start_addr;
+    z_computed = *(uint32_t *)(z_addr);
+
+    if ( z_computed != z_oup[z/2] ) {
+      diff_1 = z_computed - z_oup[z/2];
+      if (diff_1 > 3) {
+        diff_2 = z_oup[z/2] - z_computed;
+        if (diff_2 > 3) {
+          err++;
+        }
+      }
+    }
+  }
+
+  return err;
+
+}
+
+int redmule16_compare_int(uint32_t *actual_z, uint32_t *golden_z, int len) {
+  #define ERR 0x0011
+  uint32_t actual_word = 0;
+  uint16_t actual_MSHWord, actual_LSHWord;
+  uint32_t golden_word = 0;
+  uint16_t golden_MSHWord, golden_LSHWord;
+  uint32_t actual = 0;
+  uint32_t golden = 0;
+
+  int errors = 0;
+  int error;
+
+  for (int i=0; i<len; i++) {
+    error = 0;
+    actual_word = *(actual_z+i);
+    golden_word = *(golden_z+i);
+
+    uint16_t diff = 0;
+    
+    // Chechink Least Significant Half-Word
+    actual_LSHWord = (uint16_t)(actual_word & 0x0000FFFF);
+    golden_LSHWord = (uint16_t)(golden_word & 0x0000FFFF);
+
+    diff = (actual_LSHWord > golden_LSHWord) ? (actual_LSHWord - golden_LSHWord) : 0;
+    diff = (actual_LSHWord < golden_LSHWord) ? (golden_LSHWord - actual_LSHWord) : 0;
+
+    if (diff > ERR) {
+      error = 1;
+      #ifdef VERBOSE
+        tfp_printf ("diff: 0x%08x\n", diff);
+        tfp_printf ("LSW: Error!\n");
+      #endif
+    }
+
+    // Checking Most Significant Half-Word
+    actual_MSHWord = (uint16_t)((actual_word >> 16) & 0x0000FFFF);
+    golden_MSHWord = (uint16_t)((golden_word >> 16) & 0x0000FFFF);
+
+    diff = (actual_MSHWord > golden_MSHWord) ? (actual_MSHWord - golden_MSHWord) : 0;
+    diff = (actual_MSHWord < golden_MSHWord) ? (golden_MSHWord - actual_MSHWord) : 0;
+
+    if (diff > ERR) {
+      error = 1;
+      #ifdef VERBOSE
+        tfp_printf ("diff: 0x%08x\n", diff);
+        tfp_printf ("MSW: Error!\n");
+      #endif
+    }
+    
+    errors += error;
+
+    #ifdef DEBUG
+      tfp_printf("  Golden: 0x%08x; Actual: 0x%08x,\n", golden_word, actual_word);
+    #endif
+
+    #ifdef VERBOSE
+    if(error) {
+        if(errors==1) tfp_printf("  golden     <- actual     @ address    @ index\n");
+        tfp_printf("  0x%08x <- 0x%08x @ 0x%08x @ 0x%08x\n", golden_word, actual_word, (actual_z+i), i*4);
+    }
+    #endif
+  }
+  return errors;
+}
+
+int redmule8_compare_int(uint32_t *actual_z, uint32_t *golden_z, int len) {
+  #define ERR 0x0011
+  uint32_t actual_word = 0;
+  uint8_t  actual_Byte0,
+           actual_Byte1,
+           actual_Byte2,
+           actual_Byte3;
+  uint32_t golden_word = 0;
+  uint8_t  golden_Byte0,
+           golden_Byte1,
+           golden_Byte2,
+           golden_Byte3;
+  uint32_t actual = 0;
+  uint32_t golden = 0;
+
+  int errors = 0;
+  int error;
+
+  for (int i=0; i<len; i++) {
+    error = 0;
+    actual_word = *(actual_z+i);
+    golden_word = *(golden_z+i);
+
+    uint8_t diff = 0;
+    
+    // Cheching Byte0
+    actual_Byte0 = (uint8_t)(actual_word & 0x000000FF);
+    golden_Byte0 = (uint8_t)(golden_word & 0x000000FF);
+
+    diff = (actual_Byte0 > golden_Byte0) ? (actual_Byte0 - golden_Byte0) : 0;
+    diff = (actual_Byte0 < golden_Byte0) ? (golden_Byte0 - actual_Byte0) : 0;
+
+    if (diff > ERR) {
+      error = 1;
+      tfp_printf ("diff: 0x%08x\n", diff);
+      tfp_printf ("Byte0: Error!\n");
+    }
+
+    // Cheching Byte1
+    actual_Byte1 = (uint8_t)( (actual_word >> 8 ) & 0x000000FF);
+    golden_Byte1 = (uint8_t)( (golden_word >> 8 ) & 0x000000FF);
+
+    diff = (actual_Byte1 > golden_Byte1) ? (actual_Byte1 - golden_Byte1) : 0;
+    diff = (actual_Byte1 < golden_Byte1) ? (golden_Byte1 - actual_Byte1) : 0;
+
+    if (diff > ERR) {
+      error = 1;
+      tfp_printf ("diff: 0x%08x\n", diff);
+      tfp_printf ("Byte1: Error!\n");
+    }
+
+    // Cheching Byte2
+    actual_Byte2 = (uint8_t)( (actual_word >> 16 ) & 0x000000FF);
+    golden_Byte2 = (uint8_t)( (golden_word >> 16 ) & 0x000000FF);
+
+    diff = (actual_Byte2 > golden_Byte2) ? (actual_Byte2 - golden_Byte2) : 0;
+    diff = (actual_Byte2 < golden_Byte2) ? (golden_Byte2 - actual_Byte2) : 0;
+
+    if (diff > ERR) {
+      error = 1;
+      tfp_printf ("diff: 0x%08x\n", diff);
+      tfp_printf ("Byte2: Error!\n");
+    }
+
+    // Cheching Byte3
+    actual_Byte3 = (uint8_t)( (actual_word >> 24 ) & 0x000000FF);
+    golden_Byte3 = (uint8_t)( (golden_word >> 24 ) & 0x000000FF);
+
+    diff = (actual_Byte3 > golden_Byte3) ? (actual_Byte3 - golden_Byte3) : 0;
+    diff = (actual_Byte3 < golden_Byte3) ? (golden_Byte3 - actual_Byte3) : 0;
+
+    if (diff > ERR) {
+      error = 1;
+      tfp_printf ("diff: 0x%08x\n", diff);
+      tfp_printf ("Byte3: Error!\n");
+    }
+    
+    errors += error;
+
+    #ifdef DEBUG
+      tfp_printf("  Golden: 0x%08x; Actual: 0x%08x,\n", golden_word, actual_word);
+    #endif
+
+    #ifdef VERBOSE
+      if(error) {
+        if(errors==1) tfp_printf("  golden     <- actual     @ address    @ index\n");
+        tfp_printf("  0x%08x <- 0x%08x @ 0x%08x @ 0x%08x\n", golden_word, actual_word, (actual_z+i), i*4);
+      }
+    #endif
+  }
+  return errors;
+}
+
+#endif
diff --git a/hwpe/redmule_256iter/inc/golden.h b/hwpe/redmule_256iter/inc/golden.h
new file mode 100644
index 0000000..f664e47
--- /dev/null
+++ b/hwpe/redmule_256iter/inc/golden.h
@@ -0,0 +1,387 @@
+ /* Header file generated by RedMulE Golden Model */
+uint32_t golden [384] = {
+0x48974845,
+0x48384608,
+0x487b4855,
+0x48804869,
+0x48b046d1,
+0x483f48db,
+0x485f48c9,
+0x483a4881,
+0x472c484b,
+0x492b4762,
+0x48fd4822,
+0x492e488e,
+0x484f483e,
+0x46d749e8,
+0x489d484b,
+0x47e9490b,
+0x47d2484f,
+0x474744be,
+0x46c047c7,
+0x48af4727,
+0x482d46c5,
+0x482e483d,
+0x479f4897,
+0x4749488b,
+0x46a8489a,
+0x488b46f2,
+0x47e84891,
+0x483d4872,
+0x46fd4716,
+0x46a049b5,
+0x47a446e7,
+0x476748a1,
+0x49354939,
+0x48c14703,
+0x48bd4863,
+0x48cf4913,
+0x48b848b6,
+0x49204946,
+0x48e1495e,
+0x48b24938,
+0x4882493a,
+0x49d5483b,
+0x49724911,
+0x49df496b,
+0x488848f2,
+0x48214a46,
+0x490c48c1,
+0x48a349b2,
+0x47b0463a,
+0x476244cb,
+0x46b94765,
+0x4814466a,
+0x47964631,
+0x474b4666,
+0x47044798,
+0x47614838,
+0x459047d3,
+0x48a245ea,
+0x484447f1,
+0x4776484b,
+0x46d847d6,
+0x44d348f3,
+0x478d46fa,
+0x466e481e,
+0x481e4827,
+0x479445a2,
+0x48064727,
+0x48d5475d,
+0x48284708,
+0x480d4862,
+0x48324895,
+0x47f148bd,
+0x46a7482a,
+0x492d47b1,
+0x4884484d,
+0x485f48dc,
+0x480c476d,
+0x46d348e9,
+0x48844728,
+0x480e48a0,
+0x48134862,
+0x485a4675,
+0x473847e8,
+0x48234836,
+0x482146e7,
+0x47b34822,
+0x48554846,
+0x47174863,
+0x47c14872,
+0x488e46d5,
+0x485f47e2,
+0x48b8487c,
+0x4788481e,
+0x467748bd,
+0x47f846c9,
+0x47fc48fe,
+0x47b247a0,
+0x467e4588,
+0x46c74662,
+0x481246e8,
+0x474e4536,
+0x468f46c0,
+0x4679481f,
+0x46e246a1,
+0x45604809,
+0x47eb4630,
+0x475746b5,
+0x477f4848,
+0x46d846a6,
+0x459a4870,
+0x46784670,
+0x468c47d2,
+0x48c44762,
+0x479146e3,
+0x486d46b1,
+0x486747d0,
+0x47f6468d,
+0x475648a5,
+0x48544857,
+0x48384866,
+0x46ec484d,
+0x48f647d2,
+0x4879484a,
+0x483c4848,
+0x4806471d,
+0x473048fa,
+0x47b84768,
+0x46f94865,
+0x491848a8,
+0x486746ca,
+0x48624800,
+0x491048d3,
+0x4849474e,
+0x486b48eb,
+0x48c54966,
+0x483048f4,
+0x477848f9,
+0x499e481e,
+0x48f148cf,
+0x49234982,
+0x47cf487c,
+0x464949ea,
+0x495e4773,
+0x483f48b2,
+0x497548a7,
+0x481e4616,
+0x4866481f,
+0x486448b6,
+0x487347dc,
+0x487f485c,
+0x491f4938,
+0x48b6490d,
+0x48a148f8,
+0x492d4859,
+0x4915489c,
+0x48874899,
+0x4859486c,
+0x471e49ca,
+0x49184867,
+0x482748d3,
+0x4998488b,
+0x481d4704,
+0x488048b8,
+0x49444876,
+0x48f2470c,
+0x489b48b9,
+0x48e54956,
+0x48a548d6,
+0x485648dc,
+0x49ab484e,
+0x490e48e0,
+0x494548dd,
+0x48dd488b,
+0x47ea4a32,
+0x49114835,
+0x48194965,
+0x481e460e,
+0x4673452c,
+0x4717475c,
+0x46d046f6,
+0x46bc4696,
+0x481e4726,
+0x46ea4763,
+0x475846fe,
+0x4627478b,
+0x483f4704,
+0x47b146ad,
+0x48164792,
+0x468446f2,
+0x45a84827,
+0x47a4472f,
+0x462b4797,
+0x48ab483f,
+0x4863468f,
+0x4766485a,
+0x48cb481d,
+0x490347dc,
+0x483048fc,
+0x483e48cc,
+0x486448ab,
+0x47634966,
+0x499d4794,
+0x488b488e,
+0x496048dc,
+0x484c4854,
+0x474c499c,
+0x48bc4826,
+0x48834949,
+0x4905489d,
+0x481e4718,
+0x48f448e3,
+0x490448c1,
+0x48b347e8,
+0x48d44892,
+0x489448ff,
+0x488648d5,
+0x480348fa,
+0x492e47d2,
+0x48b24870,
+0x492b48e5,
+0x4785487b,
+0x471d49e3,
+0x48bf4837,
+0x48c4489b,
+0x4871475c,
+0x4811464a,
+0x471c47af,
+0x48174817,
+0x484e463b,
+0x464f477f,
+0x487c4704,
+0x472547a3,
+0x462a4853,
+0x4860465a,
+0x48804736,
+0x482b47e1,
+0x46c04811,
+0x475d48dc,
+0x48064668,
+0x46f44893,
+0x49594858,
+0x487b463d,
+0x484e480f,
+0x48a648c0,
+0x48944847,
+0x484a48a0,
+0x48f4491e,
+0x48b548fc,
+0x47d248ce,
+0x497f47db,
+0x49394955,
+0x48ce48a7,
+0x48844890,
+0x476349d6,
+0x4922486e,
+0x48c348f4,
+0x491c47ec,
+0x47834698,
+0x47544715,
+0x47524745,
+0x4832472f,
+0x48094817,
+0x48c347f8,
+0x480047e6,
+0x473048b6,
+0x48cb480a,
+0x488e479e,
+0x488e47c2,
+0x47ee472f,
+0x4744489d,
+0x48514755,
+0x47d34846,
+0x48a04838,
+0x47624634,
+0x48064786,
+0x482d47e3,
+0x486c4726,
+0x480347b7,
+0x481448ac,
+0x483948e0,
+0x47504827,
+0x48c546f2,
+0x4886483f,
+0x485648ad,
+0x47a947e8,
+0x47434937,
+0x481f46d0,
+0x4804484c,
+0x481f47fd,
+0x4813456d,
+0x4807474d,
+0x480e4688,
+0x481046e8,
+0x4799469f,
+0x478f4853,
+0x482447f2,
+0x471f47d0,
+0x485f46da,
+0x481c4813,
+0x4863482e,
+0x480b4786,
+0x46b848c9,
+0x46e2475a,
+0x46c54852,
+0x480245af,
+0x46c24466,
+0x4743465d,
+0x47ba46b7,
+0x46c34636,
+0x47844677,
+0x47c2485a,
+0x46ac46dc,
+0x460e47de,
+0x4834465f,
+0x476947f4,
+0x481046fc,
+0x45ea45fd,
+0x45b548d0,
+0x47834704,
+0x46c44830,
+0x47c74759,
+0x45b0453d,
+0x47024741,
+0x47934736,
+0x47ba461b,
+0x46dd470b,
+0x470b4657,
+0x4710470d,
+0x468f486c,
+0x46ba45c3,
+0x483b479d,
+0x477446c9,
+0x46a746a9,
+0x46064833,
+0x46a94690,
+0x46a746f5,
+0x48bb47ac,
+0x4803452c,
+0x4824470f,
+0x48cb47d5,
+0x484a4707,
+0x47974832,
+0x482c4851,
+0x4877487a,
+0x465d4891,
+0x48ce47f4,
+0x48994898,
+0x486a484e,
+0x47f047ac,
+0x4611493e,
+0x489e47e2,
+0x46af488c,
+0x48364665,
+0x46b645e4,
+0x46b946a1,
+0x46dd46c8,
+0x474b4658,
+0x4777467b,
+0x47984769,
+0x475e4785,
+0x4656472a,
+0x488145fb,
+0x472d46fc,
+0x47a3476e,
+0x46ca465d,
+0x45004855,
+0x479a464f,
+0x473846c3,
+0x486c481e,
+0x48014659,
+0x477a4756,
+0x487b47d5,
+0x48084706,
+0x4838484f,
+0x48634870,
+0x480648d3,
+0x47714865,
+0x494c46be,
+0x484c4915,
+0x48624900,
+0x46e8481a,
+0x46a04974,
+0x483d4775,
+0x480e487c,
+};
\ No newline at end of file
diff --git a/hwpe/redmule_256iter/inc/tensor_dim.h b/hwpe/redmule_256iter/inc/tensor_dim.h
new file mode 100644
index 0000000..21bd0d8
--- /dev/null
+++ b/hwpe/redmule_256iter/inc/tensor_dim.h
@@ -0,0 +1,13 @@
+ /* Header file generated by RedMulE Golden Model */
+#ifndef __TENSOR_DIM__
+#define __TENSOR_DIM__
+
+#define M_SIZE  24 
+#define N_SIZE  32 
+#define K_SIZE  32
+#define SRC_FMT FP16
+#define DST_FMT FP16
+#define FPFORMAT 16
+uint8_t gemm_ops = GEMM; 
+
+#endif
diff --git a/hwpe/redmule_256iter/inc/w_2D.h b/hwpe/redmule_256iter/inc/w_2D.h
new file mode 100644
index 0000000..9409c64
--- /dev/null
+++ b/hwpe/redmule_256iter/inc/w_2D.h
@@ -0,0 +1,35 @@
+ /* Header file generated by RedMulE Golden Model */
+uint16_t w_inp_2D [32][32] = {
+0x311a, 0x39e0, 0x387d, 0x3a4a, 0x386f, 0x3ada, 0x392f, 0x3854, 0x3014, 0x2fd2, 0x31c9, 0x2fca, 0x2e55, 0x3bc8, 0x396d, 0x3b1d, 0x39f6, 0x333a, 0x3908, 0x3628, 0x3bab, 0x3b8b, 0x3b4a, 0x322d, 0x3925, 0x317a, 0x3725, 0x31c2, 0x3066, 0x38f3, 0x3a17, 0x3476, 
+0x3bda, 0x3196, 0x3922, 0x3680, 0x396a, 0x3021, 0x3761, 0x374d, 0x2fc2, 0x3967, 0x3b94, 0x33b5, 0x3797, 0x34d6, 0x3655, 0x2176, 0x39bc, 0x3999, 0x3658, 0x3904, 0x3759, 0x2ade, 0x3a5a, 0x3b78, 0x36c7, 0x2d01, 0x3b58, 0x2d9a, 0x373d, 0x3952, 0x38e8, 0x3887, 
+0x37b6, 0x3a88, 0x2f8a, 0x2d79, 0x3413, 0x3421, 0x3976, 0x32b2, 0x3446, 0x2d99, 0x3a56, 0x3322, 0x3b49, 0x39fa, 0x3acd, 0x3af6, 0x304c, 0x3abb, 0x3a83, 0x38b2, 0x3ab9, 0x363e, 0x389f, 0x31bb, 0x38e1, 0x3bc4, 0x3b9b, 0x2984, 0x3a43, 0x3b2f, 0x35d6, 0x3bda, 
+0x2df3, 0x3bf8, 0x2acc, 0x378b, 0x3555, 0x2e59, 0x31d4, 0x34ec, 0x3a46, 0x3bab, 0x3214, 0x3161, 0x3470, 0x3a03, 0x368e, 0x31ad, 0x27cb, 0x2ecb, 0x3422, 0x39f7, 0x3644, 0x3a77, 0x313f, 0x34f2, 0x39b3, 0x3bf2, 0x379a, 0x3456, 0x35fe, 0x3ae7, 0x3964, 0x385f, 
+0x3b16, 0x3999, 0x3833, 0x2eda, 0x3afd, 0x3a4a, 0x3ba2, 0x2bd4, 0x3b38, 0x31a2, 0x32dd, 0x353c, 0x366f, 0x375e, 0x3821, 0x367a, 0x3b44, 0x39e6, 0x3787, 0x339e, 0x39d7, 0x38c6, 0x37d5, 0x342f, 0x3984, 0x319b, 0x33b5, 0x35ab, 0x398a, 0x374e, 0x36b6, 0x3b21, 
+0x3bbb, 0x2ab3, 0x2ad5, 0x33bc, 0x2bef, 0x3780, 0x3738, 0x3a0b, 0x3b09, 0x30ca, 0x384e, 0x3ab3, 0x39bd, 0x3453, 0x3a6d, 0x3957, 0x2c10, 0x30e9, 0x35d4, 0x3aef, 0x3be9, 0x39ad, 0x3a74, 0x3af9, 0x3739, 0x2d4d, 0x39fe, 0x3b72, 0x2c57, 0x398c, 0x381f, 0x3930, 
+0x3820, 0x321b, 0x3964, 0x2964, 0x33a0, 0x2d00, 0x2490, 0x336b, 0x3465, 0x3b2e, 0x3aa0, 0x371f, 0x300e, 0x3a09, 0x3bf1, 0x25cc, 0x3b6f, 0x3384, 0x3a88, 0x3acb, 0x3814, 0x36d0, 0x3081, 0x3a2c, 0x3353, 0x39cb, 0x31ed, 0x3af6, 0x3721, 0x36c7, 0x2ce2, 0x390d, 
+0x3698, 0x3ab2, 0x3b3e, 0x2eb4, 0x3998, 0x39e3, 0x3a77, 0x3632, 0x2c12, 0x3bd5, 0x3ba3, 0x3bba, 0x323c, 0x367b, 0x3557, 0x39c8, 0x37db, 0x3b45, 0x3b6e, 0x3931, 0x3121, 0x3a8d, 0x3a55, 0x3b9b, 0x358a, 0x3925, 0x3491, 0x3912, 0x3b6b, 0x3584, 0x32df, 0x3120, 
+0x32b2, 0x3b0a, 0x2cad, 0x3465, 0x3ad3, 0x3bcd, 0x363b, 0x3afe, 0x354b, 0x3374, 0x39af, 0x3b7f, 0x308c, 0x2e72, 0x3380, 0x3b70, 0x3902, 0x38d8, 0x39f3, 0x3a4b, 0x3853, 0x397b, 0x2ebe, 0x387f, 0x2845, 0x37e2, 0x360f, 0x370b, 0x3acb, 0x35d4, 0x36e6, 0x3262, 
+0x2e88, 0x3a54, 0x2ee3, 0x3575, 0x3afe, 0x2aee, 0x39a0, 0x3aae, 0x3693, 0x3432, 0x3834, 0x3b9b, 0x3bcb, 0x2e3a, 0x356d, 0x374e, 0x3924, 0x383c, 0x311e, 0x3ac5, 0x352d, 0x311e, 0x38ca, 0x34d4, 0x36ca, 0x34ed, 0x3a13, 0x33eb, 0x3639, 0x3828, 0x3b3c, 0x3939, 
+0x3837, 0x3521, 0x2cb5, 0x3629, 0x3924, 0x384c, 0x366a, 0x3bbf, 0x2e9e, 0x3ba8, 0x33ad, 0x38c8, 0x3934, 0x3907, 0x249a, 0x3690, 0x3a09, 0x3215, 0x3898, 0x325d, 0x37d5, 0x3195, 0x361c, 0x3ae4, 0x351f, 0x3452, 0x3bc0, 0x375c, 0x39bf, 0x317a, 0x3aae, 0x283a, 
+0x3476, 0x3b92, 0x3472, 0x383e, 0x280f, 0x39d6, 0x2fd1, 0x31f4, 0x2ffb, 0x3b97, 0x3692, 0x36c0, 0x3989, 0x33cf, 0x3ba6, 0x3239, 0x35d7, 0x33ab, 0x31eb, 0x3b47, 0x389b, 0x3b88, 0x3580, 0x354c, 0x3802, 0x3b9a, 0x3b94, 0x2a92, 0x2db1, 0x38bd, 0x2dfb, 0x3900, 
+0x344f, 0x3739, 0x27a5, 0x3b2e, 0x342b, 0x34bb, 0x30c8, 0x3ae8, 0x3b26, 0x3982, 0x38c0, 0x3408, 0x38c8, 0x36ef, 0x3bf0, 0x3acf, 0x3a3c, 0x3825, 0x31a5, 0x3ada, 0x3b5b, 0x37db, 0x3a01, 0x3663, 0x3a7d, 0x327b, 0x3a1f, 0x3862, 0x38af, 0x3204, 0x372e, 0x3b19, 
+0x3708, 0x3622, 0x2e62, 0x39ab, 0x2d4d, 0x31b4, 0x3552, 0x3bbc, 0x36f2, 0x36eb, 0x38ef, 0x3755, 0x3bbe, 0x2c17, 0x3815, 0x2f53, 0x363f, 0x38c1, 0x3246, 0x386b, 0x34de, 0x34e4, 0x3baa, 0x349e, 0x32ce, 0x3a68, 0x373f, 0x2cce, 0x3b36, 0x28ba, 0x3b50, 0x3232, 
+0x1f34, 0x3928, 0x35cd, 0x3b38, 0x30ce, 0x35a1, 0x3a06, 0x3a32, 0x3a53, 0x3489, 0x3241, 0x372f, 0x390c, 0x3a1b, 0x378a, 0x3713, 0x3769, 0x37a8, 0x3418, 0x3ad4, 0x3a4e, 0x3bf7, 0x37a5, 0x34dc, 0x39b2, 0x351b, 0x3372, 0x349f, 0x2f50, 0x3ab1, 0x3795, 0x2db7, 
+0x3864, 0x3157, 0x3900, 0x323e, 0x389e, 0x3880, 0x3b1f, 0x37a1, 0x396c, 0x2e43, 0x2c2a, 0x3b78, 0x3988, 0x3a14, 0x39c1, 0x3b51, 0x3780, 0x3bf2, 0x2d19, 0x3815, 0x3a5f, 0x3641, 0x2f62, 0x37d5, 0x3564, 0x139a, 0x3ab8, 0x28f7, 0x3785, 0x34e1, 0x3097, 0x3768, 
+0x3971, 0x3ae2, 0x32ae, 0x2fd5, 0x382a, 0x346c, 0x3133, 0x3167, 0x3940, 0x2d12, 0x389a, 0x3bd0, 0x3943, 0x391c, 0x3a75, 0x2a11, 0x391e, 0x372d, 0x3a79, 0x3b72, 0x3373, 0x39b7, 0x35d7, 0x372b, 0x3a6d, 0x38a1, 0x3279, 0x3434, 0x3694, 0x3b45, 0x3abb, 0x392d, 
+0x34a8, 0x3757, 0x32ca, 0x345d, 0x36a5, 0x3854, 0x2dcd, 0x30af, 0x38dd, 0x3067, 0x3411, 0x3997, 0x397a, 0x3a64, 0x38b8, 0x3962, 0x3509, 0x3bb6, 0x3a66, 0x339f, 0x372a, 0x31a8, 0x37da, 0x36ff, 0x33c6, 0x31da, 0x3977, 0x3b72, 0x3841, 0x3567, 0x3433, 0x33b8, 
+0x39fe, 0x3a10, 0x3bf2, 0x35e7, 0x3a4a, 0x3b3e, 0x2ec7, 0x3aa4, 0x3846, 0x3af9, 0x38a9, 0x2c1f, 0x39ab, 0x349f, 0x31d6, 0x39ae, 0x3b79, 0x352d, 0x3516, 0x347c, 0x2f33, 0x35ad, 0x31c4, 0x3b52, 0x354b, 0x3786, 0x3ab7, 0x3896, 0x34ac, 0x352f, 0x37e6, 0x326a, 
+0x2e44, 0x34c7, 0x388d, 0x3bf4, 0x363f, 0x3b3d, 0x33b1, 0x3b8b, 0x3340, 0x37f7, 0x3b07, 0x25bf, 0x398e, 0x3505, 0x3bd7, 0x366d, 0x388a, 0x2cc0, 0x359a, 0x3b9a, 0x3b99, 0x379d, 0x3b6b, 0x39b8, 0x3223, 0x2703, 0x3ba9, 0x2ecb, 0x3759, 0x39d8, 0x37ac, 0x32cf, 
+0x35f2, 0x38a3, 0x399e, 0x3bd2, 0x3780, 0x3af3, 0x3b5e, 0x337b, 0x3a08, 0x35da, 0x3446, 0x3b25, 0x3ad0, 0x3bee, 0x3141, 0x32d8, 0x34ce, 0x2ac9, 0x3800, 0x3a8a, 0x2d53, 0x368a, 0x3561, 0x3998, 0x35a3, 0x3677, 0x3ab2, 0x3269, 0x3236, 0x3b3e, 0x3aba, 0x3bac, 
+0x395d, 0x3820, 0x1df6, 0x3bb5, 0x35b5, 0x3675, 0x3b74, 0x360f, 0x34de, 0x3a0c, 0x3aeb, 0x299d, 0x3207, 0x3bd8, 0x2178, 0x3995, 0x3948, 0x3908, 0x3843, 0x2ea5, 0x3045, 0x3989, 0x345d, 0x39c5, 0x3a89, 0x3863, 0x3be0, 0x397a, 0x38f1, 0x39e2, 0x3b08, 0x352e, 
+0x385f, 0x28f2, 0x3bc3, 0x35e0, 0x380c, 0x3b9c, 0x3afc, 0x390a, 0x3689, 0x34fd, 0x2cf5, 0x308e, 0x342b, 0x3921, 0x3a67, 0x3ad6, 0x2986, 0x32fc, 0x35aa, 0x3507, 0x3608, 0x33fd, 0x3bf3, 0x39e2, 0x3b0f, 0x30b7, 0x3896, 0x3ae4, 0x2145, 0x35b6, 0x2e1d, 0x3ad1, 
+0x333d, 0x3afb, 0x2703, 0x3413, 0x1d7d, 0x3b7f, 0x3ae1, 0x303c, 0x3004, 0x39d3, 0x3554, 0x31a4, 0x354e, 0x3662, 0x39c5, 0x2eb7, 0x2c6e, 0x397f, 0x31d8, 0x1f0c, 0x38e3, 0x35f0, 0x2714, 0x28d1, 0x375e, 0x3a75, 0x3830, 0x3578, 0x397d, 0x3b18, 0x383c, 0x3498, 
+0x39ad, 0x3598, 0x23c4, 0x34ea, 0x3a61, 0x2b00, 0x3707, 0x3ae1, 0x37ae, 0x389d, 0x37fa, 0x3673, 0x3278, 0xf3e, 0x3809, 0x33c6, 0x3bf5, 0x3279, 0x3816, 0x360c, 0x39c8, 0x381f, 0x3741, 0x2d66, 0x38c0, 0x37d3, 0x377a, 0x3621, 0x2faf, 0x392e, 0x2de6, 0x33c5, 
+0x3803, 0x2600, 0x32e9, 0x39b4, 0x38d2, 0x34e8, 0x2fe6, 0x3199, 0x3643, 0x3a77, 0x27cc, 0x39d7, 0x34c6, 0x2ea8, 0x364e, 0x3b07, 0x31c7, 0x30a1, 0x31b1, 0x3b8f, 0x3571, 0x3b75, 0x3989, 0x3805, 0x39fb, 0x3945, 0x352b, 0x31d8, 0x3904, 0x3440, 0x3a57, 0x2cf7, 
+0x3b39, 0x2fcd, 0x2b89, 0x2edd, 0x3682, 0x36a9, 0x32c8, 0x37ac, 0x32a5, 0x3311, 0x394b, 0x3b84, 0x3aec, 0x3601, 0x2765, 0x3b69, 0x396b, 0x3727, 0x3bfe, 0x3907, 0x376f, 0x3674, 0x3973, 0x3671, 0x3491, 0x3993, 0x383f, 0x3335, 0x3989, 0x3550, 0x3077, 0x35f5, 
+0x3a59, 0x3950, 0x380c, 0x37cd, 0x30bf, 0x3607, 0x3afa, 0x3b5d, 0x32b9, 0x386b, 0x35bd, 0x3aca, 0x3ba5, 0x3b2d, 0x3b19, 0x3b8b, 0x345e, 0x2845, 0x34aa, 0x372a, 0x3448, 0x34f5, 0x3ae2, 0x3637, 0x2cb5, 0x354b, 0x3b15, 0x2ca8, 0x2641, 0x3178, 0x2cfe, 0x39b4, 
+0x3bdd, 0x3acb, 0x3a05, 0x38a2, 0x3b4a, 0x34e5, 0x395f, 0x394b, 0x34c4, 0x3aa5, 0x29bb, 0x2d96, 0x339d, 0x387c, 0x382e, 0x385a, 0x396b, 0x3aa9, 0x2f1e, 0x33a7, 0x3b90, 0x3b7b, 0x3b5f, 0x39d3, 0x3b18, 0x354f, 0x2cdb, 0x3a6f, 0x3434, 0x34ff, 0x3a5b, 0x3b84, 
+0x3a33, 0x384b, 0x2e67, 0x3b85, 0x3853, 0x380c, 0x346a, 0x3aaa, 0x3492, 0x33e8, 0x3bf2, 0x38ae, 0x3a29, 0x3830, 0x3221, 0x35b1, 0x3a48, 0x2c68, 0x2ced, 0x3a7e, 0x3539, 0x3922, 0x374c, 0x3aaa, 0x2dae, 0x395d, 0x3b3d, 0x3890, 0x2cfe, 0x2dd6, 0x3bad, 0x33c5, 
+0x2c07, 0x3a2c, 0x37a8, 0x390f, 0x2fc8, 0x35ae, 0x388c, 0x30ee, 0x3674, 0x391d, 0x3bfc, 0x36bf, 0x322d, 0x3a78, 0x35c0, 0x3492, 0x3ac8, 0x3504, 0x3315, 0x381d, 0x3a7a, 0x3a08, 0x343c, 0x3bda, 0x341b, 0x39f0, 0x3b9e, 0x395d, 0x3c00, 0x38ab, 0x3bcf, 0x3564, 
+0x33c4, 0x3b0d, 0x3623, 0x33b9, 0x3b92, 0x1e71, 0x2c57, 0x36d0, 0x314b, 0x3a16, 0x3372, 0x341b, 0x3aaa, 0x3444, 0x396b, 0x2dd7, 0x3b30, 0x3559, 0x3b5b, 0x3a29, 0x2d19, 0x38b7, 0x3b01, 0x3afa, 0x398a, 0x3839, 0x3ac9, 0x2e31, 0x3924, 0x39f2, 0x3a7f, 0x3285 
+};
\ No newline at end of file
diff --git a/hwpe/redmule_256iter/inc/w_input.h b/hwpe/redmule_256iter/inc/w_input.h
new file mode 100644
index 0000000..dc4d3be
--- /dev/null
+++ b/hwpe/redmule_256iter/inc/w_input.h
@@ -0,0 +1,35 @@
+ /* Header file generated by RedMulE Golden Model */
+uint16_t w_inp [1024] = {
+0x311a, 0x39e0, 0x387d, 0x3a4a, 0x386f, 0x3ada, 0x392f, 0x3854, 0x3014, 0x2fd2, 0x31c9, 0x2fca, 0x2e55, 0x3bc8, 0x396d, 0x3b1d, 0x39f6, 0x333a, 0x3908, 0x3628, 0x3bab, 0x3b8b, 0x3b4a, 0x322d, 0x3925, 0x317a, 0x3725, 0x31c2, 0x3066, 0x38f3, 0x3a17, 0x3476, 
+0x3bda, 0x3196, 0x3922, 0x3680, 0x396a, 0x3021, 0x3761, 0x374d, 0x2fc2, 0x3967, 0x3b94, 0x33b5, 0x3797, 0x34d6, 0x3655, 0x2176, 0x39bc, 0x3999, 0x3658, 0x3904, 0x3759, 0x2ade, 0x3a5a, 0x3b78, 0x36c7, 0x2d01, 0x3b58, 0x2d9a, 0x373d, 0x3952, 0x38e8, 0x3887, 
+0x37b6, 0x3a88, 0x2f8a, 0x2d79, 0x3413, 0x3421, 0x3976, 0x32b2, 0x3446, 0x2d99, 0x3a56, 0x3322, 0x3b49, 0x39fa, 0x3acd, 0x3af6, 0x304c, 0x3abb, 0x3a83, 0x38b2, 0x3ab9, 0x363e, 0x389f, 0x31bb, 0x38e1, 0x3bc4, 0x3b9b, 0x2984, 0x3a43, 0x3b2f, 0x35d6, 0x3bda, 
+0x2df3, 0x3bf8, 0x2acc, 0x378b, 0x3555, 0x2e59, 0x31d4, 0x34ec, 0x3a46, 0x3bab, 0x3214, 0x3161, 0x3470, 0x3a03, 0x368e, 0x31ad, 0x27cb, 0x2ecb, 0x3422, 0x39f7, 0x3644, 0x3a77, 0x313f, 0x34f2, 0x39b3, 0x3bf2, 0x379a, 0x3456, 0x35fe, 0x3ae7, 0x3964, 0x385f, 
+0x3b16, 0x3999, 0x3833, 0x2eda, 0x3afd, 0x3a4a, 0x3ba2, 0x2bd4, 0x3b38, 0x31a2, 0x32dd, 0x353c, 0x366f, 0x375e, 0x3821, 0x367a, 0x3b44, 0x39e6, 0x3787, 0x339e, 0x39d7, 0x38c6, 0x37d5, 0x342f, 0x3984, 0x319b, 0x33b5, 0x35ab, 0x398a, 0x374e, 0x36b6, 0x3b21, 
+0x3bbb, 0x2ab3, 0x2ad5, 0x33bc, 0x2bef, 0x3780, 0x3738, 0x3a0b, 0x3b09, 0x30ca, 0x384e, 0x3ab3, 0x39bd, 0x3453, 0x3a6d, 0x3957, 0x2c10, 0x30e9, 0x35d4, 0x3aef, 0x3be9, 0x39ad, 0x3a74, 0x3af9, 0x3739, 0x2d4d, 0x39fe, 0x3b72, 0x2c57, 0x398c, 0x381f, 0x3930, 
+0x3820, 0x321b, 0x3964, 0x2964, 0x33a0, 0x2d00, 0x2490, 0x336b, 0x3465, 0x3b2e, 0x3aa0, 0x371f, 0x300e, 0x3a09, 0x3bf1, 0x25cc, 0x3b6f, 0x3384, 0x3a88, 0x3acb, 0x3814, 0x36d0, 0x3081, 0x3a2c, 0x3353, 0x39cb, 0x31ed, 0x3af6, 0x3721, 0x36c7, 0x2ce2, 0x390d, 
+0x3698, 0x3ab2, 0x3b3e, 0x2eb4, 0x3998, 0x39e3, 0x3a77, 0x3632, 0x2c12, 0x3bd5, 0x3ba3, 0x3bba, 0x323c, 0x367b, 0x3557, 0x39c8, 0x37db, 0x3b45, 0x3b6e, 0x3931, 0x3121, 0x3a8d, 0x3a55, 0x3b9b, 0x358a, 0x3925, 0x3491, 0x3912, 0x3b6b, 0x3584, 0x32df, 0x3120, 
+0x32b2, 0x3b0a, 0x2cad, 0x3465, 0x3ad3, 0x3bcd, 0x363b, 0x3afe, 0x354b, 0x3374, 0x39af, 0x3b7f, 0x308c, 0x2e72, 0x3380, 0x3b70, 0x3902, 0x38d8, 0x39f3, 0x3a4b, 0x3853, 0x397b, 0x2ebe, 0x387f, 0x2845, 0x37e2, 0x360f, 0x370b, 0x3acb, 0x35d4, 0x36e6, 0x3262, 
+0x2e88, 0x3a54, 0x2ee3, 0x3575, 0x3afe, 0x2aee, 0x39a0, 0x3aae, 0x3693, 0x3432, 0x3834, 0x3b9b, 0x3bcb, 0x2e3a, 0x356d, 0x374e, 0x3924, 0x383c, 0x311e, 0x3ac5, 0x352d, 0x311e, 0x38ca, 0x34d4, 0x36ca, 0x34ed, 0x3a13, 0x33eb, 0x3639, 0x3828, 0x3b3c, 0x3939, 
+0x3837, 0x3521, 0x2cb5, 0x3629, 0x3924, 0x384c, 0x366a, 0x3bbf, 0x2e9e, 0x3ba8, 0x33ad, 0x38c8, 0x3934, 0x3907, 0x249a, 0x3690, 0x3a09, 0x3215, 0x3898, 0x325d, 0x37d5, 0x3195, 0x361c, 0x3ae4, 0x351f, 0x3452, 0x3bc0, 0x375c, 0x39bf, 0x317a, 0x3aae, 0x283a, 
+0x3476, 0x3b92, 0x3472, 0x383e, 0x280f, 0x39d6, 0x2fd1, 0x31f4, 0x2ffb, 0x3b97, 0x3692, 0x36c0, 0x3989, 0x33cf, 0x3ba6, 0x3239, 0x35d7, 0x33ab, 0x31eb, 0x3b47, 0x389b, 0x3b88, 0x3580, 0x354c, 0x3802, 0x3b9a, 0x3b94, 0x2a92, 0x2db1, 0x38bd, 0x2dfb, 0x3900, 
+0x344f, 0x3739, 0x27a5, 0x3b2e, 0x342b, 0x34bb, 0x30c8, 0x3ae8, 0x3b26, 0x3982, 0x38c0, 0x3408, 0x38c8, 0x36ef, 0x3bf0, 0x3acf, 0x3a3c, 0x3825, 0x31a5, 0x3ada, 0x3b5b, 0x37db, 0x3a01, 0x3663, 0x3a7d, 0x327b, 0x3a1f, 0x3862, 0x38af, 0x3204, 0x372e, 0x3b19, 
+0x3708, 0x3622, 0x2e62, 0x39ab, 0x2d4d, 0x31b4, 0x3552, 0x3bbc, 0x36f2, 0x36eb, 0x38ef, 0x3755, 0x3bbe, 0x2c17, 0x3815, 0x2f53, 0x363f, 0x38c1, 0x3246, 0x386b, 0x34de, 0x34e4, 0x3baa, 0x349e, 0x32ce, 0x3a68, 0x373f, 0x2cce, 0x3b36, 0x28ba, 0x3b50, 0x3232, 
+0x1f34, 0x3928, 0x35cd, 0x3b38, 0x30ce, 0x35a1, 0x3a06, 0x3a32, 0x3a53, 0x3489, 0x3241, 0x372f, 0x390c, 0x3a1b, 0x378a, 0x3713, 0x3769, 0x37a8, 0x3418, 0x3ad4, 0x3a4e, 0x3bf7, 0x37a5, 0x34dc, 0x39b2, 0x351b, 0x3372, 0x349f, 0x2f50, 0x3ab1, 0x3795, 0x2db7, 
+0x3864, 0x3157, 0x3900, 0x323e, 0x389e, 0x3880, 0x3b1f, 0x37a1, 0x396c, 0x2e43, 0x2c2a, 0x3b78, 0x3988, 0x3a14, 0x39c1, 0x3b51, 0x3780, 0x3bf2, 0x2d19, 0x3815, 0x3a5f, 0x3641, 0x2f62, 0x37d5, 0x3564, 0x139a, 0x3ab8, 0x28f7, 0x3785, 0x34e1, 0x3097, 0x3768, 
+0x3971, 0x3ae2, 0x32ae, 0x2fd5, 0x382a, 0x346c, 0x3133, 0x3167, 0x3940, 0x2d12, 0x389a, 0x3bd0, 0x3943, 0x391c, 0x3a75, 0x2a11, 0x391e, 0x372d, 0x3a79, 0x3b72, 0x3373, 0x39b7, 0x35d7, 0x372b, 0x3a6d, 0x38a1, 0x3279, 0x3434, 0x3694, 0x3b45, 0x3abb, 0x392d, 
+0x34a8, 0x3757, 0x32ca, 0x345d, 0x36a5, 0x3854, 0x2dcd, 0x30af, 0x38dd, 0x3067, 0x3411, 0x3997, 0x397a, 0x3a64, 0x38b8, 0x3962, 0x3509, 0x3bb6, 0x3a66, 0x339f, 0x372a, 0x31a8, 0x37da, 0x36ff, 0x33c6, 0x31da, 0x3977, 0x3b72, 0x3841, 0x3567, 0x3433, 0x33b8, 
+0x39fe, 0x3a10, 0x3bf2, 0x35e7, 0x3a4a, 0x3b3e, 0x2ec7, 0x3aa4, 0x3846, 0x3af9, 0x38a9, 0x2c1f, 0x39ab, 0x349f, 0x31d6, 0x39ae, 0x3b79, 0x352d, 0x3516, 0x347c, 0x2f33, 0x35ad, 0x31c4, 0x3b52, 0x354b, 0x3786, 0x3ab7, 0x3896, 0x34ac, 0x352f, 0x37e6, 0x326a, 
+0x2e44, 0x34c7, 0x388d, 0x3bf4, 0x363f, 0x3b3d, 0x33b1, 0x3b8b, 0x3340, 0x37f7, 0x3b07, 0x25bf, 0x398e, 0x3505, 0x3bd7, 0x366d, 0x388a, 0x2cc0, 0x359a, 0x3b9a, 0x3b99, 0x379d, 0x3b6b, 0x39b8, 0x3223, 0x2703, 0x3ba9, 0x2ecb, 0x3759, 0x39d8, 0x37ac, 0x32cf, 
+0x35f2, 0x38a3, 0x399e, 0x3bd2, 0x3780, 0x3af3, 0x3b5e, 0x337b, 0x3a08, 0x35da, 0x3446, 0x3b25, 0x3ad0, 0x3bee, 0x3141, 0x32d8, 0x34ce, 0x2ac9, 0x3800, 0x3a8a, 0x2d53, 0x368a, 0x3561, 0x3998, 0x35a3, 0x3677, 0x3ab2, 0x3269, 0x3236, 0x3b3e, 0x3aba, 0x3bac, 
+0x395d, 0x3820, 0x1df6, 0x3bb5, 0x35b5, 0x3675, 0x3b74, 0x360f, 0x34de, 0x3a0c, 0x3aeb, 0x299d, 0x3207, 0x3bd8, 0x2178, 0x3995, 0x3948, 0x3908, 0x3843, 0x2ea5, 0x3045, 0x3989, 0x345d, 0x39c5, 0x3a89, 0x3863, 0x3be0, 0x397a, 0x38f1, 0x39e2, 0x3b08, 0x352e, 
+0x385f, 0x28f2, 0x3bc3, 0x35e0, 0x380c, 0x3b9c, 0x3afc, 0x390a, 0x3689, 0x34fd, 0x2cf5, 0x308e, 0x342b, 0x3921, 0x3a67, 0x3ad6, 0x2986, 0x32fc, 0x35aa, 0x3507, 0x3608, 0x33fd, 0x3bf3, 0x39e2, 0x3b0f, 0x30b7, 0x3896, 0x3ae4, 0x2145, 0x35b6, 0x2e1d, 0x3ad1, 
+0x333d, 0x3afb, 0x2703, 0x3413, 0x1d7d, 0x3b7f, 0x3ae1, 0x303c, 0x3004, 0x39d3, 0x3554, 0x31a4, 0x354e, 0x3662, 0x39c5, 0x2eb7, 0x2c6e, 0x397f, 0x31d8, 0x1f0c, 0x38e3, 0x35f0, 0x2714, 0x28d1, 0x375e, 0x3a75, 0x3830, 0x3578, 0x397d, 0x3b18, 0x383c, 0x3498, 
+0x39ad, 0x3598, 0x23c4, 0x34ea, 0x3a61, 0x2b00, 0x3707, 0x3ae1, 0x37ae, 0x389d, 0x37fa, 0x3673, 0x3278, 0xf3e, 0x3809, 0x33c6, 0x3bf5, 0x3279, 0x3816, 0x360c, 0x39c8, 0x381f, 0x3741, 0x2d66, 0x38c0, 0x37d3, 0x377a, 0x3621, 0x2faf, 0x392e, 0x2de6, 0x33c5, 
+0x3803, 0x2600, 0x32e9, 0x39b4, 0x38d2, 0x34e8, 0x2fe6, 0x3199, 0x3643, 0x3a77, 0x27cc, 0x39d7, 0x34c6, 0x2ea8, 0x364e, 0x3b07, 0x31c7, 0x30a1, 0x31b1, 0x3b8f, 0x3571, 0x3b75, 0x3989, 0x3805, 0x39fb, 0x3945, 0x352b, 0x31d8, 0x3904, 0x3440, 0x3a57, 0x2cf7, 
+0x3b39, 0x2fcd, 0x2b89, 0x2edd, 0x3682, 0x36a9, 0x32c8, 0x37ac, 0x32a5, 0x3311, 0x394b, 0x3b84, 0x3aec, 0x3601, 0x2765, 0x3b69, 0x396b, 0x3727, 0x3bfe, 0x3907, 0x376f, 0x3674, 0x3973, 0x3671, 0x3491, 0x3993, 0x383f, 0x3335, 0x3989, 0x3550, 0x3077, 0x35f5, 
+0x3a59, 0x3950, 0x380c, 0x37cd, 0x30bf, 0x3607, 0x3afa, 0x3b5d, 0x32b9, 0x386b, 0x35bd, 0x3aca, 0x3ba5, 0x3b2d, 0x3b19, 0x3b8b, 0x345e, 0x2845, 0x34aa, 0x372a, 0x3448, 0x34f5, 0x3ae2, 0x3637, 0x2cb5, 0x354b, 0x3b15, 0x2ca8, 0x2641, 0x3178, 0x2cfe, 0x39b4, 
+0x3bdd, 0x3acb, 0x3a05, 0x38a2, 0x3b4a, 0x34e5, 0x395f, 0x394b, 0x34c4, 0x3aa5, 0x29bb, 0x2d96, 0x339d, 0x387c, 0x382e, 0x385a, 0x396b, 0x3aa9, 0x2f1e, 0x33a7, 0x3b90, 0x3b7b, 0x3b5f, 0x39d3, 0x3b18, 0x354f, 0x2cdb, 0x3a6f, 0x3434, 0x34ff, 0x3a5b, 0x3b84, 
+0x3a33, 0x384b, 0x2e67, 0x3b85, 0x3853, 0x380c, 0x346a, 0x3aaa, 0x3492, 0x33e8, 0x3bf2, 0x38ae, 0x3a29, 0x3830, 0x3221, 0x35b1, 0x3a48, 0x2c68, 0x2ced, 0x3a7e, 0x3539, 0x3922, 0x374c, 0x3aaa, 0x2dae, 0x395d, 0x3b3d, 0x3890, 0x2cfe, 0x2dd6, 0x3bad, 0x33c5, 
+0x2c07, 0x3a2c, 0x37a8, 0x390f, 0x2fc8, 0x35ae, 0x388c, 0x30ee, 0x3674, 0x391d, 0x3bfc, 0x36bf, 0x322d, 0x3a78, 0x35c0, 0x3492, 0x3ac8, 0x3504, 0x3315, 0x381d, 0x3a7a, 0x3a08, 0x343c, 0x3bda, 0x341b, 0x39f0, 0x3b9e, 0x395d, 0x3c00, 0x38ab, 0x3bcf, 0x3564, 
+0x33c4, 0x3b0d, 0x3623, 0x33b9, 0x3b92, 0x1e71, 0x2c57, 0x36d0, 0x314b, 0x3a16, 0x3372, 0x341b, 0x3aaa, 0x3444, 0x396b, 0x2dd7, 0x3b30, 0x3559, 0x3b5b, 0x3a29, 0x2d19, 0x38b7, 0x3b01, 0x3afa, 0x398a, 0x3839, 0x3ac9, 0x2e31, 0x3924, 0x39f2, 0x3a7f, 0x3285 
+};
\ No newline at end of file
diff --git a/hwpe/redmule_256iter/inc/x_2D.h b/hwpe/redmule_256iter/inc/x_2D.h
new file mode 100644
index 0000000..0b589f8
--- /dev/null
+++ b/hwpe/redmule_256iter/inc/x_2D.h
@@ -0,0 +1,27 @@
+ /* Header file generated by RedMulE Golden Model */
+uint16_t x_inp_2D [24][32] = {
+0x2153, 0x3bb5, 0x3896, 0x365f, 0x2483, 0x3518, 0x2dd1, 0x3bca, 0x397b, 0x29b1, 0x3705, 0x36c8, 0x398b, 0x3661, 0x2f05, 0x365a, 0x3bf9, 0x34df, 0x363b, 0x38d9, 0x39c6, 0x3abb, 0x3952, 0x38f2, 0x392d, 0x3b3e, 0x2afb, 0x3a9d, 0x353b, 0x3b73, 0x3a01, 0x3679, 
+0x3934, 0x397d, 0x2904, 0x3822, 0x3462, 0x3b44, 0x39e9, 0x28be, 0x331e, 0x3a1d, 0x39e5, 0x34da, 0x3a19, 0x3906, 0x1d35, 0x3871, 0x31e7, 0x3b29, 0x325d, 0x3797, 0x2b2f, 0x38b4, 0x232f, 0x38aa, 0x3aca, 0x316f, 0x3811, 0x3950, 0x32ea, 0x3bc7, 0x382c, 0x38a2, 
+0x29ce, 0x3afa, 0x3a39, 0x2ccc, 0x39fd, 0x3b3d, 0x384a, 0x3a35, 0x3802, 0x366a, 0x37ec, 0x3598, 0x3bf8, 0x3a85, 0x3a1b, 0x386e, 0x3b4c, 0x39de, 0x38c2, 0x2f93, 0x3b4c, 0x39c4, 0x3b9e, 0x3844, 0x346d, 0x3bff, 0x32ce, 0x296d, 0x3130, 0x3b3d, 0x3b44, 0x369d, 
+0x3b13, 0x31ed, 0x330a, 0x3831, 0x34e7, 0x37b3, 0x331a, 0x3918, 0x32d3, 0x3995, 0x3991, 0x3919, 0x3a26, 0x385b, 0x2b76, 0x3a3b, 0x37f2, 0x26a7, 0x3225, 0x3b64, 0x28f0, 0x3456, 0x3822, 0x341e, 0x381a, 0x38d8, 0x2c11, 0x33be, 0x33ac, 0x353f, 0x3476, 0x3abc, 
+0x36ec, 0x3a1d, 0x39d3, 0x3821, 0x36ac, 0x3bce, 0x3ad2, 0x3616, 0x36a1, 0x2cb3, 0x38d2, 0x314f, 0x385c, 0x3b63, 0x3bb6, 0x2951, 0x372d, 0x2c42, 0x3823, 0x3883, 0x3872, 0x31ee, 0x36c5, 0x399a, 0x31b0, 0x3887, 0x3884, 0x3865, 0x3896, 0x36c3, 0x32e3, 0x346c, 
+0x3935, 0x3b50, 0x2b6d, 0x38cd, 0x388f, 0x3389, 0x395d, 0x31cd, 0x2efd, 0x3154, 0x2f35, 0x3444, 0x3293, 0x3b6b, 0x1bec, 0x3b69, 0x3bf3, 0x3611, 0x3508, 0x3742, 0x3a50, 0x3ab7, 0x3457, 0x38d3, 0x3344, 0x38e8, 0x33c0, 0x3668, 0x3bee, 0x3b21, 0x3727, 0x3121, 
+0x316c, 0x3288, 0x2d50, 0x2e74, 0x35d5, 0x37e2, 0x303d, 0x36af, 0x341f, 0x3436, 0x2df7, 0x399d, 0x30f4, 0x3aaf, 0x34e4, 0x2c2a, 0x3116, 0x34d3, 0x36ac, 0x35e3, 0x3760, 0x36e1, 0x3ad2, 0x3547, 0x38f4, 0x369c, 0x3ba9, 0x34f0, 0x3a39, 0x3b19, 0x36e6, 0x395d, 
+0x3be8, 0x3293, 0x3bfc, 0x3435, 0x2eb3, 0x3360, 0x3919, 0x3bed, 0x396a, 0x37fc, 0x3242, 0x384b, 0x38cb, 0x3b2c, 0x3b28, 0x28cf, 0x3828, 0x3855, 0x3ba9, 0x2fa7, 0x340b, 0x32f1, 0x3ada, 0x36fa, 0x31f5, 0x3436, 0x29d0, 0x33e6, 0x3232, 0x3bec, 0x3904, 0x2797, 
+0x3b81, 0x3bac, 0x38d2, 0x343d, 0x31af, 0x3b1e, 0x33fc, 0x3864, 0x3624, 0x3905, 0x2945, 0x3b52, 0x2d08, 0x3a17, 0x3b84, 0x3804, 0x3a24, 0x38a3, 0x3562, 0x3ae6, 0x3bba, 0x3a45, 0x3679, 0x31fa, 0x3994, 0x2c3d, 0x383f, 0x399d, 0x34f7, 0x360e, 0x35f3, 0x38f0, 
+0x38d4, 0x399a, 0x3a48, 0x3987, 0x3b54, 0x382c, 0x3210, 0x35ef, 0x36ca, 0x31b4, 0x3625, 0x371f, 0x37bd, 0x3680, 0x3a3a, 0x3ac0, 0x3bbf, 0x3bf5, 0x39f2, 0x29c2, 0x363e, 0x3a4e, 0x3596, 0x3b1b, 0x3459, 0x3669, 0x3aa1, 0x39c3, 0x3376, 0x390d, 0x2456, 0x39b5, 
+0x3a66, 0x3ad8, 0x3b51, 0x36aa, 0x32be, 0x3ac8, 0x392b, 0x3740, 0x3a48, 0x38f5, 0x3b2d, 0x3a5f, 0x2ff3, 0x366f, 0x39d3, 0x35e5, 0x3822, 0x38db, 0x3b8a, 0x34be, 0x2d33, 0x36dd, 0x3578, 0x3bdf, 0x2c7e, 0x39cf, 0x32ff, 0x35c9, 0x3970, 0x3bcb, 0x351e, 0x3956, 
+0x2c42, 0x3308, 0x377a, 0x361c, 0x39a0, 0x36c9, 0x2dcb, 0x3bf2, 0x3b5f, 0x33ee, 0x24c1, 0x2ce9, 0x3927, 0x305d, 0x3702, 0x3119, 0x35f9, 0x3855, 0x3374, 0x349b, 0x3bcf, 0x2dea, 0x34f0, 0x363f, 0x37da, 0x3a74, 0x35fc, 0x35fa, 0x316b, 0x3804, 0x37a7, 0x3986, 
+0x3073, 0x3aed, 0x31c7, 0x3844, 0x34a4, 0x387d, 0x3a20, 0x3037, 0x3a00, 0x3b70, 0x377f, 0x3686, 0x3b7e, 0x38b3, 0x32e3, 0x3323, 0x391e, 0x3228, 0x3930, 0x3997, 0x3a5e, 0x398b, 0x3512, 0x35b0, 0x365c, 0x325d, 0x3b61, 0x38b8, 0x39a4, 0x3423, 0x3bd7, 0x38af, 
+0x2d3d, 0x382d, 0x38ac, 0x26ca, 0x395e, 0x21a8, 0x3520, 0x386f, 0x3b95, 0x32c0, 0x3b84, 0x3a51, 0x3b4b, 0x31d2, 0x3747, 0x3b96, 0x3b40, 0x3535, 0x38d1, 0x3899, 0x3b00, 0x3827, 0x3ae3, 0x38c8, 0x3a07, 0x338d, 0x2e96, 0x3a46, 0x394a, 0x39de, 0x2951, 0x3a02, 
+0x3838, 0x2d45, 0x28c0, 0x3958, 0x3070, 0x2aa2, 0x3510, 0x38ce, 0x271c, 0x3440, 0x3954, 0x30bc, 0x3b35, 0x2f1d, 0x3afb, 0x2dae, 0x356f, 0x2e13, 0x3981, 0x326d, 0x3a28, 0x3a36, 0x3a95, 0x38cb, 0x38db, 0x3150, 0x2c9e, 0x34c5, 0x3adb, 0x3bdf, 0x38f2, 0x3994, 
+0x36f8, 0x31c0, 0x3a4f, 0x3825, 0x394b, 0x3a8b, 0x38ac, 0x3167, 0x2e2d, 0x3a93, 0x34f3, 0x37bd, 0x3b63, 0x2f2f, 0x3ae0, 0x3ad8, 0x34a8, 0x2e1c, 0x3890, 0x3705, 0x3b69, 0x3bc1, 0x28af, 0x3b36, 0x348b, 0x3111, 0x3a8d, 0x389c, 0x3916, 0x36dc, 0x3bae, 0x3874, 
+0x3593, 0x3638, 0x3018, 0x3a56, 0x38a3, 0x2ad4, 0x3a25, 0x38d7, 0x3864, 0x31c1, 0x28d1, 0x39c8, 0x37d6, 0x2c7f, 0x3ba5, 0x34b8, 0x3bef, 0x3b83, 0x3ab5, 0x3062, 0x38bc, 0x399c, 0x2ce4, 0x2f2c, 0x39bf, 0x2ed1, 0x385f, 0x37e0, 0x35ee, 0x397d, 0x3b0c, 0x3049, 
+0x39d5, 0x322e, 0x3936, 0x3747, 0x2e15, 0x3b41, 0x3874, 0x3bd0, 0x2c04, 0x3800, 0x375b, 0x3b2d, 0x38d8, 0x3a51, 0x3406, 0x38da, 0x38ba, 0x3497, 0x382e, 0x35fc, 0x39d4, 0x3775, 0x3b1e, 0x3813, 0x3649, 0x31af, 0x37bb, 0x334a, 0x3a6e, 0x3284, 0x26e0, 0x2e01, 
+0x2ebb, 0x344b, 0x3821, 0x381a, 0x385a, 0x2534, 0x3635, 0x2a92, 0x3b8c, 0x31f0, 0x3947, 0x3ac7, 0x3743, 0x3924, 0x39e4, 0x358f, 0x2b62, 0x392c, 0x3955, 0x3341, 0x3676, 0x38ac, 0x3957, 0x335b, 0x2ca2, 0x39ff, 0x37cb, 0x341f, 0x3ac9, 0x3b6c, 0x2f14, 0x34c3, 
+0x3018, 0x3169, 0x355b, 0x3624, 0x31ed, 0x379e, 0x3268, 0x309b, 0x35db, 0x3872, 0x3bdb, 0x34c7, 0x3408, 0x3359, 0x3920, 0x331f, 0x3866, 0x3af0, 0x2a1a, 0x39e0, 0x3b14, 0x34fa, 0x2d18, 0x3963, 0x35e8, 0x2539, 0x38f5, 0x37b3, 0x378f, 0x31b5, 0x3a6c, 0x3685, 
+0x3a06, 0x318a, 0x2934, 0x33c1, 0x3be8, 0x375b, 0x3860, 0x3543, 0x3702, 0x3951, 0x3677, 0x37ff, 0x2e27, 0x2e3a, 0x340f, 0x3817, 0x2f04, 0x357e, 0x3a1d, 0x2dd6, 0x252a, 0x3945, 0x162a, 0x3b19, 0x3a53, 0x35d2, 0x3a5d, 0x3474, 0x38e9, 0x374b, 0x387c, 0x1f1a, 
+0x38ac, 0x3291, 0x3393, 0x3b53, 0x3169, 0x3bca, 0x2f1a, 0x3551, 0x38a3, 0x28e3, 0x369d, 0x34a1, 0x38a8, 0x34c3, 0x3841, 0x390d, 0x3b13, 0x3282, 0x3a29, 0x3a78, 0x2df3, 0x3a37, 0x35f4, 0x35a6, 0x38e8, 0x3328, 0x3beb, 0x390b, 0x32dc, 0x34dc, 0x396d, 0x3a78, 
+0x39ba, 0x3a06, 0x2cdd, 0x3bc3, 0x2d43, 0x2992, 0x3663, 0x3a68, 0x2c3e, 0x394e, 0x2c9f, 0x380e, 0x37f5, 0x3557, 0x2873, 0x390f, 0x39e7, 0x3939, 0x3669, 0x385c, 0x3a68, 0x32c4, 0x2b04, 0x2d6d, 0x39d3, 0x3895, 0x331d, 0x3b59, 0x3463, 0x2b6a, 0x31de, 0x3296, 
+0x3aae, 0x3bcd, 0x345a, 0x3897, 0x374b, 0x3bd4, 0x38a2, 0x357f, 0x3402, 0x3a0c, 0x3507, 0x3865, 0x3a54, 0x3878, 0x3859, 0x383e, 0x32b5, 0x34ea, 0x328d, 0x38b6, 0x3464, 0x2f5b, 0x35ff, 0x3817, 0x2f24, 0x3533, 0x3b21, 0x37ba, 0x3837, 0x2e34, 0x3bad, 0x34bc 
+};
\ No newline at end of file
diff --git a/hwpe/redmule_256iter/inc/x_input.h b/hwpe/redmule_256iter/inc/x_input.h
new file mode 100644
index 0000000..1e38d23
--- /dev/null
+++ b/hwpe/redmule_256iter/inc/x_input.h
@@ -0,0 +1,27 @@
+ /* Header file generated by RedMulE Golden Model */
+uint16_t x_inp [768] = {
+0x2153, 0x3bb5, 0x3896, 0x365f, 0x2483, 0x3518, 0x2dd1, 0x3bca, 0x397b, 0x29b1, 0x3705, 0x36c8, 0x398b, 0x3661, 0x2f05, 0x365a, 0x3bf9, 0x34df, 0x363b, 0x38d9, 0x39c6, 0x3abb, 0x3952, 0x38f2, 0x392d, 0x3b3e, 0x2afb, 0x3a9d, 0x353b, 0x3b73, 0x3a01, 0x3679, 
+0x3934, 0x397d, 0x2904, 0x3822, 0x3462, 0x3b44, 0x39e9, 0x28be, 0x331e, 0x3a1d, 0x39e5, 0x34da, 0x3a19, 0x3906, 0x1d35, 0x3871, 0x31e7, 0x3b29, 0x325d, 0x3797, 0x2b2f, 0x38b4, 0x232f, 0x38aa, 0x3aca, 0x316f, 0x3811, 0x3950, 0x32ea, 0x3bc7, 0x382c, 0x38a2, 
+0x29ce, 0x3afa, 0x3a39, 0x2ccc, 0x39fd, 0x3b3d, 0x384a, 0x3a35, 0x3802, 0x366a, 0x37ec, 0x3598, 0x3bf8, 0x3a85, 0x3a1b, 0x386e, 0x3b4c, 0x39de, 0x38c2, 0x2f93, 0x3b4c, 0x39c4, 0x3b9e, 0x3844, 0x346d, 0x3bff, 0x32ce, 0x296d, 0x3130, 0x3b3d, 0x3b44, 0x369d, 
+0x3b13, 0x31ed, 0x330a, 0x3831, 0x34e7, 0x37b3, 0x331a, 0x3918, 0x32d3, 0x3995, 0x3991, 0x3919, 0x3a26, 0x385b, 0x2b76, 0x3a3b, 0x37f2, 0x26a7, 0x3225, 0x3b64, 0x28f0, 0x3456, 0x3822, 0x341e, 0x381a, 0x38d8, 0x2c11, 0x33be, 0x33ac, 0x353f, 0x3476, 0x3abc, 
+0x36ec, 0x3a1d, 0x39d3, 0x3821, 0x36ac, 0x3bce, 0x3ad2, 0x3616, 0x36a1, 0x2cb3, 0x38d2, 0x314f, 0x385c, 0x3b63, 0x3bb6, 0x2951, 0x372d, 0x2c42, 0x3823, 0x3883, 0x3872, 0x31ee, 0x36c5, 0x399a, 0x31b0, 0x3887, 0x3884, 0x3865, 0x3896, 0x36c3, 0x32e3, 0x346c, 
+0x3935, 0x3b50, 0x2b6d, 0x38cd, 0x388f, 0x3389, 0x395d, 0x31cd, 0x2efd, 0x3154, 0x2f35, 0x3444, 0x3293, 0x3b6b, 0x1bec, 0x3b69, 0x3bf3, 0x3611, 0x3508, 0x3742, 0x3a50, 0x3ab7, 0x3457, 0x38d3, 0x3344, 0x38e8, 0x33c0, 0x3668, 0x3bee, 0x3b21, 0x3727, 0x3121, 
+0x316c, 0x3288, 0x2d50, 0x2e74, 0x35d5, 0x37e2, 0x303d, 0x36af, 0x341f, 0x3436, 0x2df7, 0x399d, 0x30f4, 0x3aaf, 0x34e4, 0x2c2a, 0x3116, 0x34d3, 0x36ac, 0x35e3, 0x3760, 0x36e1, 0x3ad2, 0x3547, 0x38f4, 0x369c, 0x3ba9, 0x34f0, 0x3a39, 0x3b19, 0x36e6, 0x395d, 
+0x3be8, 0x3293, 0x3bfc, 0x3435, 0x2eb3, 0x3360, 0x3919, 0x3bed, 0x396a, 0x37fc, 0x3242, 0x384b, 0x38cb, 0x3b2c, 0x3b28, 0x28cf, 0x3828, 0x3855, 0x3ba9, 0x2fa7, 0x340b, 0x32f1, 0x3ada, 0x36fa, 0x31f5, 0x3436, 0x29d0, 0x33e6, 0x3232, 0x3bec, 0x3904, 0x2797, 
+0x3b81, 0x3bac, 0x38d2, 0x343d, 0x31af, 0x3b1e, 0x33fc, 0x3864, 0x3624, 0x3905, 0x2945, 0x3b52, 0x2d08, 0x3a17, 0x3b84, 0x3804, 0x3a24, 0x38a3, 0x3562, 0x3ae6, 0x3bba, 0x3a45, 0x3679, 0x31fa, 0x3994, 0x2c3d, 0x383f, 0x399d, 0x34f7, 0x360e, 0x35f3, 0x38f0, 
+0x38d4, 0x399a, 0x3a48, 0x3987, 0x3b54, 0x382c, 0x3210, 0x35ef, 0x36ca, 0x31b4, 0x3625, 0x371f, 0x37bd, 0x3680, 0x3a3a, 0x3ac0, 0x3bbf, 0x3bf5, 0x39f2, 0x29c2, 0x363e, 0x3a4e, 0x3596, 0x3b1b, 0x3459, 0x3669, 0x3aa1, 0x39c3, 0x3376, 0x390d, 0x2456, 0x39b5, 
+0x3a66, 0x3ad8, 0x3b51, 0x36aa, 0x32be, 0x3ac8, 0x392b, 0x3740, 0x3a48, 0x38f5, 0x3b2d, 0x3a5f, 0x2ff3, 0x366f, 0x39d3, 0x35e5, 0x3822, 0x38db, 0x3b8a, 0x34be, 0x2d33, 0x36dd, 0x3578, 0x3bdf, 0x2c7e, 0x39cf, 0x32ff, 0x35c9, 0x3970, 0x3bcb, 0x351e, 0x3956, 
+0x2c42, 0x3308, 0x377a, 0x361c, 0x39a0, 0x36c9, 0x2dcb, 0x3bf2, 0x3b5f, 0x33ee, 0x24c1, 0x2ce9, 0x3927, 0x305d, 0x3702, 0x3119, 0x35f9, 0x3855, 0x3374, 0x349b, 0x3bcf, 0x2dea, 0x34f0, 0x363f, 0x37da, 0x3a74, 0x35fc, 0x35fa, 0x316b, 0x3804, 0x37a7, 0x3986, 
+0x3073, 0x3aed, 0x31c7, 0x3844, 0x34a4, 0x387d, 0x3a20, 0x3037, 0x3a00, 0x3b70, 0x377f, 0x3686, 0x3b7e, 0x38b3, 0x32e3, 0x3323, 0x391e, 0x3228, 0x3930, 0x3997, 0x3a5e, 0x398b, 0x3512, 0x35b0, 0x365c, 0x325d, 0x3b61, 0x38b8, 0x39a4, 0x3423, 0x3bd7, 0x38af, 
+0x2d3d, 0x382d, 0x38ac, 0x26ca, 0x395e, 0x21a8, 0x3520, 0x386f, 0x3b95, 0x32c0, 0x3b84, 0x3a51, 0x3b4b, 0x31d2, 0x3747, 0x3b96, 0x3b40, 0x3535, 0x38d1, 0x3899, 0x3b00, 0x3827, 0x3ae3, 0x38c8, 0x3a07, 0x338d, 0x2e96, 0x3a46, 0x394a, 0x39de, 0x2951, 0x3a02, 
+0x3838, 0x2d45, 0x28c0, 0x3958, 0x3070, 0x2aa2, 0x3510, 0x38ce, 0x271c, 0x3440, 0x3954, 0x30bc, 0x3b35, 0x2f1d, 0x3afb, 0x2dae, 0x356f, 0x2e13, 0x3981, 0x326d, 0x3a28, 0x3a36, 0x3a95, 0x38cb, 0x38db, 0x3150, 0x2c9e, 0x34c5, 0x3adb, 0x3bdf, 0x38f2, 0x3994, 
+0x36f8, 0x31c0, 0x3a4f, 0x3825, 0x394b, 0x3a8b, 0x38ac, 0x3167, 0x2e2d, 0x3a93, 0x34f3, 0x37bd, 0x3b63, 0x2f2f, 0x3ae0, 0x3ad8, 0x34a8, 0x2e1c, 0x3890, 0x3705, 0x3b69, 0x3bc1, 0x28af, 0x3b36, 0x348b, 0x3111, 0x3a8d, 0x389c, 0x3916, 0x36dc, 0x3bae, 0x3874, 
+0x3593, 0x3638, 0x3018, 0x3a56, 0x38a3, 0x2ad4, 0x3a25, 0x38d7, 0x3864, 0x31c1, 0x28d1, 0x39c8, 0x37d6, 0x2c7f, 0x3ba5, 0x34b8, 0x3bef, 0x3b83, 0x3ab5, 0x3062, 0x38bc, 0x399c, 0x2ce4, 0x2f2c, 0x39bf, 0x2ed1, 0x385f, 0x37e0, 0x35ee, 0x397d, 0x3b0c, 0x3049, 
+0x39d5, 0x322e, 0x3936, 0x3747, 0x2e15, 0x3b41, 0x3874, 0x3bd0, 0x2c04, 0x3800, 0x375b, 0x3b2d, 0x38d8, 0x3a51, 0x3406, 0x38da, 0x38ba, 0x3497, 0x382e, 0x35fc, 0x39d4, 0x3775, 0x3b1e, 0x3813, 0x3649, 0x31af, 0x37bb, 0x334a, 0x3a6e, 0x3284, 0x26e0, 0x2e01, 
+0x2ebb, 0x344b, 0x3821, 0x381a, 0x385a, 0x2534, 0x3635, 0x2a92, 0x3b8c, 0x31f0, 0x3947, 0x3ac7, 0x3743, 0x3924, 0x39e4, 0x358f, 0x2b62, 0x392c, 0x3955, 0x3341, 0x3676, 0x38ac, 0x3957, 0x335b, 0x2ca2, 0x39ff, 0x37cb, 0x341f, 0x3ac9, 0x3b6c, 0x2f14, 0x34c3, 
+0x3018, 0x3169, 0x355b, 0x3624, 0x31ed, 0x379e, 0x3268, 0x309b, 0x35db, 0x3872, 0x3bdb, 0x34c7, 0x3408, 0x3359, 0x3920, 0x331f, 0x3866, 0x3af0, 0x2a1a, 0x39e0, 0x3b14, 0x34fa, 0x2d18, 0x3963, 0x35e8, 0x2539, 0x38f5, 0x37b3, 0x378f, 0x31b5, 0x3a6c, 0x3685, 
+0x3a06, 0x318a, 0x2934, 0x33c1, 0x3be8, 0x375b, 0x3860, 0x3543, 0x3702, 0x3951, 0x3677, 0x37ff, 0x2e27, 0x2e3a, 0x340f, 0x3817, 0x2f04, 0x357e, 0x3a1d, 0x2dd6, 0x252a, 0x3945, 0x162a, 0x3b19, 0x3a53, 0x35d2, 0x3a5d, 0x3474, 0x38e9, 0x374b, 0x387c, 0x1f1a, 
+0x38ac, 0x3291, 0x3393, 0x3b53, 0x3169, 0x3bca, 0x2f1a, 0x3551, 0x38a3, 0x28e3, 0x369d, 0x34a1, 0x38a8, 0x34c3, 0x3841, 0x390d, 0x3b13, 0x3282, 0x3a29, 0x3a78, 0x2df3, 0x3a37, 0x35f4, 0x35a6, 0x38e8, 0x3328, 0x3beb, 0x390b, 0x32dc, 0x34dc, 0x396d, 0x3a78, 
+0x39ba, 0x3a06, 0x2cdd, 0x3bc3, 0x2d43, 0x2992, 0x3663, 0x3a68, 0x2c3e, 0x394e, 0x2c9f, 0x380e, 0x37f5, 0x3557, 0x2873, 0x390f, 0x39e7, 0x3939, 0x3669, 0x385c, 0x3a68, 0x32c4, 0x2b04, 0x2d6d, 0x39d3, 0x3895, 0x331d, 0x3b59, 0x3463, 0x2b6a, 0x31de, 0x3296, 
+0x3aae, 0x3bcd, 0x345a, 0x3897, 0x374b, 0x3bd4, 0x38a2, 0x357f, 0x3402, 0x3a0c, 0x3507, 0x3865, 0x3a54, 0x3878, 0x3859, 0x383e, 0x32b5, 0x34ea, 0x328d, 0x38b6, 0x3464, 0x2f5b, 0x35ff, 0x3817, 0x2f24, 0x3533, 0x3b21, 0x37ba, 0x3837, 0x2e34, 0x3bad, 0x34bc 
+};
\ No newline at end of file
diff --git a/hwpe/redmule_256iter/inc/y_2D.h b/hwpe/redmule_256iter/inc/y_2D.h
new file mode 100644
index 0000000..9484a10
--- /dev/null
+++ b/hwpe/redmule_256iter/inc/y_2D.h
@@ -0,0 +1,27 @@
+ /* Header file generated by RedMulE Golden Model */
+uint16_t y_inp_2D [32][32] = {
+0x3150, 0x2dc1, 0x3033, 0x31f5, 0x3bb6, 0x3bff, 0x39f9, 0x3662, 0x3720, 0x351d, 0x384b, 0x3093, 0x3b9d, 0x35ad, 0x3695, 0x3466, 0x2300, 0x3445, 0x33ae, 0x3586, 0x38a3, 0x3bdb, 0x33a2, 0x379b, 0x3a0e, 0x38b0, 0x39ba, 0x379b, 0x39d3, 0x3a51, 0x3b30, 0x3794, 
+0x3b76, 0x3042, 0x38cc, 0x2dfc, 0x3b1a, 0x37fb, 0x38f7, 0x3824, 0x386f, 0x38c7, 0x36ee, 0x3a9c, 0x38d3, 0x2c67, 0x3a80, 0x2f30, 0x3328, 0x3721, 0x3790, 0x34e5, 0x3a6c, 0x3643, 0x3934, 0x3034, 0x38d4, 0x362e, 0x3b4b, 0x3408, 0x30c2, 0x370e, 0x3b31, 0x3b16, 
+0x3b6b, 0x39d4, 0x339c, 0x381e, 0x313e, 0x3671, 0x3ae2, 0x3479, 0x3940, 0x342d, 0x3925, 0x370a, 0x35d8, 0x2dad, 0x3888, 0x24b9, 0x375d, 0x34bd, 0x3243, 0x2ebb, 0x3970, 0x3a21, 0x3a07, 0x3877, 0x3888, 0x3569, 0x372d, 0x2ac1, 0x331e, 0x384d, 0x3996, 0x34a4, 
+0x35c1, 0x33a9, 0x21ed, 0x3a42, 0x388d, 0x34e4, 0x33c3, 0x34f9, 0x3a7b, 0x33fb, 0x2cdd, 0x3b0e, 0x333b, 0x3973, 0x34fc, 0x3771, 0x32ea, 0x2de4, 0x31a8, 0x3946, 0x3657, 0x3a4e, 0x36f6, 0x2829, 0x3ba2, 0x3bdc, 0x3bb3, 0x306c, 0x398d, 0x3a1f, 0x3991, 0x3846, 
+0x3547, 0x3292, 0x2e85, 0x31ed, 0x3979, 0x3a90, 0x28a4, 0x3bed, 0x36d8, 0x340e, 0x3b6a, 0x3ab6, 0x3824, 0x382b, 0x3ac3, 0x3811, 0x36d7, 0x3519, 0x3a92, 0x3a42, 0x29d1, 0x383a, 0x3a9b, 0x300e, 0x2cd3, 0x39cd, 0x3874, 0x3a07, 0x2eb1, 0x3b86, 0x3ad8, 0x3a5d, 
+0x3712, 0x284a, 0x38c1, 0x3bec, 0x39c0, 0x32cd, 0x3ad8, 0x3bce, 0x3817, 0x3896, 0x3aa7, 0x3870, 0x3996, 0x32cc, 0x3a4c, 0x3757, 0x3814, 0x3b65, 0x3acb, 0x376e, 0x34c0, 0x3609, 0x3bf0, 0x3b24, 0x3b29, 0x3848, 0x34b7, 0x398a, 0x220c, 0x3498, 0x3a8c, 0x3883, 
+0x38c4, 0x3af6, 0x3a42, 0x2dd6, 0x3147, 0x3717, 0x3a8e, 0x3af9, 0x3296, 0x38ef, 0x34fa, 0x3555, 0x3b29, 0x38de, 0x315e, 0x3773, 0x3b67, 0x3116, 0x38ec, 0x357c, 0x35d0, 0x2518, 0x3958, 0x2a03, 0x37d9, 0x3699, 0x3a1e, 0x3230, 0x3b13, 0x36d4, 0x3b2a, 0x39ad, 
+0x3b10, 0x351a, 0x3b97, 0x3326, 0x2b54, 0x3b7d, 0x386f, 0x373e, 0x37fa, 0x389b, 0x3b90, 0x3292, 0x3975, 0x38f3, 0x37f1, 0x3590, 0x3810, 0x2fd7, 0x3bf7, 0x3a5a, 0x3a1c, 0x34dd, 0x354c, 0x32f8, 0x3095, 0x321e, 0x39e0, 0x395c, 0x3717, 0x357f, 0x394a, 0x34b1, 
+0x3ba4, 0x380c, 0x3604, 0x2f50, 0x348d, 0x3828, 0x3a9f, 0x39ce, 0x32ca, 0x3906, 0x3ab2, 0x2ca5, 0x38c9, 0x362a, 0x34b2, 0x29dc, 0x3a36, 0x3052, 0x31b7, 0x3589, 0x387c, 0x3401, 0x3b22, 0x3ad6, 0x3ae8, 0x3238, 0x3494, 0x3502, 0x3717, 0x3a6c, 0x3229, 0x368c, 
+0x3056, 0x3a56, 0x3498, 0x39eb, 0x2864, 0x342d, 0x39e0, 0x34a1, 0x2b99, 0x3a04, 0x38ff, 0x328c, 0x34d9, 0x387d, 0x3a3c, 0x32e5, 0x39eb, 0x3984, 0x34dd, 0x38a7, 0x373f, 0x39b4, 0x3235, 0x2f58, 0x2f39, 0x3800, 0x3758, 0x3939, 0x39fc, 0x3a4b, 0x38bf, 0x30ee, 
+0x345e, 0x39c8, 0x3a6d, 0x3262, 0x3b81, 0x31dc, 0x3a15, 0x3bd0, 0x36af, 0x36de, 0x37d5, 0x39d7, 0x3ad3, 0x3ac1, 0x3109, 0x35ea, 0x31c6, 0x398d, 0x3987, 0x3a4a, 0x34d2, 0x2ed2, 0x35e6, 0x352c, 0x39eb, 0x3bd6, 0x3a5b, 0x39d1, 0x34aa, 0x3ade, 0x394b, 0x38a1, 
+0x2bed, 0x38de, 0x3811, 0x3813, 0x391a, 0x374b, 0x3829, 0x3725, 0x38f0, 0x3583, 0x3966, 0x3a7d, 0x375a, 0x38fe, 0x3696, 0x361c, 0x39a8, 0x35f0, 0x38e1, 0x3003, 0x3595, 0x316e, 0x3862, 0x3af8, 0x3af2, 0x34c8, 0x381d, 0x37d8, 0x3893, 0x3a9c, 0x3989, 0x308c, 
+0x30cc, 0x2538, 0x399d, 0x3919, 0x399e, 0x21cc, 0x38e9, 0x30f8, 0x3a20, 0x3b3c, 0x3990, 0x259c, 0x3143, 0x3080, 0x3967, 0x3afb, 0x3a1b, 0x3779, 0x2eeb, 0x39f3, 0x379a, 0x369c, 0x3985, 0x3a1b, 0x3ba6, 0x3a53, 0x28d5, 0x3881, 0x31d9, 0x3a34, 0x3bd9, 0x393a, 
+0x3601, 0x2c6e, 0x3636, 0x3298, 0x39bb, 0x3a08, 0x38db, 0x35ad, 0x3a09, 0x36a6, 0x3bc7, 0x3bac, 0x34ae, 0x3291, 0x290b, 0x3250, 0x2648, 0x333d, 0x2bf3, 0x34b1, 0x30e0, 0x351f, 0x3a74, 0x38dc, 0x3883, 0x2841, 0x35e1, 0x390d, 0x3a50, 0x3abd, 0x386d, 0x3bb7, 
+0x3b94, 0x36b7, 0x3a49, 0x332f, 0x3a1d, 0x354b, 0x3bab, 0x3346, 0x3417, 0x351e, 0x3b6d, 0x391a, 0x2db3, 0x3b1c, 0x3a4a, 0x37b7, 0x36cf, 0x3a56, 0x39c4, 0x3be9, 0x34f0, 0x39be, 0x3691, 0x1ba5, 0x3888, 0x3040, 0x3ae1, 0x3b9b, 0x398f, 0x3a49, 0x3a16, 0x38c0, 
+0x386c, 0x39ab, 0x37fa, 0x382c, 0x3a6f, 0x393f, 0x340d, 0x38ef, 0x39d1, 0x3845, 0x398f, 0x363e, 0x3687, 0x3052, 0x3a2b, 0x392c, 0x2f5c, 0x3412, 0x3a1f, 0x3b2f, 0x3bcc, 0x3a63, 0x3a89, 0x36e9, 0x3921, 0x3b80, 0x2dc0, 0x3a03, 0x3beb, 0x38d3, 0x36cb, 0x39a3, 
+0x3978, 0x3a88, 0x3ba4, 0x3561, 0x28c5, 0x33a0, 0x37be, 0x2c39, 0x30ee, 0x3782, 0x2c07, 0x354e, 0x3491, 0x3a92, 0x331a, 0x3b15, 0x32e1, 0x3839, 0x3afb, 0x36c2, 0x2fd0, 0x29ad, 0x3b2e, 0x39c1, 0x2a8c, 0x341a, 0x2f90, 0x395a, 0x3969, 0x37ea, 0x3a5c, 0x3b6d, 
+0x3971, 0x3a93, 0x304e, 0x3623, 0x3a22, 0x31ee, 0x29df, 0x2c93, 0x3a01, 0x3a62, 0x366c, 0x371d, 0x3af3, 0x2e08, 0x3ac0, 0x3642, 0x3a28, 0x368d, 0x2d3d, 0x36d9, 0x32c3, 0x373f, 0x36fe, 0x3487, 0x2c81, 0x3623, 0x3b59, 0x3a91, 0x350a, 0x34f4, 0x3b09, 0x2c25, 
+0x3b13, 0x325a, 0x379e, 0x3a7d, 0x34b1, 0x39d5, 0x2ba8, 0x322b, 0x3b5e, 0x37ab, 0x2e24, 0x3ba9, 0x3a3d, 0x34f7, 0x3ba1, 0x3877, 0x3071, 0x39fb, 0x3bbd, 0x3633, 0x3b36, 0x2daa, 0x3b9b, 0x3aa0, 0x395c, 0x3b8f, 0x38d5, 0x3ab0, 0x3a8f, 0x36c2, 0x3b1f, 0x3489, 
+0x2acc, 0x3845, 0x3715, 0x37d8, 0x3992, 0x3bff, 0x350e, 0x3ad7, 0x39b0, 0x35ac, 0x3287, 0x385f, 0x3bd4, 0x37a3, 0x3438, 0x39a5, 0x3bcf, 0x38c3, 0x34f6, 0x3ae3, 0x3b57, 0x39af, 0x35eb, 0x3bed, 0x34d4, 0x2a95, 0x3b13, 0x384e, 0x3a3b, 0x33da, 0x3bce, 0x3b99, 
+0x3559, 0x3335, 0x3a2e, 0x3123, 0x38db, 0x33d0, 0x3638, 0x3b17, 0x3a72, 0x3afc, 0x3936, 0x3838, 0x2b69, 0x3895, 0x3a1a, 0x3192, 0x39d5, 0x37a5, 0x2eb0, 0x2e8b, 0x329a, 0x3b90, 0x390a, 0x3a1e, 0x3847, 0x375d, 0x3873, 0x35e2, 0x3771, 0x30f5, 0x3231, 0x3bd7, 
+0x2bbc, 0x3ace, 0x31ad, 0x3a6b, 0x28a4, 0x3b48, 0x3ba3, 0x3a84, 0x3353, 0x39f6, 0x381f, 0x2dd6, 0x314c, 0x34af, 0x3929, 0x3921, 0x383b, 0x34b0, 0x3923, 0x32c9, 0x3ae7, 0x318f, 0x3480, 0x2ad8, 0x3042, 0x3a4c, 0x349d, 0x2c12, 0x3abb, 0x3a57, 0x3b0d, 0x3111, 
+0x3359, 0x3a84, 0x38f2, 0x368d, 0x2f4b, 0x3ba0, 0x395c, 0x3026, 0x3a15, 0x2a04, 0x326e, 0x3522, 0x31a2, 0x382f, 0x2ada, 0x3b7c, 0x2f80, 0x3af5, 0x2d35, 0x38fa, 0x39ab, 0x2c6d, 0x2e7a, 0x39f6, 0x31a4, 0x3a53, 0x358c, 0x3951, 0x3a4e, 0x3916, 0x2a3f, 0x3ae9, 
+0x3b03, 0x39f8, 0x39fe, 0x3a61, 0x39fb, 0x3704, 0x360d, 0x39a7, 0x37a9, 0x348f, 0x3a30, 0x3af5, 0x366f, 0x3b29, 0x3a6a, 0x33d5, 0x370a, 0x39cd, 0x3444, 0x3bea, 0x3b2b, 0x312e, 0x3b8e, 0x32cf, 0x3b79, 0x3302, 0x3bba, 0x3962, 0x3413, 0x37a1, 0x39e0, 0x3805 
+};
\ No newline at end of file
diff --git a/hwpe/redmule_256iter/inc/y_input.h b/hwpe/redmule_256iter/inc/y_input.h
new file mode 100644
index 0000000..45a2375
--- /dev/null
+++ b/hwpe/redmule_256iter/inc/y_input.h
@@ -0,0 +1,27 @@
+ /* Header file generated by RedMulE Golden Model */
+uint16_t y_inp [768] = {
+0x3150, 0x2dc1, 0x3033, 0x31f5, 0x3bb6, 0x3bff, 0x39f9, 0x3662, 0x3720, 0x351d, 0x384b, 0x3093, 0x3b9d, 0x35ad, 0x3695, 0x3466, 0x2300, 0x3445, 0x33ae, 0x3586, 0x38a3, 0x3bdb, 0x33a2, 0x379b, 0x3a0e, 0x38b0, 0x39ba, 0x379b, 0x39d3, 0x3a51, 0x3b30, 0x3794, 
+0x3b76, 0x3042, 0x38cc, 0x2dfc, 0x3b1a, 0x37fb, 0x38f7, 0x3824, 0x386f, 0x38c7, 0x36ee, 0x3a9c, 0x38d3, 0x2c67, 0x3a80, 0x2f30, 0x3328, 0x3721, 0x3790, 0x34e5, 0x3a6c, 0x3643, 0x3934, 0x3034, 0x38d4, 0x362e, 0x3b4b, 0x3408, 0x30c2, 0x370e, 0x3b31, 0x3b16, 
+0x3b6b, 0x39d4, 0x339c, 0x381e, 0x313e, 0x3671, 0x3ae2, 0x3479, 0x3940, 0x342d, 0x3925, 0x370a, 0x35d8, 0x2dad, 0x3888, 0x24b9, 0x375d, 0x34bd, 0x3243, 0x2ebb, 0x3970, 0x3a21, 0x3a07, 0x3877, 0x3888, 0x3569, 0x372d, 0x2ac1, 0x331e, 0x384d, 0x3996, 0x34a4, 
+0x35c1, 0x33a9, 0x21ed, 0x3a42, 0x388d, 0x34e4, 0x33c3, 0x34f9, 0x3a7b, 0x33fb, 0x2cdd, 0x3b0e, 0x333b, 0x3973, 0x34fc, 0x3771, 0x32ea, 0x2de4, 0x31a8, 0x3946, 0x3657, 0x3a4e, 0x36f6, 0x2829, 0x3ba2, 0x3bdc, 0x3bb3, 0x306c, 0x398d, 0x3a1f, 0x3991, 0x3846, 
+0x3547, 0x3292, 0x2e85, 0x31ed, 0x3979, 0x3a90, 0x28a4, 0x3bed, 0x36d8, 0x340e, 0x3b6a, 0x3ab6, 0x3824, 0x382b, 0x3ac3, 0x3811, 0x36d7, 0x3519, 0x3a92, 0x3a42, 0x29d1, 0x383a, 0x3a9b, 0x300e, 0x2cd3, 0x39cd, 0x3874, 0x3a07, 0x2eb1, 0x3b86, 0x3ad8, 0x3a5d, 
+0x3712, 0x284a, 0x38c1, 0x3bec, 0x39c0, 0x32cd, 0x3ad8, 0x3bce, 0x3817, 0x3896, 0x3aa7, 0x3870, 0x3996, 0x32cc, 0x3a4c, 0x3757, 0x3814, 0x3b65, 0x3acb, 0x376e, 0x34c0, 0x3609, 0x3bf0, 0x3b24, 0x3b29, 0x3848, 0x34b7, 0x398a, 0x220c, 0x3498, 0x3a8c, 0x3883, 
+0x38c4, 0x3af6, 0x3a42, 0x2dd6, 0x3147, 0x3717, 0x3a8e, 0x3af9, 0x3296, 0x38ef, 0x34fa, 0x3555, 0x3b29, 0x38de, 0x315e, 0x3773, 0x3b67, 0x3116, 0x38ec, 0x357c, 0x35d0, 0x2518, 0x3958, 0x2a03, 0x37d9, 0x3699, 0x3a1e, 0x3230, 0x3b13, 0x36d4, 0x3b2a, 0x39ad, 
+0x3b10, 0x351a, 0x3b97, 0x3326, 0x2b54, 0x3b7d, 0x386f, 0x373e, 0x37fa, 0x389b, 0x3b90, 0x3292, 0x3975, 0x38f3, 0x37f1, 0x3590, 0x3810, 0x2fd7, 0x3bf7, 0x3a5a, 0x3a1c, 0x34dd, 0x354c, 0x32f8, 0x3095, 0x321e, 0x39e0, 0x395c, 0x3717, 0x357f, 0x394a, 0x34b1, 
+0x3ba4, 0x380c, 0x3604, 0x2f50, 0x348d, 0x3828, 0x3a9f, 0x39ce, 0x32ca, 0x3906, 0x3ab2, 0x2ca5, 0x38c9, 0x362a, 0x34b2, 0x29dc, 0x3a36, 0x3052, 0x31b7, 0x3589, 0x387c, 0x3401, 0x3b22, 0x3ad6, 0x3ae8, 0x3238, 0x3494, 0x3502, 0x3717, 0x3a6c, 0x3229, 0x368c, 
+0x3056, 0x3a56, 0x3498, 0x39eb, 0x2864, 0x342d, 0x39e0, 0x34a1, 0x2b99, 0x3a04, 0x38ff, 0x328c, 0x34d9, 0x387d, 0x3a3c, 0x32e5, 0x39eb, 0x3984, 0x34dd, 0x38a7, 0x373f, 0x39b4, 0x3235, 0x2f58, 0x2f39, 0x3800, 0x3758, 0x3939, 0x39fc, 0x3a4b, 0x38bf, 0x30ee, 
+0x345e, 0x39c8, 0x3a6d, 0x3262, 0x3b81, 0x31dc, 0x3a15, 0x3bd0, 0x36af, 0x36de, 0x37d5, 0x39d7, 0x3ad3, 0x3ac1, 0x3109, 0x35ea, 0x31c6, 0x398d, 0x3987, 0x3a4a, 0x34d2, 0x2ed2, 0x35e6, 0x352c, 0x39eb, 0x3bd6, 0x3a5b, 0x39d1, 0x34aa, 0x3ade, 0x394b, 0x38a1, 
+0x2bed, 0x38de, 0x3811, 0x3813, 0x391a, 0x374b, 0x3829, 0x3725, 0x38f0, 0x3583, 0x3966, 0x3a7d, 0x375a, 0x38fe, 0x3696, 0x361c, 0x39a8, 0x35f0, 0x38e1, 0x3003, 0x3595, 0x316e, 0x3862, 0x3af8, 0x3af2, 0x34c8, 0x381d, 0x37d8, 0x3893, 0x3a9c, 0x3989, 0x308c, 
+0x30cc, 0x2538, 0x399d, 0x3919, 0x399e, 0x21cc, 0x38e9, 0x30f8, 0x3a20, 0x3b3c, 0x3990, 0x259c, 0x3143, 0x3080, 0x3967, 0x3afb, 0x3a1b, 0x3779, 0x2eeb, 0x39f3, 0x379a, 0x369c, 0x3985, 0x3a1b, 0x3ba6, 0x3a53, 0x28d5, 0x3881, 0x31d9, 0x3a34, 0x3bd9, 0x393a, 
+0x3601, 0x2c6e, 0x3636, 0x3298, 0x39bb, 0x3a08, 0x38db, 0x35ad, 0x3a09, 0x36a6, 0x3bc7, 0x3bac, 0x34ae, 0x3291, 0x290b, 0x3250, 0x2648, 0x333d, 0x2bf3, 0x34b1, 0x30e0, 0x351f, 0x3a74, 0x38dc, 0x3883, 0x2841, 0x35e1, 0x390d, 0x3a50, 0x3abd, 0x386d, 0x3bb7, 
+0x3b94, 0x36b7, 0x3a49, 0x332f, 0x3a1d, 0x354b, 0x3bab, 0x3346, 0x3417, 0x351e, 0x3b6d, 0x391a, 0x2db3, 0x3b1c, 0x3a4a, 0x37b7, 0x36cf, 0x3a56, 0x39c4, 0x3be9, 0x34f0, 0x39be, 0x3691, 0x1ba5, 0x3888, 0x3040, 0x3ae1, 0x3b9b, 0x398f, 0x3a49, 0x3a16, 0x38c0, 
+0x386c, 0x39ab, 0x37fa, 0x382c, 0x3a6f, 0x393f, 0x340d, 0x38ef, 0x39d1, 0x3845, 0x398f, 0x363e, 0x3687, 0x3052, 0x3a2b, 0x392c, 0x2f5c, 0x3412, 0x3a1f, 0x3b2f, 0x3bcc, 0x3a63, 0x3a89, 0x36e9, 0x3921, 0x3b80, 0x2dc0, 0x3a03, 0x3beb, 0x38d3, 0x36cb, 0x39a3, 
+0x3978, 0x3a88, 0x3ba4, 0x3561, 0x28c5, 0x33a0, 0x37be, 0x2c39, 0x30ee, 0x3782, 0x2c07, 0x354e, 0x3491, 0x3a92, 0x331a, 0x3b15, 0x32e1, 0x3839, 0x3afb, 0x36c2, 0x2fd0, 0x29ad, 0x3b2e, 0x39c1, 0x2a8c, 0x341a, 0x2f90, 0x395a, 0x3969, 0x37ea, 0x3a5c, 0x3b6d, 
+0x3971, 0x3a93, 0x304e, 0x3623, 0x3a22, 0x31ee, 0x29df, 0x2c93, 0x3a01, 0x3a62, 0x366c, 0x371d, 0x3af3, 0x2e08, 0x3ac0, 0x3642, 0x3a28, 0x368d, 0x2d3d, 0x36d9, 0x32c3, 0x373f, 0x36fe, 0x3487, 0x2c81, 0x3623, 0x3b59, 0x3a91, 0x350a, 0x34f4, 0x3b09, 0x2c25, 
+0x3b13, 0x325a, 0x379e, 0x3a7d, 0x34b1, 0x39d5, 0x2ba8, 0x322b, 0x3b5e, 0x37ab, 0x2e24, 0x3ba9, 0x3a3d, 0x34f7, 0x3ba1, 0x3877, 0x3071, 0x39fb, 0x3bbd, 0x3633, 0x3b36, 0x2daa, 0x3b9b, 0x3aa0, 0x395c, 0x3b8f, 0x38d5, 0x3ab0, 0x3a8f, 0x36c2, 0x3b1f, 0x3489, 
+0x2acc, 0x3845, 0x3715, 0x37d8, 0x3992, 0x3bff, 0x350e, 0x3ad7, 0x39b0, 0x35ac, 0x3287, 0x385f, 0x3bd4, 0x37a3, 0x3438, 0x39a5, 0x3bcf, 0x38c3, 0x34f6, 0x3ae3, 0x3b57, 0x39af, 0x35eb, 0x3bed, 0x34d4, 0x2a95, 0x3b13, 0x384e, 0x3a3b, 0x33da, 0x3bce, 0x3b99, 
+0x3559, 0x3335, 0x3a2e, 0x3123, 0x38db, 0x33d0, 0x3638, 0x3b17, 0x3a72, 0x3afc, 0x3936, 0x3838, 0x2b69, 0x3895, 0x3a1a, 0x3192, 0x39d5, 0x37a5, 0x2eb0, 0x2e8b, 0x329a, 0x3b90, 0x390a, 0x3a1e, 0x3847, 0x375d, 0x3873, 0x35e2, 0x3771, 0x30f5, 0x3231, 0x3bd7, 
+0x2bbc, 0x3ace, 0x31ad, 0x3a6b, 0x28a4, 0x3b48, 0x3ba3, 0x3a84, 0x3353, 0x39f6, 0x381f, 0x2dd6, 0x314c, 0x34af, 0x3929, 0x3921, 0x383b, 0x34b0, 0x3923, 0x32c9, 0x3ae7, 0x318f, 0x3480, 0x2ad8, 0x3042, 0x3a4c, 0x349d, 0x2c12, 0x3abb, 0x3a57, 0x3b0d, 0x3111, 
+0x3359, 0x3a84, 0x38f2, 0x368d, 0x2f4b, 0x3ba0, 0x395c, 0x3026, 0x3a15, 0x2a04, 0x326e, 0x3522, 0x31a2, 0x382f, 0x2ada, 0x3b7c, 0x2f80, 0x3af5, 0x2d35, 0x38fa, 0x39ab, 0x2c6d, 0x2e7a, 0x39f6, 0x31a4, 0x3a53, 0x358c, 0x3951, 0x3a4e, 0x3916, 0x2a3f, 0x3ae9, 
+0x3b03, 0x39f8, 0x39fe, 0x3a61, 0x39fb, 0x3704, 0x360d, 0x39a7, 0x37a9, 0x348f, 0x3a30, 0x3af5, 0x366f, 0x3b29, 0x3a6a, 0x33d5, 0x370a, 0x39cd, 0x3444, 0x3bea, 0x3b2b, 0x312e, 0x3b8e, 0x32cf, 0x3b79, 0x3302, 0x3bba, 0x3962, 0x3413, 0x37a1, 0x39e0, 0x3805 
+};
\ No newline at end of file
diff --git a/hwpe/redmule_256iter/inc/z_2D.h b/hwpe/redmule_256iter/inc/z_2D.h
new file mode 100644
index 0000000..aff808a
--- /dev/null
+++ b/hwpe/redmule_256iter/inc/z_2D.h
@@ -0,0 +1,27 @@
+ /* Header file generated by RedMulE Golden Model */
+uint16_t z_oup_2D [24][32] = {
+0x4845, 0x4897, 0x4608, 0x4838, 0x4855, 0x487b, 0x4869, 0x4880, 0x46d1, 0x48b0, 0x48db, 0x483f, 0x48c9, 0x485f, 0x4881, 0x483a, 0x484b, 0x472c, 0x4762, 0x492b, 0x4822, 0x48fd, 0x488e, 0x492e, 0x483e, 0x484f, 0x49e8, 0x46d7, 0x484b, 0x489d, 0x490b, 0x47e9, 
+0x484f, 0x47d2, 0x44be, 0x4747, 0x47c7, 0x46c0, 0x4727, 0x48af, 0x46c5, 0x482d, 0x483d, 0x482e, 0x4897, 0x479f, 0x488b, 0x4749, 0x489a, 0x46a8, 0x46f2, 0x488b, 0x4891, 0x47e8, 0x4872, 0x483d, 0x4716, 0x46fd, 0x49b5, 0x46a0, 0x46e7, 0x47a4, 0x48a1, 0x4767, 
+0x4939, 0x4935, 0x4703, 0x48c1, 0x4863, 0x48bd, 0x4913, 0x48cf, 0x48b6, 0x48b8, 0x4946, 0x4920, 0x495e, 0x48e1, 0x4938, 0x48b2, 0x493a, 0x4882, 0x483b, 0x49d5, 0x4911, 0x4972, 0x496b, 0x49df, 0x48f2, 0x4888, 0x4a46, 0x4821, 0x48c1, 0x490c, 0x49b2, 0x48a3, 
+0x463a, 0x47b0, 0x44cb, 0x4762, 0x4765, 0x46b9, 0x466a, 0x4814, 0x4631, 0x4796, 0x4666, 0x474b, 0x4798, 0x4704, 0x4838, 0x4761, 0x47d3, 0x4590, 0x45ea, 0x48a2, 0x47f1, 0x4844, 0x484b, 0x4776, 0x47d6, 0x46d8, 0x48f3, 0x44d3, 0x46fa, 0x478d, 0x481e, 0x466e, 
+0x4827, 0x481e, 0x45a2, 0x4794, 0x4727, 0x4806, 0x475d, 0x48d5, 0x4708, 0x4828, 0x4862, 0x480d, 0x4895, 0x4832, 0x48bd, 0x47f1, 0x482a, 0x46a7, 0x47b1, 0x492d, 0x484d, 0x4884, 0x48dc, 0x485f, 0x476d, 0x480c, 0x48e9, 0x46d3, 0x4728, 0x4884, 0x48a0, 0x480e, 
+0x4862, 0x4813, 0x4675, 0x485a, 0x47e8, 0x4738, 0x4836, 0x4823, 0x46e7, 0x4821, 0x4822, 0x47b3, 0x4846, 0x4855, 0x4863, 0x4717, 0x4872, 0x47c1, 0x46d5, 0x488e, 0x47e2, 0x485f, 0x487c, 0x48b8, 0x481e, 0x4788, 0x48bd, 0x4677, 0x46c9, 0x47f8, 0x48fe, 0x47fc, 
+0x47a0, 0x47b2, 0x4588, 0x467e, 0x4662, 0x46c7, 0x46e8, 0x4812, 0x4536, 0x474e, 0x46c0, 0x468f, 0x481f, 0x4679, 0x46a1, 0x46e2, 0x4809, 0x4560, 0x4630, 0x47eb, 0x46b5, 0x4757, 0x4848, 0x477f, 0x46a6, 0x46d8, 0x4870, 0x459a, 0x4670, 0x4678, 0x47d2, 0x468c, 
+0x4762, 0x48c4, 0x46e3, 0x4791, 0x46b1, 0x486d, 0x47d0, 0x4867, 0x468d, 0x47f6, 0x48a5, 0x4756, 0x4857, 0x4854, 0x4866, 0x4838, 0x484d, 0x46ec, 0x47d2, 0x48f6, 0x484a, 0x4879, 0x4848, 0x483c, 0x471d, 0x4806, 0x48fa, 0x4730, 0x4768, 0x47b8, 0x4865, 0x46f9, 
+0x48a8, 0x4918, 0x46ca, 0x4867, 0x4800, 0x4862, 0x48d3, 0x4910, 0x474e, 0x4849, 0x48eb, 0x486b, 0x4966, 0x48c5, 0x48f4, 0x4830, 0x48f9, 0x4778, 0x481e, 0x499e, 0x48cf, 0x48f1, 0x4982, 0x4923, 0x487c, 0x47cf, 0x49ea, 0x4649, 0x4773, 0x495e, 0x48b2, 0x483f, 
+0x48a7, 0x4975, 0x4616, 0x481e, 0x481f, 0x4866, 0x48b6, 0x4864, 0x47dc, 0x4873, 0x485c, 0x487f, 0x4938, 0x491f, 0x490d, 0x48b6, 0x48f8, 0x48a1, 0x4859, 0x492d, 0x489c, 0x4915, 0x4899, 0x4887, 0x486c, 0x4859, 0x49ca, 0x471e, 0x4867, 0x4918, 0x48d3, 0x4827, 
+0x488b, 0x4998, 0x4704, 0x481d, 0x48b8, 0x4880, 0x4876, 0x4944, 0x470c, 0x48f2, 0x48b9, 0x489b, 0x4956, 0x48e5, 0x48d6, 0x48a5, 0x48dc, 0x4856, 0x484e, 0x49ab, 0x48e0, 0x490e, 0x48dd, 0x4945, 0x488b, 0x48dd, 0x4a32, 0x47ea, 0x4835, 0x4911, 0x4965, 0x4819, 
+0x460e, 0x481e, 0x452c, 0x4673, 0x475c, 0x4717, 0x46f6, 0x46d0, 0x4696, 0x46bc, 0x4726, 0x481e, 0x4763, 0x46ea, 0x46fe, 0x4758, 0x478b, 0x4627, 0x4704, 0x483f, 0x46ad, 0x47b1, 0x4792, 0x4816, 0x46f2, 0x4684, 0x4827, 0x45a8, 0x472f, 0x47a4, 0x4797, 0x462b, 
+0x483f, 0x48ab, 0x468f, 0x4863, 0x485a, 0x4766, 0x481d, 0x48cb, 0x47dc, 0x4903, 0x48fc, 0x4830, 0x48cc, 0x483e, 0x48ab, 0x4864, 0x4966, 0x4763, 0x4794, 0x499d, 0x488e, 0x488b, 0x48dc, 0x4960, 0x4854, 0x484c, 0x499c, 0x474c, 0x4826, 0x48bc, 0x4949, 0x4883, 
+0x489d, 0x4905, 0x4718, 0x481e, 0x48e3, 0x48f4, 0x48c1, 0x4904, 0x47e8, 0x48b3, 0x4892, 0x48d4, 0x48ff, 0x4894, 0x48d5, 0x4886, 0x48fa, 0x4803, 0x47d2, 0x492e, 0x4870, 0x48b2, 0x48e5, 0x492b, 0x487b, 0x4785, 0x49e3, 0x471d, 0x4837, 0x48bf, 0x489b, 0x48c4, 
+0x475c, 0x4871, 0x464a, 0x4811, 0x47af, 0x471c, 0x4817, 0x4817, 0x463b, 0x484e, 0x477f, 0x464f, 0x4704, 0x487c, 0x47a3, 0x4725, 0x4853, 0x462a, 0x465a, 0x4860, 0x4736, 0x4880, 0x47e1, 0x482b, 0x4811, 0x46c0, 0x48dc, 0x475d, 0x4668, 0x4806, 0x4893, 0x46f4, 
+0x4858, 0x4959, 0x463d, 0x487b, 0x480f, 0x484e, 0x48c0, 0x48a6, 0x4847, 0x4894, 0x48a0, 0x484a, 0x491e, 0x48f4, 0x48fc, 0x48b5, 0x48ce, 0x47d2, 0x47db, 0x497f, 0x4955, 0x4939, 0x48a7, 0x48ce, 0x4890, 0x4884, 0x49d6, 0x4763, 0x486e, 0x4922, 0x48f4, 0x48c3, 
+0x47ec, 0x491c, 0x4698, 0x4783, 0x4715, 0x4754, 0x4745, 0x4752, 0x472f, 0x4832, 0x4817, 0x4809, 0x47f8, 0x48c3, 0x47e6, 0x4800, 0x48b6, 0x4730, 0x480a, 0x48cb, 0x479e, 0x488e, 0x47c2, 0x488e, 0x472f, 0x47ee, 0x489d, 0x4744, 0x4755, 0x4851, 0x4846, 0x47d3, 
+0x4838, 0x48a0, 0x4634, 0x4762, 0x4786, 0x4806, 0x47e3, 0x482d, 0x4726, 0x486c, 0x47b7, 0x4803, 0x48ac, 0x4814, 0x48e0, 0x4839, 0x4827, 0x4750, 0x46f2, 0x48c5, 0x483f, 0x4886, 0x48ad, 0x4856, 0x47e8, 0x47a9, 0x4937, 0x4743, 0x46d0, 0x481f, 0x484c, 0x4804, 
+0x47fd, 0x481f, 0x456d, 0x4813, 0x474d, 0x4807, 0x4688, 0x480e, 0x46e8, 0x4810, 0x469f, 0x4799, 0x4853, 0x478f, 0x47f2, 0x4824, 0x47d0, 0x471f, 0x46da, 0x485f, 0x4813, 0x481c, 0x482e, 0x4863, 0x4786, 0x480b, 0x48c9, 0x46b8, 0x475a, 0x46e2, 0x4852, 0x46c5, 
+0x45af, 0x4802, 0x4466, 0x46c2, 0x465d, 0x4743, 0x46b7, 0x47ba, 0x4636, 0x46c3, 0x4677, 0x4784, 0x485a, 0x47c2, 0x46dc, 0x46ac, 0x47de, 0x460e, 0x465f, 0x4834, 0x47f4, 0x4769, 0x46fc, 0x4810, 0x45fd, 0x45ea, 0x48d0, 0x45b5, 0x4704, 0x4783, 0x4830, 0x46c4, 
+0x4759, 0x47c7, 0x453d, 0x45b0, 0x4741, 0x4702, 0x4736, 0x4793, 0x461b, 0x47ba, 0x470b, 0x46dd, 0x4657, 0x470b, 0x470d, 0x4710, 0x486c, 0x468f, 0x45c3, 0x46ba, 0x479d, 0x483b, 0x46c9, 0x4774, 0x46a9, 0x46a7, 0x4833, 0x4606, 0x4690, 0x46a9, 0x46f5, 0x46a7, 
+0x47ac, 0x48bb, 0x452c, 0x4803, 0x470f, 0x4824, 0x47d5, 0x48cb, 0x4707, 0x484a, 0x4832, 0x4797, 0x4851, 0x482c, 0x487a, 0x4877, 0x4891, 0x465d, 0x47f4, 0x48ce, 0x4898, 0x4899, 0x484e, 0x486a, 0x47ac, 0x47f0, 0x493e, 0x4611, 0x47e2, 0x489e, 0x488c, 0x46af, 
+0x4665, 0x4836, 0x45e4, 0x46b6, 0x46a1, 0x46b9, 0x46c8, 0x46dd, 0x4658, 0x474b, 0x467b, 0x4777, 0x4769, 0x4798, 0x4785, 0x475e, 0x472a, 0x4656, 0x45fb, 0x4881, 0x46fc, 0x472d, 0x476e, 0x47a3, 0x465d, 0x46ca, 0x4855, 0x4500, 0x464f, 0x479a, 0x46c3, 0x4738, 
+0x481e, 0x486c, 0x4659, 0x4801, 0x4756, 0x477a, 0x47d5, 0x487b, 0x4706, 0x4808, 0x484f, 0x4838, 0x4870, 0x4863, 0x48d3, 0x4806, 0x4865, 0x4771, 0x46be, 0x494c, 0x4915, 0x484c, 0x4900, 0x4862, 0x481a, 0x46e8, 0x4974, 0x46a0, 0x4775, 0x483d, 0x487c, 0x480e 
+};
\ No newline at end of file
diff --git a/hwpe/redmule_256iter/inc/z_output.h b/hwpe/redmule_256iter/inc/z_output.h
new file mode 100644
index 0000000..96c7e5f
--- /dev/null
+++ b/hwpe/redmule_256iter/inc/z_output.h
@@ -0,0 +1,27 @@
+ /* Header file generated by RedMulE Golden Model */
+uint16_t z_oup [768] = {
+0x4845, 0x4897, 0x4608, 0x4838, 0x4855, 0x487b, 0x4869, 0x4880, 0x46d1, 0x48b0, 0x48db, 0x483f, 0x48c9, 0x485f, 0x4881, 0x483a, 0x484b, 0x472c, 0x4762, 0x492b, 0x4822, 0x48fd, 0x488e, 0x492e, 0x483e, 0x484f, 0x49e8, 0x46d7, 0x484b, 0x489d, 0x490b, 0x47e9, 
+0x484f, 0x47d2, 0x44be, 0x4747, 0x47c7, 0x46c0, 0x4727, 0x48af, 0x46c5, 0x482d, 0x483d, 0x482e, 0x4897, 0x479f, 0x488b, 0x4749, 0x489a, 0x46a8, 0x46f2, 0x488b, 0x4891, 0x47e8, 0x4872, 0x483d, 0x4716, 0x46fd, 0x49b5, 0x46a0, 0x46e7, 0x47a4, 0x48a1, 0x4767, 
+0x4939, 0x4935, 0x4703, 0x48c1, 0x4863, 0x48bd, 0x4913, 0x48cf, 0x48b6, 0x48b8, 0x4946, 0x4920, 0x495e, 0x48e1, 0x4938, 0x48b2, 0x493a, 0x4882, 0x483b, 0x49d5, 0x4911, 0x4972, 0x496b, 0x49df, 0x48f2, 0x4888, 0x4a46, 0x4821, 0x48c1, 0x490c, 0x49b2, 0x48a3, 
+0x463a, 0x47b0, 0x44cb, 0x4762, 0x4765, 0x46b9, 0x466a, 0x4814, 0x4631, 0x4796, 0x4666, 0x474b, 0x4798, 0x4704, 0x4838, 0x4761, 0x47d3, 0x4590, 0x45ea, 0x48a2, 0x47f1, 0x4844, 0x484b, 0x4776, 0x47d6, 0x46d8, 0x48f3, 0x44d3, 0x46fa, 0x478d, 0x481e, 0x466e, 
+0x4827, 0x481e, 0x45a2, 0x4794, 0x4727, 0x4806, 0x475d, 0x48d5, 0x4708, 0x4828, 0x4862, 0x480d, 0x4895, 0x4832, 0x48bd, 0x47f1, 0x482a, 0x46a7, 0x47b1, 0x492d, 0x484d, 0x4884, 0x48dc, 0x485f, 0x476d, 0x480c, 0x48e9, 0x46d3, 0x4728, 0x4884, 0x48a0, 0x480e, 
+0x4862, 0x4813, 0x4675, 0x485a, 0x47e8, 0x4738, 0x4836, 0x4823, 0x46e7, 0x4821, 0x4822, 0x47b3, 0x4846, 0x4855, 0x4863, 0x4717, 0x4872, 0x47c1, 0x46d5, 0x488e, 0x47e2, 0x485f, 0x487c, 0x48b8, 0x481e, 0x4788, 0x48bd, 0x4677, 0x46c9, 0x47f8, 0x48fe, 0x47fc, 
+0x47a0, 0x47b2, 0x4588, 0x467e, 0x4662, 0x46c7, 0x46e8, 0x4812, 0x4536, 0x474e, 0x46c0, 0x468f, 0x481f, 0x4679, 0x46a1, 0x46e2, 0x4809, 0x4560, 0x4630, 0x47eb, 0x46b5, 0x4757, 0x4848, 0x477f, 0x46a6, 0x46d8, 0x4870, 0x459a, 0x4670, 0x4678, 0x47d2, 0x468c, 
+0x4762, 0x48c4, 0x46e3, 0x4791, 0x46b1, 0x486d, 0x47d0, 0x4867, 0x468d, 0x47f6, 0x48a5, 0x4756, 0x4857, 0x4854, 0x4866, 0x4838, 0x484d, 0x46ec, 0x47d2, 0x48f6, 0x484a, 0x4879, 0x4848, 0x483c, 0x471d, 0x4806, 0x48fa, 0x4730, 0x4768, 0x47b8, 0x4865, 0x46f9, 
+0x48a8, 0x4918, 0x46ca, 0x4867, 0x4800, 0x4862, 0x48d3, 0x4910, 0x474e, 0x4849, 0x48eb, 0x486b, 0x4966, 0x48c5, 0x48f4, 0x4830, 0x48f9, 0x4778, 0x481e, 0x499e, 0x48cf, 0x48f1, 0x4982, 0x4923, 0x487c, 0x47cf, 0x49ea, 0x4649, 0x4773, 0x495e, 0x48b2, 0x483f, 
+0x48a7, 0x4975, 0x4616, 0x481e, 0x481f, 0x4866, 0x48b6, 0x4864, 0x47dc, 0x4873, 0x485c, 0x487f, 0x4938, 0x491f, 0x490d, 0x48b6, 0x48f8, 0x48a1, 0x4859, 0x492d, 0x489c, 0x4915, 0x4899, 0x4887, 0x486c, 0x4859, 0x49ca, 0x471e, 0x4867, 0x4918, 0x48d3, 0x4827, 
+0x488b, 0x4998, 0x4704, 0x481d, 0x48b8, 0x4880, 0x4876, 0x4944, 0x470c, 0x48f2, 0x48b9, 0x489b, 0x4956, 0x48e5, 0x48d6, 0x48a5, 0x48dc, 0x4856, 0x484e, 0x49ab, 0x48e0, 0x490e, 0x48dd, 0x4945, 0x488b, 0x48dd, 0x4a32, 0x47ea, 0x4835, 0x4911, 0x4965, 0x4819, 
+0x460e, 0x481e, 0x452c, 0x4673, 0x475c, 0x4717, 0x46f6, 0x46d0, 0x4696, 0x46bc, 0x4726, 0x481e, 0x4763, 0x46ea, 0x46fe, 0x4758, 0x478b, 0x4627, 0x4704, 0x483f, 0x46ad, 0x47b1, 0x4792, 0x4816, 0x46f2, 0x4684, 0x4827, 0x45a8, 0x472f, 0x47a4, 0x4797, 0x462b, 
+0x483f, 0x48ab, 0x468f, 0x4863, 0x485a, 0x4766, 0x481d, 0x48cb, 0x47dc, 0x4903, 0x48fc, 0x4830, 0x48cc, 0x483e, 0x48ab, 0x4864, 0x4966, 0x4763, 0x4794, 0x499d, 0x488e, 0x488b, 0x48dc, 0x4960, 0x4854, 0x484c, 0x499c, 0x474c, 0x4826, 0x48bc, 0x4949, 0x4883, 
+0x489d, 0x4905, 0x4718, 0x481e, 0x48e3, 0x48f4, 0x48c1, 0x4904, 0x47e8, 0x48b3, 0x4892, 0x48d4, 0x48ff, 0x4894, 0x48d5, 0x4886, 0x48fa, 0x4803, 0x47d2, 0x492e, 0x4870, 0x48b2, 0x48e5, 0x492b, 0x487b, 0x4785, 0x49e3, 0x471d, 0x4837, 0x48bf, 0x489b, 0x48c4, 
+0x475c, 0x4871, 0x464a, 0x4811, 0x47af, 0x471c, 0x4817, 0x4817, 0x463b, 0x484e, 0x477f, 0x464f, 0x4704, 0x487c, 0x47a3, 0x4725, 0x4853, 0x462a, 0x465a, 0x4860, 0x4736, 0x4880, 0x47e1, 0x482b, 0x4811, 0x46c0, 0x48dc, 0x475d, 0x4668, 0x4806, 0x4893, 0x46f4, 
+0x4858, 0x4959, 0x463d, 0x487b, 0x480f, 0x484e, 0x48c0, 0x48a6, 0x4847, 0x4894, 0x48a0, 0x484a, 0x491e, 0x48f4, 0x48fc, 0x48b5, 0x48ce, 0x47d2, 0x47db, 0x497f, 0x4955, 0x4939, 0x48a7, 0x48ce, 0x4890, 0x4884, 0x49d6, 0x4763, 0x486e, 0x4922, 0x48f4, 0x48c3, 
+0x47ec, 0x491c, 0x4698, 0x4783, 0x4715, 0x4754, 0x4745, 0x4752, 0x472f, 0x4832, 0x4817, 0x4809, 0x47f8, 0x48c3, 0x47e6, 0x4800, 0x48b6, 0x4730, 0x480a, 0x48cb, 0x479e, 0x488e, 0x47c2, 0x488e, 0x472f, 0x47ee, 0x489d, 0x4744, 0x4755, 0x4851, 0x4846, 0x47d3, 
+0x4838, 0x48a0, 0x4634, 0x4762, 0x4786, 0x4806, 0x47e3, 0x482d, 0x4726, 0x486c, 0x47b7, 0x4803, 0x48ac, 0x4814, 0x48e0, 0x4839, 0x4827, 0x4750, 0x46f2, 0x48c5, 0x483f, 0x4886, 0x48ad, 0x4856, 0x47e8, 0x47a9, 0x4937, 0x4743, 0x46d0, 0x481f, 0x484c, 0x4804, 
+0x47fd, 0x481f, 0x456d, 0x4813, 0x474d, 0x4807, 0x4688, 0x480e, 0x46e8, 0x4810, 0x469f, 0x4799, 0x4853, 0x478f, 0x47f2, 0x4824, 0x47d0, 0x471f, 0x46da, 0x485f, 0x4813, 0x481c, 0x482e, 0x4863, 0x4786, 0x480b, 0x48c9, 0x46b8, 0x475a, 0x46e2, 0x4852, 0x46c5, 
+0x45af, 0x4802, 0x4466, 0x46c2, 0x465d, 0x4743, 0x46b7, 0x47ba, 0x4636, 0x46c3, 0x4677, 0x4784, 0x485a, 0x47c2, 0x46dc, 0x46ac, 0x47de, 0x460e, 0x465f, 0x4834, 0x47f4, 0x4769, 0x46fc, 0x4810, 0x45fd, 0x45ea, 0x48d0, 0x45b5, 0x4704, 0x4783, 0x4830, 0x46c4, 
+0x4759, 0x47c7, 0x453d, 0x45b0, 0x4741, 0x4702, 0x4736, 0x4793, 0x461b, 0x47ba, 0x470b, 0x46dd, 0x4657, 0x470b, 0x470d, 0x4710, 0x486c, 0x468f, 0x45c3, 0x46ba, 0x479d, 0x483b, 0x46c9, 0x4774, 0x46a9, 0x46a7, 0x4833, 0x4606, 0x4690, 0x46a9, 0x46f5, 0x46a7, 
+0x47ac, 0x48bb, 0x452c, 0x4803, 0x470f, 0x4824, 0x47d5, 0x48cb, 0x4707, 0x484a, 0x4832, 0x4797, 0x4851, 0x482c, 0x487a, 0x4877, 0x4891, 0x465d, 0x47f4, 0x48ce, 0x4898, 0x4899, 0x484e, 0x486a, 0x47ac, 0x47f0, 0x493e, 0x4611, 0x47e2, 0x489e, 0x488c, 0x46af, 
+0x4665, 0x4836, 0x45e4, 0x46b6, 0x46a1, 0x46b9, 0x46c8, 0x46dd, 0x4658, 0x474b, 0x467b, 0x4777, 0x4769, 0x4798, 0x4785, 0x475e, 0x472a, 0x4656, 0x45fb, 0x4881, 0x46fc, 0x472d, 0x476e, 0x47a3, 0x465d, 0x46ca, 0x4855, 0x4500, 0x464f, 0x479a, 0x46c3, 0x4738, 
+0x481e, 0x486c, 0x4659, 0x4801, 0x4756, 0x477a, 0x47d5, 0x487b, 0x4706, 0x4808, 0x484f, 0x4838, 0x4870, 0x4863, 0x48d3, 0x4806, 0x4865, 0x4771, 0x46be, 0x494c, 0x4915, 0x484c, 0x4900, 0x4862, 0x481a, 0x46e8, 0x4974, 0x46a0, 0x4775, 0x483d, 0x487c, 0x480e 
+};
\ No newline at end of file
diff --git a/hwpe/redmule_256iter/pulp_inject_fault.tcl b/hwpe/redmule_256iter/pulp_inject_fault.tcl
new file mode 100644
index 0000000..61ccadf
--- /dev/null
+++ b/hwpe/redmule_256iter/pulp_inject_fault.tcl
@@ -0,0 +1,53 @@
+# Copyright 2023 ETH Zurich and University of Bologna.
+# Solderpad Hardware License, Version 0.51, see LICENSE for details.
+# SPDX-License-Identifier: SHL-0.51
+#
+# Author: Michael Rogenmoser (michaero@iis.ee.ethz.ch)
+
+transcript quietly
+if {! [info exists ::env(VSIM_PATH)]} {error "Define VSIM_PATH"}
+set utils_base_path  [file join $::env(VSIM_PATH) scripts fault_injection_utils]
+set script_base_path [file join $::env(VSIM_PATH) fault_injection_sim scripts]
+
+set verbosity            2
+set log_injections       1
+# Easy way to generate a variable seed
+# set seed                 [clock seconds]
+# Default value
+set seed                 12345
+set print_statistics     1
+
+set inject_start_time 550000000000ps
+set inject_stop_time 750000000000ps
+set injection_clock "pulp_cluster_tb/cluster_i/clk_i"
+set injection_clock_trigger 0
+set fault_period 150
+set rand_initial_injection_phase 0
+# max_num set to 0 means until stop_time
+set max_num_fault_inject 0
+set signal_fault_duration 20ns
+set register_fault_duration 0ns
+
+set allow_multi_bit_upset $::env(MULTI_BIT_UPSET)
+set use_bitwidth_as_weight 0
+set check_core_output_modification 0
+set check_core_next_state_modification 0
+set reg_to_sig_ratio 1
+
+source [file join $utils_base_path pulp_extract_nets.tcl]
+
+set inject_signals_netlist []
+set inject_register_netlist []
+set output_netlist []
+set next_state_netlist []
+set assertion_disable_list []
+
+# for {set idx 0} {$idx < 12} {incr idx} {
+#     set inject_signals_netlist [list {*}$inject_signals_netlist {*}[get_all_core_nets $idx]]
+#     set output_netlist [list {*}$output_netlist {*}[get_core_output_nets $idx]]
+# }
+
+set inject_register_netlist [list {*}$inject_register_netlist {*}[get_memory_slice {0 16} {256 336}]]
+
+source [file join $script_base_path inject_fault.tcl]
+
diff --git a/hwpe/redmule_256iter/redmule.c b/hwpe/redmule_256iter/redmule.c
new file mode 100644
index 0000000..77019ae
--- /dev/null
+++ b/hwpe/redmule_256iter/redmule.c
@@ -0,0 +1,141 @@
+/*
+ * Copyright (C) 2022-2023 ETH Zurich and University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ * SPDX-License-Identifier: Apache-2.0
+ * 
+ * Author: Yvan Tortorella  <yvan.tortorella@unibo.it>
+ *
+ * RedMulE SW test
+ */
+
+#include <stdint.h>
+#include "stdio.h"
+#include "archi_redmule.h"
+#include "hal_redmule.h"
+#include "pulp.h"
+
+#define NB_ITER 255
+
+int main() {
+
+  volatile int errors = 0;
+  unsigned int cluster_id = rt_cluster_id();
+  unsigned int intc_data_correctable_cnt, redmule_data_correctable_cnt = 0;
+  unsigned int intc_meta_correctable_cnt = 0;
+  unsigned int intc_data_uncorrectable_cnt, redmule_data_uncorrectable_cnt = 0;
+  unsigned int intc_meta_uncorrectable_cnt = 0;
+
+  if(get_core_id() == 0){
+
+    uint16_t m_size = M_SIZE;
+    uint16_t n_size = N_SIZE;
+    uint16_t k_size = K_SIZE;
+
+    uint8_t *x_ext = x_inp;
+    uint8_t *w_ext = w_inp;
+    uint8_t *y_ext = y_inp;
+    uint8_t *z_ext = z_oup;
+
+    uint8_t volatile *x = (uint8_t volatile *) pi_l1_malloc(0, (2*m_size*n_size));
+    uint8_t volatile *w = (uint8_t volatile *) pi_l1_malloc(0, (2*n_size*k_size));
+    uint8_t volatile *y = (uint8_t volatile *) pi_l1_malloc(0, (2*m_size*k_size));
+    uint8_t volatile *z = (uint8_t volatile *) pi_l1_malloc(0, (2*m_size*k_size));
+
+    #ifdef USE_DMA
+      volatile unsigned int dma_id = 0;
+      dma_id = mchan_alloc();
+      mchan_transfer((unsigned int) 2*(2*m_size*n_size),
+                     (unsigned int) x_ext,
+                     (unsigned int) x    );
+      mchan_barrier(dma_id);
+      mchan_free(dma_id);
+    
+      dma_id = mchan_alloc();
+      mchan_transfer((unsigned int) 2*(2*n_size*k_size),
+                     (unsigned int) w_ext,
+                     (unsigned int) w    );
+      mchan_barrier(dma_id);
+      mchan_free(dma_id);
+    
+      dma_id = mchan_alloc();
+      mchan_transfer((unsigned int) 2*(2*m_size*k_size),
+                     (unsigned int) y_ext,
+                     (unsigned int) y    );
+      mchan_barrier(dma_id);
+    #else
+      generate_test_data16((int) x, (int) w, (int) y, (int) m_size, (int) n_size, (int) k_size);
+    #endif
+
+    int gold_sum = 0, check_sum = 0;
+    int i,j;
+
+    int offload_id_tmp, offload_id;
+
+    // Enable RedMulE
+    hwpe_cg_enable();
+    asm volatile("": : :"memory");
+
+    hwpe_soft_clear();
+    asm volatile("": : :"memory");
+
+    volatile int job_id = -1;
+
+    // job 0
+    job_id = hwpe_wait_acquire();
+    asm volatile("": : :"memory");
+    redmule_x_add_set ((unsigned int) x);
+    redmule_w_add_set ((unsigned int) w);
+    redmule_y_add_set ((unsigned int) y);
+    redmule_z_add_set ((unsigned int) z);
+    redmule_cfg (m_size, n_size, k_size, gemm_ops);
+    asm volatile("": : :"memory");
+    hwpe_trigger_job();
+    asm volatile("": : :"memory");
+
+    // job 1
+    job_id = hwpe_wait_acquire();
+    asm volatile("": : :"memory");
+    redmule_x_add_set ((unsigned int) x);
+    redmule_w_add_set ((unsigned int) w);
+    redmule_y_add_set ((unsigned int) y);
+    redmule_z_add_set ((unsigned int) z);
+    redmule_cfg (m_size, n_size, k_size, gemm_ops);
+    asm volatile("": : :"memory");
+    hwpe_trigger_job();
+    asm volatile("": : :"memory");
+
+    // jobs 2-255
+    do {
+      job_id = hwpe_wait_acquire();
+      asm volatile("": : :"memory");
+      hwpe_trigger_job();
+      asm volatile("": : :"memory");
+    } while(job_id < NB_ITER);
+
+    // Wait for end of computation
+    redmule_evt_wait();
+
+    // Disable RedMulE
+    hwpe_cg_disable();
+
+    errors = redmule_compare16((int) z, (int) m_size, (int) k_size);
+
+    *(int *) 0x1A1040A0 = errors;
+
+    printf ("Terminated test with %d errors. See you!\n", errors);
+
+  }
+  synch_barrier();
+  return (errors != 0);
+}
diff --git a/hwpe/redmule_softclear/Makefile b/hwpe/redmule_softclear/Makefile
new file mode 100644
index 0000000..88346b6
--- /dev/null
+++ b/hwpe/redmule_softclear/Makefile
@@ -0,0 +1,20 @@
+PULP_APP = test
+PULP_APP_SRCS = redmule.c
+PULP_CFLAGS = -O3
+
+ifeq ($(use_dma),1)
+ PULP_CFLAGS += -DUSE_DMA
+endif
+
+ifeq ($(fault_inject),1)
+  export FAULT_INJECTION=1
+  export FAULT_INJECTION_SCRIPT=$(CURDIR)/pulp_inject_fault.tcl
+endif
+
+ifeq ($(multi_bit_upset),1)
+  export MULTI_BIT_UPSET=1
+else
+  export MULTI_BIT_UPSET=0
+endif
+
+include $(PULP_SDK_HOME)/install/rules/pulp_rt.mk
diff --git a/hwpe/redmule_softclear/archi_redmule.h b/hwpe/redmule_softclear/archi_redmule.h
new file mode 100644
index 0000000..40eceee
--- /dev/null
+++ b/hwpe/redmule_softclear/archi_redmule.h
@@ -0,0 +1,170 @@
+/*
+ * Copyright (C) 2022-2023 ETH Zurich and University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ * SPDX-License-Identifier: Apache-2.0
+ * 
+ * Author: Yvan Tortorella  <yvan.tortorella@unibo.it>
+ *
+ * High-level architecture of RedMulE
+ *
+ */
+
+#ifndef __ARCHI_REDMULE_H__
+#define __ARCHI_REDMULE_H__
+
+/*
+ * |========================================================================|
+ * ||                                                                      ||
+ * ||Control and generic configuration register layout                     ||
+ * |========================================================================|
+ * || # reg |  offset  |  bits   |   bitmask    ||  content                ||
+ * ||-------+----------+---------+--------------++-------------------------||
+ * ||    0  |  0x0000  |  31: 0  |  0xFFFFFFFF  ||  TRIGGER                ||
+ * ||    1  |  0x0004  |  31: 0  |  0xFFFFFFFF  ||  ACQUIRE                ||
+ * ||    2  |  0x0008  |  31: 0  |  0xFFFFFFFF  ||  EVT_ENABLE             ||
+ * ||    3  |  0x000c  |  31: 0  |  0xFFFFFFFF  ||  STATUS                 ||
+ * ||    4  |  0x0010  |  31: 0  |  0xFFFFFFFF  ||  RUNNING_JOB            ||
+ * ||    5  |  0x0014  |  31: 0  |  0xFFFFFFFF  ||  SOFT_CLEAR             ||
+ * |========================================================================|
+ * ||                                                                      ||
+ * ||Job-dependent registers layout                                        ||
+ * |========================================================================|
+ * || # reg |  offset  |  bits   |   bitmask    ||  content                ||
+ * ||-------+----------+---------+--------------++-------------------------||
+ * ||    0  |  0x0040  |  31: 0  |  0xFFFFFFFF  ||  X_ADDR                 ||
+ * ||-------+----------+---------+--------------++-------------------------||
+ * ||    1  |  0x0044  |  31: 0  |  0xFFFFFFFF  ||  W_ADDR                 ||
+ * ||-------+----------+---------+--------------++-------------------------||
+ * ||    2  |  0x0048  |  31: 0  |  0xFFFFFFFF  ||  Z_ADDR                 ||
+ * ||-------+----------+---------+--------------++-------------------------||
+ * ||    3  |  0x004C  |         |              ||  Matrix Config 0 Reg    ||
+ * ||       |          |  31:16  |  0xFFFF0000  ||  K Size (W Columns)     ||
+ * ||       |          |  15: 0  |  0x0000FFFF  ||  M Size (X Rows)        ||
+ * ||-------+----------+---------+--------------++-------------------------||
+ * ||    4  |  0x0050  |         |              ||  Matrix Config 1 Reg    ||
+ * ||       |          |  31:16  |  0xFFFFFFFF  ||  N Size (X Cols/W Rows) ||
+ * ||-------+----------+---------+--------------++-------------------------||
+ * ||    5  |  0x0054  |         |              ||  Matrix Arithmetic Reg  ||
+ * ||       |          |  12:10  |  0x00001C00  ||  Operation selection    ||
+ * ||       |          |   9: 7  |  0x00000380  ||  Input/Output format    ||
+ * |========================================================================|
+ *
+ */
+
+/* PULP Cluster Archi defines */
+#define ARCHI_CLUST_CTRL_BASE ARCHI_CLUSTER_CTRL_ADDR
+#define ARCHI_CLUST_HWPE_BASE ARCHI_HWCE_ADDR
+#define DMA_COMMAND_QUEUE     ARCHI_MCHAN_DEMUX_ADDR
+#define DMA_STATUS_REGISTER   (ARCHI_MCHAN_DEMUX_ADDR + 4)
+#define ARCHI_CL_HWPE_EVT0 12
+#define ARCHI_CL_HWPE_EVT1 13
+#define FC_DMA_EVENT 8
+#define CL_DMA_EVENT 22
+#define CLUST_CTRL_HWPE_EN 0x18
+#define CLUST_CTRL_HWPE_EN_MASK 0x800
+#define __builtin_bitinsert(a,b,c,d) (a | (((b << (32-c)) >> (32-c)) << d))
+
+// RedMulE architecture
+#define ADDR_WIDTH   32
+#define DATA_WIDTH   256
+#define REDMULE_FMT  16
+#define ARRAY_HEIGHT 4
+#define PIPE_REGS    3
+#define ARRAY_WIDTH  12 /* Superior limit is ARRAY_HEIGHT*PIPE_REGS */
+
+// Commands
+#define REDMULE_TRIGGER     0x00
+#define REDMULE_ACQUIRE     0x04
+#define REDMULE_FINISHED    0x08
+#define REDMULE_STATUS      0x0C
+#define REDMULE_RUNNING_JOB 0x10
+#define REDMULE_SOFT_CLEAR  0x14
+
+// Registers
+#define REDMULE_REG_OFFS  0x40
+// #define REDMULE_REG_X_PTR 0x00
+// #define REDMULE_REG_W_PTR 0x04
+// #define REDMULE_REG_Z_PTR 0x08
+// #define REDMULE_MCFG0_PTR 0x0C
+// #define REDMULE_MCFG1_PTR 0x10
+// #define REDMULE_ARITH_PTR 0x14
+#define REDMULE_REG_X_PTR              0x00
+#define REDMULE_REG_W_PTR              0x04
+#define REDMULE_REG_Y_PTR              0x08
+#define REDMULE_REG_Z_PTR              0x0C
+#define REDMULE_REG_X_ITER_PTR         0x10
+#define REDMULE_REG_W_ITER_PTR         0x14
+#define REDMULE_REG_LEFTOVERS_PTR      0x18
+#define REDMULE_REG_LEFT_PARAMS_PTR    0x1C
+#define REDMULE_REG_X_D1_STRIDE_PTR    0x20
+#define REDMULE_REG_W_TOT_LEN_PTR      0x24
+#define REDMULE_REG_TOT_X_READ_PTR     0x28
+#define REDMULE_REG_W_D0_STRIDE_PTR    0x2C
+#define REDMULE_REG_YZ_TOT_LEN_PTR     0x30
+#define REDMULE_REG_YZ_D0_STRIDE_PTR   0x34
+#define REDMULE_REG_YZ_D2_STRIDE_PTR   0x38
+#define REDMULE_REG_X_ROWS_OFFS_PTR    0x3C
+#define REDMULE_REG_X_BUFFER_SLOTS_PTR 0x40
+#define REDMULE_REG_X_TOT_LEN_PTR      0x44
+#define REDMULE_REG_OP_SELECTION       0x48
+
+#define REDMULE_ECC_REG_OFFS           0x90
+#define DATA_CORR_ERR                  0x00
+#define DATA_UNCORR_ERR                0x04
+#define METADATA_CORR_ERR              0x08
+#define METADATA_UNCORR_ERR            0x0c
+
+// OPs definition
+#define MATMUL 0x0
+#define GEMM   0x1
+#define ADDMAX 0x2
+#define ADDMIN 0x3
+#define MULMAX 0x4
+#define MULMIN 0x5
+#define MAXMIN 0x6
+#define MINMAX 0x7
+
+// GEMM formats
+#define Float8     0x0
+#define Float16    0x1
+#define Float8Alt  0x2
+#define Float16Alt 0x3
+
+#define RNE       0x0
+#define RTZ       0x1
+#define OP_FMADD  0x0
+#define OP_ADD    0x2
+#define OP_MUL    0x3
+#define OP_MINMAX 0x7
+
+// FP Formats encoding
+#define FP16    0x2
+#define FP8     0x3
+#define FP16ALT 0x4
+#define FP8ALT  0x5
+
+/* DMA Archi */
+#define DMA_TX  0
+#define DMA_RX  1
+#define DMA_INC 1
+
+#define PLP_DMA_TYPE_BIT    0x00000011
+#define PLP_DMA_INCR_BIT    0x00000012
+#define PLP_DMA_2D_BIT      0x00000013
+#define PLP_DMA_ELE_BIT     0x00000014
+#define PLP_DMA_ILE_BIT     0x00000015
+#define PLP_DMA_BLE_BIT     0x00000016
+#define PLP_DMA_2D_TCDM_BIT 0x0000017
+
+#endif
diff --git a/hwpe/redmule_softclear/hal_redmule.h b/hwpe/redmule_softclear/hal_redmule.h
new file mode 100644
index 0000000..8fc5000
--- /dev/null
+++ b/hwpe/redmule_softclear/hal_redmule.h
@@ -0,0 +1,556 @@
+/*
+ * Copyright (C) 2022-2023 ETH Zurich and University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ * SPDX-License-Identifier: Apache-2.0
+ * 
+ * Author: Yvan Tortorella  <yvan.tortorella@unibo.it>
+ *
+ * RedMulE Hardware Abstraction Layer (HAL)
+ */
+
+#ifndef __HAL_REDMULE_H__
+#define __HAL_REDMULE_H__
+
+#include <pulp.h>
+#include "inc/x_input.h"
+#include "inc/w_input.h"
+#include "inc/y_input.h"
+#include "inc/z_output.h"
+#include "inc/golden.h"
+#include "inc/tensor_dim.h"
+
+/*
+ * 
+ * For control, generic configuration register layout,
+ * and job-dependent register map, look at redmule_archi.h
+ *
+ */
+
+// For all the following functions we use __builtin_pulp_OffsetedWrite and __builtin_pulp_OffsetedRead
+// instead of classic load/store because otherwise the compiler is not able to correctly factorize
+// the HWPE base in case several accesses are done, ending up with twice more code
+
+#define HWPE_WRITE(value, offset) *(int *)(ARCHI_CLUST_HWPE_BASE + offset) = value
+#define HWPE_READ(offset) *(int *)(ARCHI_CLUST_HWPE_BASE + offset)
+
+static inline void redmule_x_add_set (unsigned int value) {
+  HWPE_WRITE(value, REDMULE_REG_OFFS + REDMULE_REG_X_PTR);
+}
+
+static inline void redmule_w_add_set (unsigned int value) {
+  HWPE_WRITE(value, REDMULE_REG_OFFS + REDMULE_REG_W_PTR);
+}
+
+static inline void redmule_y_add_set (unsigned int value) {
+  HWPE_WRITE(value, REDMULE_REG_OFFS + REDMULE_REG_Y_PTR);
+}
+
+static inline void redmule_z_add_set (unsigned int value) {
+  HWPE_WRITE(value, REDMULE_REG_OFFS + REDMULE_REG_Z_PTR);
+}
+
+// static inline void redmule_mcfg_set (uint32_t mcfg0, uint32_t mcfg1) {
+//   HWPE_WRITE(mcfg0, REDMULE_REG_OFFS + REDMULE_MCFG0_PTR);
+//   HWPE_WRITE(mcfg1, REDMULE_REG_OFFS + REDMULE_MCFG1_PTR);
+// }
+// 
+// static inline void redmule_arith_set (uint32_t arith) {
+//   HWPE_WRITE(arith, REDMULE_REG_OFFS + REDMULE_ARITH_PTR);
+// }
+
+static inline void hwpe_trigger_job() {
+  HWPE_WRITE(0, REDMULE_TRIGGER);
+}
+
+static inline int hwpe_acquire_job() {
+  return HWPE_READ(REDMULE_ACQUIRE);
+}
+
+static inline unsigned int hwpe_get_status() {
+  return HWPE_READ(REDMULE_STATUS);
+}
+
+static inline unsigned int hwpe_get_running_job() {
+  return HWPE_READ(REDMULE_RUNNING_JOB);
+}
+
+static inline void hwpe_soft_clear() {
+  HWPE_WRITE(0, REDMULE_SOFT_CLEAR);
+}
+
+static inline void hwpe_cg_enable() {
+  *(volatile int*) (ARCHI_CLUST_CTRL_BASE + CLUST_CTRL_HWPE_EN) |= CLUST_CTRL_HWPE_EN_MASK;
+}
+
+static inline void hwpe_cg_disable() {
+  *(volatile int*) (ARCHI_CLUST_CTRL_BASE + CLUST_CTRL_HWPE_EN) &= ~CLUST_CTRL_HWPE_EN_MASK;
+}
+
+static inline void redmule_evt_wait() {
+  do {
+    eu_evt_maskWaitAndClr (1 << ARCHI_CL_HWPE_EVT0);
+  } while((*(int volatile *)(ARCHI_CLUST_HWPE_BASE + REDMULE_STATUS)) != 0);
+}
+
+static inline int hwpe_wait_acquire() {
+  int job_id = hwpe_acquire_job();
+  while(job_id < 0) {
+    eu_evt_maskWaitAndClr (1 << ARCHI_CL_HWPE_EVT0);
+    job_id = hwpe_acquire_job();
+  }
+  return job_id;
+}
+
+static inline unsigned int redmule_get_data_correctable_count () {
+  return HWPE_READ(REDMULE_ECC_REG_OFFS + DATA_CORR_ERR);
+}
+
+static inline unsigned int redmule_get_data_uncorrectable_count () {
+  return HWPE_READ(REDMULE_ECC_REG_OFFS + DATA_UNCORR_ERR);
+}
+
+static inline unsigned int redmule_get_meta_correctable_count () {
+  return HWPE_READ(REDMULE_ECC_REG_OFFS + METADATA_CORR_ERR);
+}
+
+static inline unsigned int redmule_get_meta_uncorrectable_count () {
+  return HWPE_READ(REDMULE_ECC_REG_OFFS + METADATA_UNCORR_ERR);
+}
+
+/* DMA APIs */
+static inline int mchan_alloc(){
+  return *(volatile int*) DMA_COMMAND_QUEUE;
+}
+
+static inline void mchan_transfer(unsigned int len,
+                                  unsigned int ext_addr,
+                                  unsigned int tcdm_addr) {
+
+  *(volatile int*) DMA_COMMAND_QUEUE = len |
+                                       (DMA_RX << PLP_DMA_TYPE_BIT)  |
+                                       (DMA_INC << PLP_DMA_INCR_BIT) |
+                                       (0 << PLP_DMA_2D_BIT)         |
+                                       (1 << PLP_DMA_ELE_BIT)        |
+                                       (1 << PLP_DMA_ILE_BIT)        |
+                                       (0 << PLP_DMA_BLE_BIT)        |
+                                       (0 << PLP_DMA_2D_TCDM_BIT);
+  *(volatile int*) DMA_COMMAND_QUEUE = tcdm_addr;
+  *(volatile int*) DMA_COMMAND_QUEUE = ext_addr;
+}
+
+static inline void mchan_barrier(int id) {
+  while(((*(volatile int*)(DMA_STATUS_REGISTER)) >> id ) & 0x1 ) {
+    eu_evt_maskWaitAndClr(1 << FC_DMA_EVENT);
+ }
+}
+
+static inline void mchan_free(int id) {
+  *(volatile int*) DMA_STATUS_REGISTER = 0x1 << id;
+}
+
+// void redmule_cfg (unsigned int x,  unsigned int w,  unsigned int z,
+//                   uint16_t m_size, uint16_t n_size, uint16_t k_size,
+//                   uint8_t gemm_op, uint8_t gemm_fmt){
+// 
+//   uint32_t mcfg_reg0 = 0;
+//   uint32_t mcfg_reg1 = 0;
+//   uint32_t arith_reg = 0;
+// 
+//   mcfg_reg0 = (k_size << 16) |
+//               (m_size <<  0);
+//   mcfg_reg1 = n_size << 0;
+// 
+//   arith_reg = (gemm_op << 10) |
+//               (gemm_fmt << 7);
+// 
+//   redmule_x_add_set ((unsigned int) x);
+//   redmule_w_add_set ((unsigned int) w);
+//   redmule_z_add_set ((unsigned int) z);
+//   redmule_mcfg_set  ((unsigned int) mcfg_reg0,
+//                      (unsigned int) mcfg_reg1);
+//   redmule_arith_set ((unsigned int) arith_reg);
+//   
+// }
+
+void redmule_cfg (uint16_t m_size, uint16_t n_size, uint16_t k_size, uint8_t gemm_ops){
+   uint32_t x_iters        = 0;
+   uint32_t w_iters        = 0;
+   uint32_t leftovers      = 0;
+   uint32_t left_params    = 0;
+   uint32_t x_d1_stride    = 0;
+   uint32_t x_rows_offs    = 0;
+   uint32_t w_tot_len      = 0;
+   uint32_t w_d1_len       = 0;
+   uint32_t w_d0_stride    = 0;
+   uint32_t yz_tot_len     = 0;
+   uint32_t yz_d0_stride   = 0;
+   uint32_t yz_d2_stride   = 0;
+   uint32_t tot_x_read     = 0;
+   uint32_t x_buffer_slots = 0;
+   uint32_t op_selection   = 0;
+   uint16_t tot_stores     = 0;
+   uint16_t w_rows         = n_size;
+   uint16_t depth          = DATA_WIDTH/(ARRAY_HEIGHT*FPFORMAT);
+   uint8_t  tile           = ARRAY_HEIGHT*(PIPE_REGS + 1);
+   _Bool    x_rows_sub     = 0;
+   _Bool    x_cols_sub     = 0;
+   _Bool    w_cols_sub     = 0;
+   uint16_t x_rows_iter,
+            x_rows_iter_tmp,
+            w_rows_iter,
+            w_rows_iter_tmp;
+   uint16_t x_cols_iter,
+            x_cols_iter_tmp,
+            w_cols_iter,
+            w_cols_iter_tmp;
+   uint8_t  x_rows_lftovr,
+            x_cols_lftovr,
+            w_rows_lftovr,
+            w_cols_lftovr,
+            slots;
+
+   // Calculating the number of iterations alng the two dimensions of the X matrix
+   x_rows_iter_tmp = m_size/ARRAY_WIDTH;
+   x_cols_iter_tmp = n_size/tile;
+
+   // Calculating the number of iterations alng the two dimensions of the W matrix
+   w_rows_iter_tmp = w_rows;
+   w_cols_iter_tmp = k_size/tile;
+
+   // Calculating the residuals along the input dimensions
+   x_rows_lftovr = m_size - (x_rows_iter_tmp*ARRAY_WIDTH);
+   x_cols_lftovr = n_size - (x_cols_iter_tmp*tile);
+
+   // Calculating the residuals along the weight dimensions
+   w_rows_lftovr = n_size - (ARRAY_HEIGHT*(w_rows/ARRAY_HEIGHT));
+   w_cols_lftovr = k_size - (w_cols_iter_tmp*tile);
+
+   if (w_cols_lftovr != 0)
+     w_cols_iter = w_cols_iter_tmp + 1;
+   else 
+     w_cols_iter = w_cols_iter_tmp;
+
+  if (w_rows_lftovr != 0)
+    w_rows_iter = w_rows_iter_tmp + ARRAY_HEIGHT - w_rows_lftovr;
+  else
+    w_rows_iter = w_rows_iter_tmp;
+
+   if (x_cols_lftovr != 0)
+     x_cols_iter = x_cols_iter_tmp + 1;
+   else 
+     x_cols_iter = x_cols_iter_tmp;
+
+   if (x_rows_lftovr != 0)
+     x_rows_iter = x_rows_iter_tmp + 1;
+   else 
+     x_rows_iter = x_rows_iter_tmp;
+
+   if (x_cols_lftovr%depth != 0)
+     x_buffer_slots = x_cols_lftovr/depth + 1;
+   else
+     x_buffer_slots = x_cols_lftovr/depth;
+
+   // Calculating the number of total stores
+   tot_stores = x_rows_iter*w_cols_iter;
+
+   // Determining if input matrixes are sub-matrixes
+   if (m_size < ARRAY_WIDTH)
+     x_rows_sub = 1;
+   if (n_size < ARRAY_HEIGHT)
+     x_cols_sub = 1;
+   if (k_size < tile)
+    w_cols_sub = 1;
+
+   // Operation selection
+   switch (gemm_ops) {
+     case MATMUL:
+       op_selection |= (RNE << 29 | RNE << 26 | OP_FMADD << 22 | OP_MINMAX << 18 | SRC_FMT << 15 | DST_FMT << 12) | 0;
+     break;
+     
+     case GEMM:
+       op_selection |= (RNE << 29 | RNE << 26 | OP_FMADD << 22 | OP_MINMAX << 18 | SRC_FMT << 15 | DST_FMT << 12) | 1;
+     break;
+     
+     case ADDMAX:
+       op_selection |= (RNE << 29 | RTZ << 26 | OP_ADD << 22 | OP_MINMAX << 18 | SRC_FMT << 15 | DST_FMT << 12) | 1;
+     break;
+     
+     case ADDMIN:
+       op_selection |= (RNE << 29 | RNE << 26 | OP_ADD << 22 | OP_MINMAX << 18 | SRC_FMT << 15 | DST_FMT << 12) | 1;
+     break;
+     
+     case MULMAX:
+       op_selection |= (RNE << 29 | RTZ << 26 | OP_MUL << 22 | OP_MINMAX << 18 | SRC_FMT << 15 | DST_FMT << 12) | 1;
+     break;
+     
+     case MULMIN:
+       op_selection |= (RNE << 29 | RNE << 26 | OP_MUL << 22 | OP_MINMAX << 18 | SRC_FMT << 15 | DST_FMT << 12) | 1;
+     break;
+     
+     case MAXMIN:
+       op_selection |= (RTZ << 29 | RNE << 26 | OP_MINMAX << 22 | OP_MINMAX << 18 | SRC_FMT << 15 | DST_FMT << 12) | 1;
+     break;
+     
+     case MINMAX:
+       op_selection |= (RNE << 29 | RTZ << 26 | OP_MINMAX << 22 | OP_MINMAX << 18 | SRC_FMT << 15 | DST_FMT << 12) | 1;
+     break;
+   }
+
+   // Storing iterations and residuals in registers
+   x_iters      |= x_rows_iter   << 16 | x_cols_iter   << 0;
+   w_iters      |= w_rows_iter   << 16 | w_cols_iter   << 0;
+   leftovers    |= x_rows_lftovr << 24 | x_cols_lftovr << 16 | w_rows_lftovr << 8  | w_cols_lftovr << 0;
+   left_params  |= tot_stores    << 16 | x_rows_sub    << 15 | x_cols_sub    << 14 | w_cols_sub   << 13;
+   x_d1_stride   = ((4*FPFORMAT)/ADDR_WIDTH)*(((DATA_WIDTH/FPFORMAT)*x_cols_iter_tmp) + x_cols_lftovr);
+   x_rows_offs   = ARRAY_WIDTH*x_d1_stride;
+   w_tot_len     = w_rows_iter*w_cols_iter*x_rows_iter;
+   w_d0_stride   = ((4*FPFORMAT)/ADDR_WIDTH)*(((DATA_WIDTH/FPFORMAT)*w_cols_iter_tmp) + w_cols_lftovr);
+   yz_tot_len    = ARRAY_WIDTH*x_rows_iter*w_cols_iter;
+   yz_d0_stride  = w_d0_stride;
+   yz_d2_stride  = ARRAY_WIDTH*w_d0_stride;
+   tot_x_read    = x_rows_iter*x_cols_iter*w_cols_iter;
+
+   // Writing the computations in configuration register
+   HWPE_WRITE(x_iters       , REDMULE_REG_OFFS + REDMULE_REG_X_ITER_PTR         );
+   HWPE_WRITE(w_iters       , REDMULE_REG_OFFS + REDMULE_REG_W_ITER_PTR         );
+   HWPE_WRITE(leftovers     , REDMULE_REG_OFFS + REDMULE_REG_LEFTOVERS_PTR      );
+   HWPE_WRITE(left_params   , REDMULE_REG_OFFS + REDMULE_REG_LEFT_PARAMS_PTR    );
+   HWPE_WRITE(x_d1_stride   , REDMULE_REG_OFFS + REDMULE_REG_X_D1_STRIDE_PTR    );
+   HWPE_WRITE(x_rows_offs   , REDMULE_REG_OFFS + REDMULE_REG_X_ROWS_OFFS_PTR    );
+   HWPE_WRITE(tot_x_read    , REDMULE_REG_OFFS + REDMULE_REG_TOT_X_READ_PTR     );
+   HWPE_WRITE(x_buffer_slots, REDMULE_REG_OFFS + REDMULE_REG_X_BUFFER_SLOTS_PTR );
+   HWPE_WRITE(w_tot_len     , REDMULE_REG_OFFS + REDMULE_REG_W_TOT_LEN_PTR      );
+   HWPE_WRITE(w_d0_stride   , REDMULE_REG_OFFS + REDMULE_REG_W_D0_STRIDE_PTR    );
+   HWPE_WRITE(yz_tot_len    , REDMULE_REG_OFFS + REDMULE_REG_YZ_TOT_LEN_PTR     );
+   HWPE_WRITE(yz_d0_stride  , REDMULE_REG_OFFS + REDMULE_REG_YZ_D0_STRIDE_PTR   );
+   HWPE_WRITE(yz_d2_stride  , REDMULE_REG_OFFS + REDMULE_REG_YZ_D2_STRIDE_PTR   );
+   HWPE_WRITE(op_selection  , REDMULE_REG_OFFS + REDMULE_REG_OP_SELECTION       );
+}
+
+void generate_test_data16(int x_start_addr,
+                          int w_start_addr,
+                          int y_start_addr,
+                          int m_size,
+                          int n_size,
+                          int k_size) {
+  
+  int x_addr     = x_start_addr;
+  int w_addr     = w_start_addr;
+  int y_addr     = y_start_addr;
+  int x_end_addr = x_start_addr + (2*m_size*n_size);
+  int w_end_addr = w_start_addr + (2*n_size*k_size);
+  int y_end_addr = y_start_addr + (2*m_size*k_size);
+
+  // Generating input stimuli from golden model
+  for (x_addr = x_start_addr; x_addr < x_end_addr; x_addr += 2) {
+    int x = x_addr - x_start_addr;
+    *(uint32_t *)(x_addr) = x_inp[x/2];
+  }
+
+  // Generating Weight stimuli from golden model
+  for (w_addr = w_start_addr; w_addr < w_end_addr; w_addr += 2) {
+    int w = w_addr - w_start_addr;
+    *(uint32_t *)(w_addr) = w_inp[w/2];
+  }
+
+  for (y_addr = y_start_addr; y_addr < y_end_addr; y_addr += 2) {
+    int y = y_addr - y_start_addr;
+    *(uint32_t *)(y_addr) = y_inp[y/2];
+  }
+}
+
+int redmule_compare16 (int z_start_addr, int m_size, int k_size) {
+  int err = 0;
+  int z_end_addr = z_start_addr + 2*m_size*k_size;
+  uint16_t z_computed;
+  uint16_t diff, diff_1, diff_2;
+
+  for (int z_addr = z_start_addr; z_addr < z_end_addr; z_addr += 2) {
+    int z = z_addr - z_start_addr;
+    z_computed = *(uint32_t *)(z_addr);
+
+    if ( z_computed != z_oup[z/2] ) {
+      diff_1 = z_computed - z_oup[z/2];
+      if (diff_1 > 3) {
+        diff_2 = z_oup[z/2] - z_computed;
+        if (diff_2 > 3) {
+          err++;
+        }
+      }
+    }
+  }
+
+  return err;
+
+}
+
+int redmule16_compare_int(uint32_t *actual_z, uint32_t *golden_z, int len) {
+  #define ERR 0x0011
+  uint32_t actual_word = 0;
+  uint16_t actual_MSHWord, actual_LSHWord;
+  uint32_t golden_word = 0;
+  uint16_t golden_MSHWord, golden_LSHWord;
+  uint32_t actual = 0;
+  uint32_t golden = 0;
+
+  int errors = 0;
+  int error;
+
+  for (int i=0; i<len; i++) {
+    error = 0;
+    actual_word = *(actual_z+i);
+    golden_word = *(golden_z+i);
+
+    uint16_t diff = 0;
+    
+    // Chechink Least Significant Half-Word
+    actual_LSHWord = (uint16_t)(actual_word & 0x0000FFFF);
+    golden_LSHWord = (uint16_t)(golden_word & 0x0000FFFF);
+
+    diff = (actual_LSHWord > golden_LSHWord) ? (actual_LSHWord - golden_LSHWord) : 0;
+    diff = (actual_LSHWord < golden_LSHWord) ? (golden_LSHWord - actual_LSHWord) : 0;
+
+    if (diff > ERR) {
+      error = 1;
+      #ifdef VERBOSE
+        tfp_printf ("diff: 0x%08x\n", diff);
+        tfp_printf ("LSW: Error!\n");
+      #endif
+    }
+
+    // Checking Most Significant Half-Word
+    actual_MSHWord = (uint16_t)((actual_word >> 16) & 0x0000FFFF);
+    golden_MSHWord = (uint16_t)((golden_word >> 16) & 0x0000FFFF);
+
+    diff = (actual_MSHWord > golden_MSHWord) ? (actual_MSHWord - golden_MSHWord) : 0;
+    diff = (actual_MSHWord < golden_MSHWord) ? (golden_MSHWord - actual_MSHWord) : 0;
+
+    if (diff > ERR) {
+      error = 1;
+      #ifdef VERBOSE
+        tfp_printf ("diff: 0x%08x\n", diff);
+        tfp_printf ("MSW: Error!\n");
+      #endif
+    }
+    
+    errors += error;
+
+    #ifdef DEBUG
+      tfp_printf("  Golden: 0x%08x; Actual: 0x%08x,\n", golden_word, actual_word);
+    #endif
+
+    #ifdef VERBOSE
+    if(error) {
+        if(errors==1) tfp_printf("  golden     <- actual     @ address    @ index\n");
+        tfp_printf("  0x%08x <- 0x%08x @ 0x%08x @ 0x%08x\n", golden_word, actual_word, (actual_z+i), i*4);
+    }
+    #endif
+  }
+  return errors;
+}
+
+int redmule8_compare_int(uint32_t *actual_z, uint32_t *golden_z, int len) {
+  #define ERR 0x0011
+  uint32_t actual_word = 0;
+  uint8_t  actual_Byte0,
+           actual_Byte1,
+           actual_Byte2,
+           actual_Byte3;
+  uint32_t golden_word = 0;
+  uint8_t  golden_Byte0,
+           golden_Byte1,
+           golden_Byte2,
+           golden_Byte3;
+  uint32_t actual = 0;
+  uint32_t golden = 0;
+
+  int errors = 0;
+  int error;
+
+  for (int i=0; i<len; i++) {
+    error = 0;
+    actual_word = *(actual_z+i);
+    golden_word = *(golden_z+i);
+
+    uint8_t diff = 0;
+    
+    // Cheching Byte0
+    actual_Byte0 = (uint8_t)(actual_word & 0x000000FF);
+    golden_Byte0 = (uint8_t)(golden_word & 0x000000FF);
+
+    diff = (actual_Byte0 > golden_Byte0) ? (actual_Byte0 - golden_Byte0) : 0;
+    diff = (actual_Byte0 < golden_Byte0) ? (golden_Byte0 - actual_Byte0) : 0;
+
+    if (diff > ERR) {
+      error = 1;
+      tfp_printf ("diff: 0x%08x\n", diff);
+      tfp_printf ("Byte0: Error!\n");
+    }
+
+    // Cheching Byte1
+    actual_Byte1 = (uint8_t)( (actual_word >> 8 ) & 0x000000FF);
+    golden_Byte1 = (uint8_t)( (golden_word >> 8 ) & 0x000000FF);
+
+    diff = (actual_Byte1 > golden_Byte1) ? (actual_Byte1 - golden_Byte1) : 0;
+    diff = (actual_Byte1 < golden_Byte1) ? (golden_Byte1 - actual_Byte1) : 0;
+
+    if (diff > ERR) {
+      error = 1;
+      tfp_printf ("diff: 0x%08x\n", diff);
+      tfp_printf ("Byte1: Error!\n");
+    }
+
+    // Cheching Byte2
+    actual_Byte2 = (uint8_t)( (actual_word >> 16 ) & 0x000000FF);
+    golden_Byte2 = (uint8_t)( (golden_word >> 16 ) & 0x000000FF);
+
+    diff = (actual_Byte2 > golden_Byte2) ? (actual_Byte2 - golden_Byte2) : 0;
+    diff = (actual_Byte2 < golden_Byte2) ? (golden_Byte2 - actual_Byte2) : 0;
+
+    if (diff > ERR) {
+      error = 1;
+      tfp_printf ("diff: 0x%08x\n", diff);
+      tfp_printf ("Byte2: Error!\n");
+    }
+
+    // Cheching Byte3
+    actual_Byte3 = (uint8_t)( (actual_word >> 24 ) & 0x000000FF);
+    golden_Byte3 = (uint8_t)( (golden_word >> 24 ) & 0x000000FF);
+
+    diff = (actual_Byte3 > golden_Byte3) ? (actual_Byte3 - golden_Byte3) : 0;
+    diff = (actual_Byte3 < golden_Byte3) ? (golden_Byte3 - actual_Byte3) : 0;
+
+    if (diff > ERR) {
+      error = 1;
+      tfp_printf ("diff: 0x%08x\n", diff);
+      tfp_printf ("Byte3: Error!\n");
+    }
+    
+    errors += error;
+
+    #ifdef DEBUG
+      tfp_printf("  Golden: 0x%08x; Actual: 0x%08x,\n", golden_word, actual_word);
+    #endif
+
+    #ifdef VERBOSE
+      if(error) {
+        if(errors==1) tfp_printf("  golden     <- actual     @ address    @ index\n");
+        tfp_printf("  0x%08x <- 0x%08x @ 0x%08x @ 0x%08x\n", golden_word, actual_word, (actual_z+i), i*4);
+      }
+    #endif
+  }
+  return errors;
+}
+
+#endif
diff --git a/hwpe/redmule_softclear/inc/golden.h b/hwpe/redmule_softclear/inc/golden.h
new file mode 100644
index 0000000..f664e47
--- /dev/null
+++ b/hwpe/redmule_softclear/inc/golden.h
@@ -0,0 +1,387 @@
+ /* Header file generated by RedMulE Golden Model */
+uint32_t golden [384] = {
+0x48974845,
+0x48384608,
+0x487b4855,
+0x48804869,
+0x48b046d1,
+0x483f48db,
+0x485f48c9,
+0x483a4881,
+0x472c484b,
+0x492b4762,
+0x48fd4822,
+0x492e488e,
+0x484f483e,
+0x46d749e8,
+0x489d484b,
+0x47e9490b,
+0x47d2484f,
+0x474744be,
+0x46c047c7,
+0x48af4727,
+0x482d46c5,
+0x482e483d,
+0x479f4897,
+0x4749488b,
+0x46a8489a,
+0x488b46f2,
+0x47e84891,
+0x483d4872,
+0x46fd4716,
+0x46a049b5,
+0x47a446e7,
+0x476748a1,
+0x49354939,
+0x48c14703,
+0x48bd4863,
+0x48cf4913,
+0x48b848b6,
+0x49204946,
+0x48e1495e,
+0x48b24938,
+0x4882493a,
+0x49d5483b,
+0x49724911,
+0x49df496b,
+0x488848f2,
+0x48214a46,
+0x490c48c1,
+0x48a349b2,
+0x47b0463a,
+0x476244cb,
+0x46b94765,
+0x4814466a,
+0x47964631,
+0x474b4666,
+0x47044798,
+0x47614838,
+0x459047d3,
+0x48a245ea,
+0x484447f1,
+0x4776484b,
+0x46d847d6,
+0x44d348f3,
+0x478d46fa,
+0x466e481e,
+0x481e4827,
+0x479445a2,
+0x48064727,
+0x48d5475d,
+0x48284708,
+0x480d4862,
+0x48324895,
+0x47f148bd,
+0x46a7482a,
+0x492d47b1,
+0x4884484d,
+0x485f48dc,
+0x480c476d,
+0x46d348e9,
+0x48844728,
+0x480e48a0,
+0x48134862,
+0x485a4675,
+0x473847e8,
+0x48234836,
+0x482146e7,
+0x47b34822,
+0x48554846,
+0x47174863,
+0x47c14872,
+0x488e46d5,
+0x485f47e2,
+0x48b8487c,
+0x4788481e,
+0x467748bd,
+0x47f846c9,
+0x47fc48fe,
+0x47b247a0,
+0x467e4588,
+0x46c74662,
+0x481246e8,
+0x474e4536,
+0x468f46c0,
+0x4679481f,
+0x46e246a1,
+0x45604809,
+0x47eb4630,
+0x475746b5,
+0x477f4848,
+0x46d846a6,
+0x459a4870,
+0x46784670,
+0x468c47d2,
+0x48c44762,
+0x479146e3,
+0x486d46b1,
+0x486747d0,
+0x47f6468d,
+0x475648a5,
+0x48544857,
+0x48384866,
+0x46ec484d,
+0x48f647d2,
+0x4879484a,
+0x483c4848,
+0x4806471d,
+0x473048fa,
+0x47b84768,
+0x46f94865,
+0x491848a8,
+0x486746ca,
+0x48624800,
+0x491048d3,
+0x4849474e,
+0x486b48eb,
+0x48c54966,
+0x483048f4,
+0x477848f9,
+0x499e481e,
+0x48f148cf,
+0x49234982,
+0x47cf487c,
+0x464949ea,
+0x495e4773,
+0x483f48b2,
+0x497548a7,
+0x481e4616,
+0x4866481f,
+0x486448b6,
+0x487347dc,
+0x487f485c,
+0x491f4938,
+0x48b6490d,
+0x48a148f8,
+0x492d4859,
+0x4915489c,
+0x48874899,
+0x4859486c,
+0x471e49ca,
+0x49184867,
+0x482748d3,
+0x4998488b,
+0x481d4704,
+0x488048b8,
+0x49444876,
+0x48f2470c,
+0x489b48b9,
+0x48e54956,
+0x48a548d6,
+0x485648dc,
+0x49ab484e,
+0x490e48e0,
+0x494548dd,
+0x48dd488b,
+0x47ea4a32,
+0x49114835,
+0x48194965,
+0x481e460e,
+0x4673452c,
+0x4717475c,
+0x46d046f6,
+0x46bc4696,
+0x481e4726,
+0x46ea4763,
+0x475846fe,
+0x4627478b,
+0x483f4704,
+0x47b146ad,
+0x48164792,
+0x468446f2,
+0x45a84827,
+0x47a4472f,
+0x462b4797,
+0x48ab483f,
+0x4863468f,
+0x4766485a,
+0x48cb481d,
+0x490347dc,
+0x483048fc,
+0x483e48cc,
+0x486448ab,
+0x47634966,
+0x499d4794,
+0x488b488e,
+0x496048dc,
+0x484c4854,
+0x474c499c,
+0x48bc4826,
+0x48834949,
+0x4905489d,
+0x481e4718,
+0x48f448e3,
+0x490448c1,
+0x48b347e8,
+0x48d44892,
+0x489448ff,
+0x488648d5,
+0x480348fa,
+0x492e47d2,
+0x48b24870,
+0x492b48e5,
+0x4785487b,
+0x471d49e3,
+0x48bf4837,
+0x48c4489b,
+0x4871475c,
+0x4811464a,
+0x471c47af,
+0x48174817,
+0x484e463b,
+0x464f477f,
+0x487c4704,
+0x472547a3,
+0x462a4853,
+0x4860465a,
+0x48804736,
+0x482b47e1,
+0x46c04811,
+0x475d48dc,
+0x48064668,
+0x46f44893,
+0x49594858,
+0x487b463d,
+0x484e480f,
+0x48a648c0,
+0x48944847,
+0x484a48a0,
+0x48f4491e,
+0x48b548fc,
+0x47d248ce,
+0x497f47db,
+0x49394955,
+0x48ce48a7,
+0x48844890,
+0x476349d6,
+0x4922486e,
+0x48c348f4,
+0x491c47ec,
+0x47834698,
+0x47544715,
+0x47524745,
+0x4832472f,
+0x48094817,
+0x48c347f8,
+0x480047e6,
+0x473048b6,
+0x48cb480a,
+0x488e479e,
+0x488e47c2,
+0x47ee472f,
+0x4744489d,
+0x48514755,
+0x47d34846,
+0x48a04838,
+0x47624634,
+0x48064786,
+0x482d47e3,
+0x486c4726,
+0x480347b7,
+0x481448ac,
+0x483948e0,
+0x47504827,
+0x48c546f2,
+0x4886483f,
+0x485648ad,
+0x47a947e8,
+0x47434937,
+0x481f46d0,
+0x4804484c,
+0x481f47fd,
+0x4813456d,
+0x4807474d,
+0x480e4688,
+0x481046e8,
+0x4799469f,
+0x478f4853,
+0x482447f2,
+0x471f47d0,
+0x485f46da,
+0x481c4813,
+0x4863482e,
+0x480b4786,
+0x46b848c9,
+0x46e2475a,
+0x46c54852,
+0x480245af,
+0x46c24466,
+0x4743465d,
+0x47ba46b7,
+0x46c34636,
+0x47844677,
+0x47c2485a,
+0x46ac46dc,
+0x460e47de,
+0x4834465f,
+0x476947f4,
+0x481046fc,
+0x45ea45fd,
+0x45b548d0,
+0x47834704,
+0x46c44830,
+0x47c74759,
+0x45b0453d,
+0x47024741,
+0x47934736,
+0x47ba461b,
+0x46dd470b,
+0x470b4657,
+0x4710470d,
+0x468f486c,
+0x46ba45c3,
+0x483b479d,
+0x477446c9,
+0x46a746a9,
+0x46064833,
+0x46a94690,
+0x46a746f5,
+0x48bb47ac,
+0x4803452c,
+0x4824470f,
+0x48cb47d5,
+0x484a4707,
+0x47974832,
+0x482c4851,
+0x4877487a,
+0x465d4891,
+0x48ce47f4,
+0x48994898,
+0x486a484e,
+0x47f047ac,
+0x4611493e,
+0x489e47e2,
+0x46af488c,
+0x48364665,
+0x46b645e4,
+0x46b946a1,
+0x46dd46c8,
+0x474b4658,
+0x4777467b,
+0x47984769,
+0x475e4785,
+0x4656472a,
+0x488145fb,
+0x472d46fc,
+0x47a3476e,
+0x46ca465d,
+0x45004855,
+0x479a464f,
+0x473846c3,
+0x486c481e,
+0x48014659,
+0x477a4756,
+0x487b47d5,
+0x48084706,
+0x4838484f,
+0x48634870,
+0x480648d3,
+0x47714865,
+0x494c46be,
+0x484c4915,
+0x48624900,
+0x46e8481a,
+0x46a04974,
+0x483d4775,
+0x480e487c,
+};
\ No newline at end of file
diff --git a/hwpe/redmule_softclear/inc/tensor_dim.h b/hwpe/redmule_softclear/inc/tensor_dim.h
new file mode 100644
index 0000000..21bd0d8
--- /dev/null
+++ b/hwpe/redmule_softclear/inc/tensor_dim.h
@@ -0,0 +1,13 @@
+ /* Header file generated by RedMulE Golden Model */
+#ifndef __TENSOR_DIM__
+#define __TENSOR_DIM__
+
+#define M_SIZE  24 
+#define N_SIZE  32 
+#define K_SIZE  32
+#define SRC_FMT FP16
+#define DST_FMT FP16
+#define FPFORMAT 16
+uint8_t gemm_ops = GEMM; 
+
+#endif
diff --git a/hwpe/redmule_softclear/inc/w_2D.h b/hwpe/redmule_softclear/inc/w_2D.h
new file mode 100644
index 0000000..9409c64
--- /dev/null
+++ b/hwpe/redmule_softclear/inc/w_2D.h
@@ -0,0 +1,35 @@
+ /* Header file generated by RedMulE Golden Model */
+uint16_t w_inp_2D [32][32] = {
+0x311a, 0x39e0, 0x387d, 0x3a4a, 0x386f, 0x3ada, 0x392f, 0x3854, 0x3014, 0x2fd2, 0x31c9, 0x2fca, 0x2e55, 0x3bc8, 0x396d, 0x3b1d, 0x39f6, 0x333a, 0x3908, 0x3628, 0x3bab, 0x3b8b, 0x3b4a, 0x322d, 0x3925, 0x317a, 0x3725, 0x31c2, 0x3066, 0x38f3, 0x3a17, 0x3476, 
+0x3bda, 0x3196, 0x3922, 0x3680, 0x396a, 0x3021, 0x3761, 0x374d, 0x2fc2, 0x3967, 0x3b94, 0x33b5, 0x3797, 0x34d6, 0x3655, 0x2176, 0x39bc, 0x3999, 0x3658, 0x3904, 0x3759, 0x2ade, 0x3a5a, 0x3b78, 0x36c7, 0x2d01, 0x3b58, 0x2d9a, 0x373d, 0x3952, 0x38e8, 0x3887, 
+0x37b6, 0x3a88, 0x2f8a, 0x2d79, 0x3413, 0x3421, 0x3976, 0x32b2, 0x3446, 0x2d99, 0x3a56, 0x3322, 0x3b49, 0x39fa, 0x3acd, 0x3af6, 0x304c, 0x3abb, 0x3a83, 0x38b2, 0x3ab9, 0x363e, 0x389f, 0x31bb, 0x38e1, 0x3bc4, 0x3b9b, 0x2984, 0x3a43, 0x3b2f, 0x35d6, 0x3bda, 
+0x2df3, 0x3bf8, 0x2acc, 0x378b, 0x3555, 0x2e59, 0x31d4, 0x34ec, 0x3a46, 0x3bab, 0x3214, 0x3161, 0x3470, 0x3a03, 0x368e, 0x31ad, 0x27cb, 0x2ecb, 0x3422, 0x39f7, 0x3644, 0x3a77, 0x313f, 0x34f2, 0x39b3, 0x3bf2, 0x379a, 0x3456, 0x35fe, 0x3ae7, 0x3964, 0x385f, 
+0x3b16, 0x3999, 0x3833, 0x2eda, 0x3afd, 0x3a4a, 0x3ba2, 0x2bd4, 0x3b38, 0x31a2, 0x32dd, 0x353c, 0x366f, 0x375e, 0x3821, 0x367a, 0x3b44, 0x39e6, 0x3787, 0x339e, 0x39d7, 0x38c6, 0x37d5, 0x342f, 0x3984, 0x319b, 0x33b5, 0x35ab, 0x398a, 0x374e, 0x36b6, 0x3b21, 
+0x3bbb, 0x2ab3, 0x2ad5, 0x33bc, 0x2bef, 0x3780, 0x3738, 0x3a0b, 0x3b09, 0x30ca, 0x384e, 0x3ab3, 0x39bd, 0x3453, 0x3a6d, 0x3957, 0x2c10, 0x30e9, 0x35d4, 0x3aef, 0x3be9, 0x39ad, 0x3a74, 0x3af9, 0x3739, 0x2d4d, 0x39fe, 0x3b72, 0x2c57, 0x398c, 0x381f, 0x3930, 
+0x3820, 0x321b, 0x3964, 0x2964, 0x33a0, 0x2d00, 0x2490, 0x336b, 0x3465, 0x3b2e, 0x3aa0, 0x371f, 0x300e, 0x3a09, 0x3bf1, 0x25cc, 0x3b6f, 0x3384, 0x3a88, 0x3acb, 0x3814, 0x36d0, 0x3081, 0x3a2c, 0x3353, 0x39cb, 0x31ed, 0x3af6, 0x3721, 0x36c7, 0x2ce2, 0x390d, 
+0x3698, 0x3ab2, 0x3b3e, 0x2eb4, 0x3998, 0x39e3, 0x3a77, 0x3632, 0x2c12, 0x3bd5, 0x3ba3, 0x3bba, 0x323c, 0x367b, 0x3557, 0x39c8, 0x37db, 0x3b45, 0x3b6e, 0x3931, 0x3121, 0x3a8d, 0x3a55, 0x3b9b, 0x358a, 0x3925, 0x3491, 0x3912, 0x3b6b, 0x3584, 0x32df, 0x3120, 
+0x32b2, 0x3b0a, 0x2cad, 0x3465, 0x3ad3, 0x3bcd, 0x363b, 0x3afe, 0x354b, 0x3374, 0x39af, 0x3b7f, 0x308c, 0x2e72, 0x3380, 0x3b70, 0x3902, 0x38d8, 0x39f3, 0x3a4b, 0x3853, 0x397b, 0x2ebe, 0x387f, 0x2845, 0x37e2, 0x360f, 0x370b, 0x3acb, 0x35d4, 0x36e6, 0x3262, 
+0x2e88, 0x3a54, 0x2ee3, 0x3575, 0x3afe, 0x2aee, 0x39a0, 0x3aae, 0x3693, 0x3432, 0x3834, 0x3b9b, 0x3bcb, 0x2e3a, 0x356d, 0x374e, 0x3924, 0x383c, 0x311e, 0x3ac5, 0x352d, 0x311e, 0x38ca, 0x34d4, 0x36ca, 0x34ed, 0x3a13, 0x33eb, 0x3639, 0x3828, 0x3b3c, 0x3939, 
+0x3837, 0x3521, 0x2cb5, 0x3629, 0x3924, 0x384c, 0x366a, 0x3bbf, 0x2e9e, 0x3ba8, 0x33ad, 0x38c8, 0x3934, 0x3907, 0x249a, 0x3690, 0x3a09, 0x3215, 0x3898, 0x325d, 0x37d5, 0x3195, 0x361c, 0x3ae4, 0x351f, 0x3452, 0x3bc0, 0x375c, 0x39bf, 0x317a, 0x3aae, 0x283a, 
+0x3476, 0x3b92, 0x3472, 0x383e, 0x280f, 0x39d6, 0x2fd1, 0x31f4, 0x2ffb, 0x3b97, 0x3692, 0x36c0, 0x3989, 0x33cf, 0x3ba6, 0x3239, 0x35d7, 0x33ab, 0x31eb, 0x3b47, 0x389b, 0x3b88, 0x3580, 0x354c, 0x3802, 0x3b9a, 0x3b94, 0x2a92, 0x2db1, 0x38bd, 0x2dfb, 0x3900, 
+0x344f, 0x3739, 0x27a5, 0x3b2e, 0x342b, 0x34bb, 0x30c8, 0x3ae8, 0x3b26, 0x3982, 0x38c0, 0x3408, 0x38c8, 0x36ef, 0x3bf0, 0x3acf, 0x3a3c, 0x3825, 0x31a5, 0x3ada, 0x3b5b, 0x37db, 0x3a01, 0x3663, 0x3a7d, 0x327b, 0x3a1f, 0x3862, 0x38af, 0x3204, 0x372e, 0x3b19, 
+0x3708, 0x3622, 0x2e62, 0x39ab, 0x2d4d, 0x31b4, 0x3552, 0x3bbc, 0x36f2, 0x36eb, 0x38ef, 0x3755, 0x3bbe, 0x2c17, 0x3815, 0x2f53, 0x363f, 0x38c1, 0x3246, 0x386b, 0x34de, 0x34e4, 0x3baa, 0x349e, 0x32ce, 0x3a68, 0x373f, 0x2cce, 0x3b36, 0x28ba, 0x3b50, 0x3232, 
+0x1f34, 0x3928, 0x35cd, 0x3b38, 0x30ce, 0x35a1, 0x3a06, 0x3a32, 0x3a53, 0x3489, 0x3241, 0x372f, 0x390c, 0x3a1b, 0x378a, 0x3713, 0x3769, 0x37a8, 0x3418, 0x3ad4, 0x3a4e, 0x3bf7, 0x37a5, 0x34dc, 0x39b2, 0x351b, 0x3372, 0x349f, 0x2f50, 0x3ab1, 0x3795, 0x2db7, 
+0x3864, 0x3157, 0x3900, 0x323e, 0x389e, 0x3880, 0x3b1f, 0x37a1, 0x396c, 0x2e43, 0x2c2a, 0x3b78, 0x3988, 0x3a14, 0x39c1, 0x3b51, 0x3780, 0x3bf2, 0x2d19, 0x3815, 0x3a5f, 0x3641, 0x2f62, 0x37d5, 0x3564, 0x139a, 0x3ab8, 0x28f7, 0x3785, 0x34e1, 0x3097, 0x3768, 
+0x3971, 0x3ae2, 0x32ae, 0x2fd5, 0x382a, 0x346c, 0x3133, 0x3167, 0x3940, 0x2d12, 0x389a, 0x3bd0, 0x3943, 0x391c, 0x3a75, 0x2a11, 0x391e, 0x372d, 0x3a79, 0x3b72, 0x3373, 0x39b7, 0x35d7, 0x372b, 0x3a6d, 0x38a1, 0x3279, 0x3434, 0x3694, 0x3b45, 0x3abb, 0x392d, 
+0x34a8, 0x3757, 0x32ca, 0x345d, 0x36a5, 0x3854, 0x2dcd, 0x30af, 0x38dd, 0x3067, 0x3411, 0x3997, 0x397a, 0x3a64, 0x38b8, 0x3962, 0x3509, 0x3bb6, 0x3a66, 0x339f, 0x372a, 0x31a8, 0x37da, 0x36ff, 0x33c6, 0x31da, 0x3977, 0x3b72, 0x3841, 0x3567, 0x3433, 0x33b8, 
+0x39fe, 0x3a10, 0x3bf2, 0x35e7, 0x3a4a, 0x3b3e, 0x2ec7, 0x3aa4, 0x3846, 0x3af9, 0x38a9, 0x2c1f, 0x39ab, 0x349f, 0x31d6, 0x39ae, 0x3b79, 0x352d, 0x3516, 0x347c, 0x2f33, 0x35ad, 0x31c4, 0x3b52, 0x354b, 0x3786, 0x3ab7, 0x3896, 0x34ac, 0x352f, 0x37e6, 0x326a, 
+0x2e44, 0x34c7, 0x388d, 0x3bf4, 0x363f, 0x3b3d, 0x33b1, 0x3b8b, 0x3340, 0x37f7, 0x3b07, 0x25bf, 0x398e, 0x3505, 0x3bd7, 0x366d, 0x388a, 0x2cc0, 0x359a, 0x3b9a, 0x3b99, 0x379d, 0x3b6b, 0x39b8, 0x3223, 0x2703, 0x3ba9, 0x2ecb, 0x3759, 0x39d8, 0x37ac, 0x32cf, 
+0x35f2, 0x38a3, 0x399e, 0x3bd2, 0x3780, 0x3af3, 0x3b5e, 0x337b, 0x3a08, 0x35da, 0x3446, 0x3b25, 0x3ad0, 0x3bee, 0x3141, 0x32d8, 0x34ce, 0x2ac9, 0x3800, 0x3a8a, 0x2d53, 0x368a, 0x3561, 0x3998, 0x35a3, 0x3677, 0x3ab2, 0x3269, 0x3236, 0x3b3e, 0x3aba, 0x3bac, 
+0x395d, 0x3820, 0x1df6, 0x3bb5, 0x35b5, 0x3675, 0x3b74, 0x360f, 0x34de, 0x3a0c, 0x3aeb, 0x299d, 0x3207, 0x3bd8, 0x2178, 0x3995, 0x3948, 0x3908, 0x3843, 0x2ea5, 0x3045, 0x3989, 0x345d, 0x39c5, 0x3a89, 0x3863, 0x3be0, 0x397a, 0x38f1, 0x39e2, 0x3b08, 0x352e, 
+0x385f, 0x28f2, 0x3bc3, 0x35e0, 0x380c, 0x3b9c, 0x3afc, 0x390a, 0x3689, 0x34fd, 0x2cf5, 0x308e, 0x342b, 0x3921, 0x3a67, 0x3ad6, 0x2986, 0x32fc, 0x35aa, 0x3507, 0x3608, 0x33fd, 0x3bf3, 0x39e2, 0x3b0f, 0x30b7, 0x3896, 0x3ae4, 0x2145, 0x35b6, 0x2e1d, 0x3ad1, 
+0x333d, 0x3afb, 0x2703, 0x3413, 0x1d7d, 0x3b7f, 0x3ae1, 0x303c, 0x3004, 0x39d3, 0x3554, 0x31a4, 0x354e, 0x3662, 0x39c5, 0x2eb7, 0x2c6e, 0x397f, 0x31d8, 0x1f0c, 0x38e3, 0x35f0, 0x2714, 0x28d1, 0x375e, 0x3a75, 0x3830, 0x3578, 0x397d, 0x3b18, 0x383c, 0x3498, 
+0x39ad, 0x3598, 0x23c4, 0x34ea, 0x3a61, 0x2b00, 0x3707, 0x3ae1, 0x37ae, 0x389d, 0x37fa, 0x3673, 0x3278, 0xf3e, 0x3809, 0x33c6, 0x3bf5, 0x3279, 0x3816, 0x360c, 0x39c8, 0x381f, 0x3741, 0x2d66, 0x38c0, 0x37d3, 0x377a, 0x3621, 0x2faf, 0x392e, 0x2de6, 0x33c5, 
+0x3803, 0x2600, 0x32e9, 0x39b4, 0x38d2, 0x34e8, 0x2fe6, 0x3199, 0x3643, 0x3a77, 0x27cc, 0x39d7, 0x34c6, 0x2ea8, 0x364e, 0x3b07, 0x31c7, 0x30a1, 0x31b1, 0x3b8f, 0x3571, 0x3b75, 0x3989, 0x3805, 0x39fb, 0x3945, 0x352b, 0x31d8, 0x3904, 0x3440, 0x3a57, 0x2cf7, 
+0x3b39, 0x2fcd, 0x2b89, 0x2edd, 0x3682, 0x36a9, 0x32c8, 0x37ac, 0x32a5, 0x3311, 0x394b, 0x3b84, 0x3aec, 0x3601, 0x2765, 0x3b69, 0x396b, 0x3727, 0x3bfe, 0x3907, 0x376f, 0x3674, 0x3973, 0x3671, 0x3491, 0x3993, 0x383f, 0x3335, 0x3989, 0x3550, 0x3077, 0x35f5, 
+0x3a59, 0x3950, 0x380c, 0x37cd, 0x30bf, 0x3607, 0x3afa, 0x3b5d, 0x32b9, 0x386b, 0x35bd, 0x3aca, 0x3ba5, 0x3b2d, 0x3b19, 0x3b8b, 0x345e, 0x2845, 0x34aa, 0x372a, 0x3448, 0x34f5, 0x3ae2, 0x3637, 0x2cb5, 0x354b, 0x3b15, 0x2ca8, 0x2641, 0x3178, 0x2cfe, 0x39b4, 
+0x3bdd, 0x3acb, 0x3a05, 0x38a2, 0x3b4a, 0x34e5, 0x395f, 0x394b, 0x34c4, 0x3aa5, 0x29bb, 0x2d96, 0x339d, 0x387c, 0x382e, 0x385a, 0x396b, 0x3aa9, 0x2f1e, 0x33a7, 0x3b90, 0x3b7b, 0x3b5f, 0x39d3, 0x3b18, 0x354f, 0x2cdb, 0x3a6f, 0x3434, 0x34ff, 0x3a5b, 0x3b84, 
+0x3a33, 0x384b, 0x2e67, 0x3b85, 0x3853, 0x380c, 0x346a, 0x3aaa, 0x3492, 0x33e8, 0x3bf2, 0x38ae, 0x3a29, 0x3830, 0x3221, 0x35b1, 0x3a48, 0x2c68, 0x2ced, 0x3a7e, 0x3539, 0x3922, 0x374c, 0x3aaa, 0x2dae, 0x395d, 0x3b3d, 0x3890, 0x2cfe, 0x2dd6, 0x3bad, 0x33c5, 
+0x2c07, 0x3a2c, 0x37a8, 0x390f, 0x2fc8, 0x35ae, 0x388c, 0x30ee, 0x3674, 0x391d, 0x3bfc, 0x36bf, 0x322d, 0x3a78, 0x35c0, 0x3492, 0x3ac8, 0x3504, 0x3315, 0x381d, 0x3a7a, 0x3a08, 0x343c, 0x3bda, 0x341b, 0x39f0, 0x3b9e, 0x395d, 0x3c00, 0x38ab, 0x3bcf, 0x3564, 
+0x33c4, 0x3b0d, 0x3623, 0x33b9, 0x3b92, 0x1e71, 0x2c57, 0x36d0, 0x314b, 0x3a16, 0x3372, 0x341b, 0x3aaa, 0x3444, 0x396b, 0x2dd7, 0x3b30, 0x3559, 0x3b5b, 0x3a29, 0x2d19, 0x38b7, 0x3b01, 0x3afa, 0x398a, 0x3839, 0x3ac9, 0x2e31, 0x3924, 0x39f2, 0x3a7f, 0x3285 
+};
\ No newline at end of file
diff --git a/hwpe/redmule_softclear/inc/w_input.h b/hwpe/redmule_softclear/inc/w_input.h
new file mode 100644
index 0000000..dc4d3be
--- /dev/null
+++ b/hwpe/redmule_softclear/inc/w_input.h
@@ -0,0 +1,35 @@
+ /* Header file generated by RedMulE Golden Model */
+uint16_t w_inp [1024] = {
+0x311a, 0x39e0, 0x387d, 0x3a4a, 0x386f, 0x3ada, 0x392f, 0x3854, 0x3014, 0x2fd2, 0x31c9, 0x2fca, 0x2e55, 0x3bc8, 0x396d, 0x3b1d, 0x39f6, 0x333a, 0x3908, 0x3628, 0x3bab, 0x3b8b, 0x3b4a, 0x322d, 0x3925, 0x317a, 0x3725, 0x31c2, 0x3066, 0x38f3, 0x3a17, 0x3476, 
+0x3bda, 0x3196, 0x3922, 0x3680, 0x396a, 0x3021, 0x3761, 0x374d, 0x2fc2, 0x3967, 0x3b94, 0x33b5, 0x3797, 0x34d6, 0x3655, 0x2176, 0x39bc, 0x3999, 0x3658, 0x3904, 0x3759, 0x2ade, 0x3a5a, 0x3b78, 0x36c7, 0x2d01, 0x3b58, 0x2d9a, 0x373d, 0x3952, 0x38e8, 0x3887, 
+0x37b6, 0x3a88, 0x2f8a, 0x2d79, 0x3413, 0x3421, 0x3976, 0x32b2, 0x3446, 0x2d99, 0x3a56, 0x3322, 0x3b49, 0x39fa, 0x3acd, 0x3af6, 0x304c, 0x3abb, 0x3a83, 0x38b2, 0x3ab9, 0x363e, 0x389f, 0x31bb, 0x38e1, 0x3bc4, 0x3b9b, 0x2984, 0x3a43, 0x3b2f, 0x35d6, 0x3bda, 
+0x2df3, 0x3bf8, 0x2acc, 0x378b, 0x3555, 0x2e59, 0x31d4, 0x34ec, 0x3a46, 0x3bab, 0x3214, 0x3161, 0x3470, 0x3a03, 0x368e, 0x31ad, 0x27cb, 0x2ecb, 0x3422, 0x39f7, 0x3644, 0x3a77, 0x313f, 0x34f2, 0x39b3, 0x3bf2, 0x379a, 0x3456, 0x35fe, 0x3ae7, 0x3964, 0x385f, 
+0x3b16, 0x3999, 0x3833, 0x2eda, 0x3afd, 0x3a4a, 0x3ba2, 0x2bd4, 0x3b38, 0x31a2, 0x32dd, 0x353c, 0x366f, 0x375e, 0x3821, 0x367a, 0x3b44, 0x39e6, 0x3787, 0x339e, 0x39d7, 0x38c6, 0x37d5, 0x342f, 0x3984, 0x319b, 0x33b5, 0x35ab, 0x398a, 0x374e, 0x36b6, 0x3b21, 
+0x3bbb, 0x2ab3, 0x2ad5, 0x33bc, 0x2bef, 0x3780, 0x3738, 0x3a0b, 0x3b09, 0x30ca, 0x384e, 0x3ab3, 0x39bd, 0x3453, 0x3a6d, 0x3957, 0x2c10, 0x30e9, 0x35d4, 0x3aef, 0x3be9, 0x39ad, 0x3a74, 0x3af9, 0x3739, 0x2d4d, 0x39fe, 0x3b72, 0x2c57, 0x398c, 0x381f, 0x3930, 
+0x3820, 0x321b, 0x3964, 0x2964, 0x33a0, 0x2d00, 0x2490, 0x336b, 0x3465, 0x3b2e, 0x3aa0, 0x371f, 0x300e, 0x3a09, 0x3bf1, 0x25cc, 0x3b6f, 0x3384, 0x3a88, 0x3acb, 0x3814, 0x36d0, 0x3081, 0x3a2c, 0x3353, 0x39cb, 0x31ed, 0x3af6, 0x3721, 0x36c7, 0x2ce2, 0x390d, 
+0x3698, 0x3ab2, 0x3b3e, 0x2eb4, 0x3998, 0x39e3, 0x3a77, 0x3632, 0x2c12, 0x3bd5, 0x3ba3, 0x3bba, 0x323c, 0x367b, 0x3557, 0x39c8, 0x37db, 0x3b45, 0x3b6e, 0x3931, 0x3121, 0x3a8d, 0x3a55, 0x3b9b, 0x358a, 0x3925, 0x3491, 0x3912, 0x3b6b, 0x3584, 0x32df, 0x3120, 
+0x32b2, 0x3b0a, 0x2cad, 0x3465, 0x3ad3, 0x3bcd, 0x363b, 0x3afe, 0x354b, 0x3374, 0x39af, 0x3b7f, 0x308c, 0x2e72, 0x3380, 0x3b70, 0x3902, 0x38d8, 0x39f3, 0x3a4b, 0x3853, 0x397b, 0x2ebe, 0x387f, 0x2845, 0x37e2, 0x360f, 0x370b, 0x3acb, 0x35d4, 0x36e6, 0x3262, 
+0x2e88, 0x3a54, 0x2ee3, 0x3575, 0x3afe, 0x2aee, 0x39a0, 0x3aae, 0x3693, 0x3432, 0x3834, 0x3b9b, 0x3bcb, 0x2e3a, 0x356d, 0x374e, 0x3924, 0x383c, 0x311e, 0x3ac5, 0x352d, 0x311e, 0x38ca, 0x34d4, 0x36ca, 0x34ed, 0x3a13, 0x33eb, 0x3639, 0x3828, 0x3b3c, 0x3939, 
+0x3837, 0x3521, 0x2cb5, 0x3629, 0x3924, 0x384c, 0x366a, 0x3bbf, 0x2e9e, 0x3ba8, 0x33ad, 0x38c8, 0x3934, 0x3907, 0x249a, 0x3690, 0x3a09, 0x3215, 0x3898, 0x325d, 0x37d5, 0x3195, 0x361c, 0x3ae4, 0x351f, 0x3452, 0x3bc0, 0x375c, 0x39bf, 0x317a, 0x3aae, 0x283a, 
+0x3476, 0x3b92, 0x3472, 0x383e, 0x280f, 0x39d6, 0x2fd1, 0x31f4, 0x2ffb, 0x3b97, 0x3692, 0x36c0, 0x3989, 0x33cf, 0x3ba6, 0x3239, 0x35d7, 0x33ab, 0x31eb, 0x3b47, 0x389b, 0x3b88, 0x3580, 0x354c, 0x3802, 0x3b9a, 0x3b94, 0x2a92, 0x2db1, 0x38bd, 0x2dfb, 0x3900, 
+0x344f, 0x3739, 0x27a5, 0x3b2e, 0x342b, 0x34bb, 0x30c8, 0x3ae8, 0x3b26, 0x3982, 0x38c0, 0x3408, 0x38c8, 0x36ef, 0x3bf0, 0x3acf, 0x3a3c, 0x3825, 0x31a5, 0x3ada, 0x3b5b, 0x37db, 0x3a01, 0x3663, 0x3a7d, 0x327b, 0x3a1f, 0x3862, 0x38af, 0x3204, 0x372e, 0x3b19, 
+0x3708, 0x3622, 0x2e62, 0x39ab, 0x2d4d, 0x31b4, 0x3552, 0x3bbc, 0x36f2, 0x36eb, 0x38ef, 0x3755, 0x3bbe, 0x2c17, 0x3815, 0x2f53, 0x363f, 0x38c1, 0x3246, 0x386b, 0x34de, 0x34e4, 0x3baa, 0x349e, 0x32ce, 0x3a68, 0x373f, 0x2cce, 0x3b36, 0x28ba, 0x3b50, 0x3232, 
+0x1f34, 0x3928, 0x35cd, 0x3b38, 0x30ce, 0x35a1, 0x3a06, 0x3a32, 0x3a53, 0x3489, 0x3241, 0x372f, 0x390c, 0x3a1b, 0x378a, 0x3713, 0x3769, 0x37a8, 0x3418, 0x3ad4, 0x3a4e, 0x3bf7, 0x37a5, 0x34dc, 0x39b2, 0x351b, 0x3372, 0x349f, 0x2f50, 0x3ab1, 0x3795, 0x2db7, 
+0x3864, 0x3157, 0x3900, 0x323e, 0x389e, 0x3880, 0x3b1f, 0x37a1, 0x396c, 0x2e43, 0x2c2a, 0x3b78, 0x3988, 0x3a14, 0x39c1, 0x3b51, 0x3780, 0x3bf2, 0x2d19, 0x3815, 0x3a5f, 0x3641, 0x2f62, 0x37d5, 0x3564, 0x139a, 0x3ab8, 0x28f7, 0x3785, 0x34e1, 0x3097, 0x3768, 
+0x3971, 0x3ae2, 0x32ae, 0x2fd5, 0x382a, 0x346c, 0x3133, 0x3167, 0x3940, 0x2d12, 0x389a, 0x3bd0, 0x3943, 0x391c, 0x3a75, 0x2a11, 0x391e, 0x372d, 0x3a79, 0x3b72, 0x3373, 0x39b7, 0x35d7, 0x372b, 0x3a6d, 0x38a1, 0x3279, 0x3434, 0x3694, 0x3b45, 0x3abb, 0x392d, 
+0x34a8, 0x3757, 0x32ca, 0x345d, 0x36a5, 0x3854, 0x2dcd, 0x30af, 0x38dd, 0x3067, 0x3411, 0x3997, 0x397a, 0x3a64, 0x38b8, 0x3962, 0x3509, 0x3bb6, 0x3a66, 0x339f, 0x372a, 0x31a8, 0x37da, 0x36ff, 0x33c6, 0x31da, 0x3977, 0x3b72, 0x3841, 0x3567, 0x3433, 0x33b8, 
+0x39fe, 0x3a10, 0x3bf2, 0x35e7, 0x3a4a, 0x3b3e, 0x2ec7, 0x3aa4, 0x3846, 0x3af9, 0x38a9, 0x2c1f, 0x39ab, 0x349f, 0x31d6, 0x39ae, 0x3b79, 0x352d, 0x3516, 0x347c, 0x2f33, 0x35ad, 0x31c4, 0x3b52, 0x354b, 0x3786, 0x3ab7, 0x3896, 0x34ac, 0x352f, 0x37e6, 0x326a, 
+0x2e44, 0x34c7, 0x388d, 0x3bf4, 0x363f, 0x3b3d, 0x33b1, 0x3b8b, 0x3340, 0x37f7, 0x3b07, 0x25bf, 0x398e, 0x3505, 0x3bd7, 0x366d, 0x388a, 0x2cc0, 0x359a, 0x3b9a, 0x3b99, 0x379d, 0x3b6b, 0x39b8, 0x3223, 0x2703, 0x3ba9, 0x2ecb, 0x3759, 0x39d8, 0x37ac, 0x32cf, 
+0x35f2, 0x38a3, 0x399e, 0x3bd2, 0x3780, 0x3af3, 0x3b5e, 0x337b, 0x3a08, 0x35da, 0x3446, 0x3b25, 0x3ad0, 0x3bee, 0x3141, 0x32d8, 0x34ce, 0x2ac9, 0x3800, 0x3a8a, 0x2d53, 0x368a, 0x3561, 0x3998, 0x35a3, 0x3677, 0x3ab2, 0x3269, 0x3236, 0x3b3e, 0x3aba, 0x3bac, 
+0x395d, 0x3820, 0x1df6, 0x3bb5, 0x35b5, 0x3675, 0x3b74, 0x360f, 0x34de, 0x3a0c, 0x3aeb, 0x299d, 0x3207, 0x3bd8, 0x2178, 0x3995, 0x3948, 0x3908, 0x3843, 0x2ea5, 0x3045, 0x3989, 0x345d, 0x39c5, 0x3a89, 0x3863, 0x3be0, 0x397a, 0x38f1, 0x39e2, 0x3b08, 0x352e, 
+0x385f, 0x28f2, 0x3bc3, 0x35e0, 0x380c, 0x3b9c, 0x3afc, 0x390a, 0x3689, 0x34fd, 0x2cf5, 0x308e, 0x342b, 0x3921, 0x3a67, 0x3ad6, 0x2986, 0x32fc, 0x35aa, 0x3507, 0x3608, 0x33fd, 0x3bf3, 0x39e2, 0x3b0f, 0x30b7, 0x3896, 0x3ae4, 0x2145, 0x35b6, 0x2e1d, 0x3ad1, 
+0x333d, 0x3afb, 0x2703, 0x3413, 0x1d7d, 0x3b7f, 0x3ae1, 0x303c, 0x3004, 0x39d3, 0x3554, 0x31a4, 0x354e, 0x3662, 0x39c5, 0x2eb7, 0x2c6e, 0x397f, 0x31d8, 0x1f0c, 0x38e3, 0x35f0, 0x2714, 0x28d1, 0x375e, 0x3a75, 0x3830, 0x3578, 0x397d, 0x3b18, 0x383c, 0x3498, 
+0x39ad, 0x3598, 0x23c4, 0x34ea, 0x3a61, 0x2b00, 0x3707, 0x3ae1, 0x37ae, 0x389d, 0x37fa, 0x3673, 0x3278, 0xf3e, 0x3809, 0x33c6, 0x3bf5, 0x3279, 0x3816, 0x360c, 0x39c8, 0x381f, 0x3741, 0x2d66, 0x38c0, 0x37d3, 0x377a, 0x3621, 0x2faf, 0x392e, 0x2de6, 0x33c5, 
+0x3803, 0x2600, 0x32e9, 0x39b4, 0x38d2, 0x34e8, 0x2fe6, 0x3199, 0x3643, 0x3a77, 0x27cc, 0x39d7, 0x34c6, 0x2ea8, 0x364e, 0x3b07, 0x31c7, 0x30a1, 0x31b1, 0x3b8f, 0x3571, 0x3b75, 0x3989, 0x3805, 0x39fb, 0x3945, 0x352b, 0x31d8, 0x3904, 0x3440, 0x3a57, 0x2cf7, 
+0x3b39, 0x2fcd, 0x2b89, 0x2edd, 0x3682, 0x36a9, 0x32c8, 0x37ac, 0x32a5, 0x3311, 0x394b, 0x3b84, 0x3aec, 0x3601, 0x2765, 0x3b69, 0x396b, 0x3727, 0x3bfe, 0x3907, 0x376f, 0x3674, 0x3973, 0x3671, 0x3491, 0x3993, 0x383f, 0x3335, 0x3989, 0x3550, 0x3077, 0x35f5, 
+0x3a59, 0x3950, 0x380c, 0x37cd, 0x30bf, 0x3607, 0x3afa, 0x3b5d, 0x32b9, 0x386b, 0x35bd, 0x3aca, 0x3ba5, 0x3b2d, 0x3b19, 0x3b8b, 0x345e, 0x2845, 0x34aa, 0x372a, 0x3448, 0x34f5, 0x3ae2, 0x3637, 0x2cb5, 0x354b, 0x3b15, 0x2ca8, 0x2641, 0x3178, 0x2cfe, 0x39b4, 
+0x3bdd, 0x3acb, 0x3a05, 0x38a2, 0x3b4a, 0x34e5, 0x395f, 0x394b, 0x34c4, 0x3aa5, 0x29bb, 0x2d96, 0x339d, 0x387c, 0x382e, 0x385a, 0x396b, 0x3aa9, 0x2f1e, 0x33a7, 0x3b90, 0x3b7b, 0x3b5f, 0x39d3, 0x3b18, 0x354f, 0x2cdb, 0x3a6f, 0x3434, 0x34ff, 0x3a5b, 0x3b84, 
+0x3a33, 0x384b, 0x2e67, 0x3b85, 0x3853, 0x380c, 0x346a, 0x3aaa, 0x3492, 0x33e8, 0x3bf2, 0x38ae, 0x3a29, 0x3830, 0x3221, 0x35b1, 0x3a48, 0x2c68, 0x2ced, 0x3a7e, 0x3539, 0x3922, 0x374c, 0x3aaa, 0x2dae, 0x395d, 0x3b3d, 0x3890, 0x2cfe, 0x2dd6, 0x3bad, 0x33c5, 
+0x2c07, 0x3a2c, 0x37a8, 0x390f, 0x2fc8, 0x35ae, 0x388c, 0x30ee, 0x3674, 0x391d, 0x3bfc, 0x36bf, 0x322d, 0x3a78, 0x35c0, 0x3492, 0x3ac8, 0x3504, 0x3315, 0x381d, 0x3a7a, 0x3a08, 0x343c, 0x3bda, 0x341b, 0x39f0, 0x3b9e, 0x395d, 0x3c00, 0x38ab, 0x3bcf, 0x3564, 
+0x33c4, 0x3b0d, 0x3623, 0x33b9, 0x3b92, 0x1e71, 0x2c57, 0x36d0, 0x314b, 0x3a16, 0x3372, 0x341b, 0x3aaa, 0x3444, 0x396b, 0x2dd7, 0x3b30, 0x3559, 0x3b5b, 0x3a29, 0x2d19, 0x38b7, 0x3b01, 0x3afa, 0x398a, 0x3839, 0x3ac9, 0x2e31, 0x3924, 0x39f2, 0x3a7f, 0x3285 
+};
\ No newline at end of file
diff --git a/hwpe/redmule_softclear/inc/x_2D.h b/hwpe/redmule_softclear/inc/x_2D.h
new file mode 100644
index 0000000..0b589f8
--- /dev/null
+++ b/hwpe/redmule_softclear/inc/x_2D.h
@@ -0,0 +1,27 @@
+ /* Header file generated by RedMulE Golden Model */
+uint16_t x_inp_2D [24][32] = {
+0x2153, 0x3bb5, 0x3896, 0x365f, 0x2483, 0x3518, 0x2dd1, 0x3bca, 0x397b, 0x29b1, 0x3705, 0x36c8, 0x398b, 0x3661, 0x2f05, 0x365a, 0x3bf9, 0x34df, 0x363b, 0x38d9, 0x39c6, 0x3abb, 0x3952, 0x38f2, 0x392d, 0x3b3e, 0x2afb, 0x3a9d, 0x353b, 0x3b73, 0x3a01, 0x3679, 
+0x3934, 0x397d, 0x2904, 0x3822, 0x3462, 0x3b44, 0x39e9, 0x28be, 0x331e, 0x3a1d, 0x39e5, 0x34da, 0x3a19, 0x3906, 0x1d35, 0x3871, 0x31e7, 0x3b29, 0x325d, 0x3797, 0x2b2f, 0x38b4, 0x232f, 0x38aa, 0x3aca, 0x316f, 0x3811, 0x3950, 0x32ea, 0x3bc7, 0x382c, 0x38a2, 
+0x29ce, 0x3afa, 0x3a39, 0x2ccc, 0x39fd, 0x3b3d, 0x384a, 0x3a35, 0x3802, 0x366a, 0x37ec, 0x3598, 0x3bf8, 0x3a85, 0x3a1b, 0x386e, 0x3b4c, 0x39de, 0x38c2, 0x2f93, 0x3b4c, 0x39c4, 0x3b9e, 0x3844, 0x346d, 0x3bff, 0x32ce, 0x296d, 0x3130, 0x3b3d, 0x3b44, 0x369d, 
+0x3b13, 0x31ed, 0x330a, 0x3831, 0x34e7, 0x37b3, 0x331a, 0x3918, 0x32d3, 0x3995, 0x3991, 0x3919, 0x3a26, 0x385b, 0x2b76, 0x3a3b, 0x37f2, 0x26a7, 0x3225, 0x3b64, 0x28f0, 0x3456, 0x3822, 0x341e, 0x381a, 0x38d8, 0x2c11, 0x33be, 0x33ac, 0x353f, 0x3476, 0x3abc, 
+0x36ec, 0x3a1d, 0x39d3, 0x3821, 0x36ac, 0x3bce, 0x3ad2, 0x3616, 0x36a1, 0x2cb3, 0x38d2, 0x314f, 0x385c, 0x3b63, 0x3bb6, 0x2951, 0x372d, 0x2c42, 0x3823, 0x3883, 0x3872, 0x31ee, 0x36c5, 0x399a, 0x31b0, 0x3887, 0x3884, 0x3865, 0x3896, 0x36c3, 0x32e3, 0x346c, 
+0x3935, 0x3b50, 0x2b6d, 0x38cd, 0x388f, 0x3389, 0x395d, 0x31cd, 0x2efd, 0x3154, 0x2f35, 0x3444, 0x3293, 0x3b6b, 0x1bec, 0x3b69, 0x3bf3, 0x3611, 0x3508, 0x3742, 0x3a50, 0x3ab7, 0x3457, 0x38d3, 0x3344, 0x38e8, 0x33c0, 0x3668, 0x3bee, 0x3b21, 0x3727, 0x3121, 
+0x316c, 0x3288, 0x2d50, 0x2e74, 0x35d5, 0x37e2, 0x303d, 0x36af, 0x341f, 0x3436, 0x2df7, 0x399d, 0x30f4, 0x3aaf, 0x34e4, 0x2c2a, 0x3116, 0x34d3, 0x36ac, 0x35e3, 0x3760, 0x36e1, 0x3ad2, 0x3547, 0x38f4, 0x369c, 0x3ba9, 0x34f0, 0x3a39, 0x3b19, 0x36e6, 0x395d, 
+0x3be8, 0x3293, 0x3bfc, 0x3435, 0x2eb3, 0x3360, 0x3919, 0x3bed, 0x396a, 0x37fc, 0x3242, 0x384b, 0x38cb, 0x3b2c, 0x3b28, 0x28cf, 0x3828, 0x3855, 0x3ba9, 0x2fa7, 0x340b, 0x32f1, 0x3ada, 0x36fa, 0x31f5, 0x3436, 0x29d0, 0x33e6, 0x3232, 0x3bec, 0x3904, 0x2797, 
+0x3b81, 0x3bac, 0x38d2, 0x343d, 0x31af, 0x3b1e, 0x33fc, 0x3864, 0x3624, 0x3905, 0x2945, 0x3b52, 0x2d08, 0x3a17, 0x3b84, 0x3804, 0x3a24, 0x38a3, 0x3562, 0x3ae6, 0x3bba, 0x3a45, 0x3679, 0x31fa, 0x3994, 0x2c3d, 0x383f, 0x399d, 0x34f7, 0x360e, 0x35f3, 0x38f0, 
+0x38d4, 0x399a, 0x3a48, 0x3987, 0x3b54, 0x382c, 0x3210, 0x35ef, 0x36ca, 0x31b4, 0x3625, 0x371f, 0x37bd, 0x3680, 0x3a3a, 0x3ac0, 0x3bbf, 0x3bf5, 0x39f2, 0x29c2, 0x363e, 0x3a4e, 0x3596, 0x3b1b, 0x3459, 0x3669, 0x3aa1, 0x39c3, 0x3376, 0x390d, 0x2456, 0x39b5, 
+0x3a66, 0x3ad8, 0x3b51, 0x36aa, 0x32be, 0x3ac8, 0x392b, 0x3740, 0x3a48, 0x38f5, 0x3b2d, 0x3a5f, 0x2ff3, 0x366f, 0x39d3, 0x35e5, 0x3822, 0x38db, 0x3b8a, 0x34be, 0x2d33, 0x36dd, 0x3578, 0x3bdf, 0x2c7e, 0x39cf, 0x32ff, 0x35c9, 0x3970, 0x3bcb, 0x351e, 0x3956, 
+0x2c42, 0x3308, 0x377a, 0x361c, 0x39a0, 0x36c9, 0x2dcb, 0x3bf2, 0x3b5f, 0x33ee, 0x24c1, 0x2ce9, 0x3927, 0x305d, 0x3702, 0x3119, 0x35f9, 0x3855, 0x3374, 0x349b, 0x3bcf, 0x2dea, 0x34f0, 0x363f, 0x37da, 0x3a74, 0x35fc, 0x35fa, 0x316b, 0x3804, 0x37a7, 0x3986, 
+0x3073, 0x3aed, 0x31c7, 0x3844, 0x34a4, 0x387d, 0x3a20, 0x3037, 0x3a00, 0x3b70, 0x377f, 0x3686, 0x3b7e, 0x38b3, 0x32e3, 0x3323, 0x391e, 0x3228, 0x3930, 0x3997, 0x3a5e, 0x398b, 0x3512, 0x35b0, 0x365c, 0x325d, 0x3b61, 0x38b8, 0x39a4, 0x3423, 0x3bd7, 0x38af, 
+0x2d3d, 0x382d, 0x38ac, 0x26ca, 0x395e, 0x21a8, 0x3520, 0x386f, 0x3b95, 0x32c0, 0x3b84, 0x3a51, 0x3b4b, 0x31d2, 0x3747, 0x3b96, 0x3b40, 0x3535, 0x38d1, 0x3899, 0x3b00, 0x3827, 0x3ae3, 0x38c8, 0x3a07, 0x338d, 0x2e96, 0x3a46, 0x394a, 0x39de, 0x2951, 0x3a02, 
+0x3838, 0x2d45, 0x28c0, 0x3958, 0x3070, 0x2aa2, 0x3510, 0x38ce, 0x271c, 0x3440, 0x3954, 0x30bc, 0x3b35, 0x2f1d, 0x3afb, 0x2dae, 0x356f, 0x2e13, 0x3981, 0x326d, 0x3a28, 0x3a36, 0x3a95, 0x38cb, 0x38db, 0x3150, 0x2c9e, 0x34c5, 0x3adb, 0x3bdf, 0x38f2, 0x3994, 
+0x36f8, 0x31c0, 0x3a4f, 0x3825, 0x394b, 0x3a8b, 0x38ac, 0x3167, 0x2e2d, 0x3a93, 0x34f3, 0x37bd, 0x3b63, 0x2f2f, 0x3ae0, 0x3ad8, 0x34a8, 0x2e1c, 0x3890, 0x3705, 0x3b69, 0x3bc1, 0x28af, 0x3b36, 0x348b, 0x3111, 0x3a8d, 0x389c, 0x3916, 0x36dc, 0x3bae, 0x3874, 
+0x3593, 0x3638, 0x3018, 0x3a56, 0x38a3, 0x2ad4, 0x3a25, 0x38d7, 0x3864, 0x31c1, 0x28d1, 0x39c8, 0x37d6, 0x2c7f, 0x3ba5, 0x34b8, 0x3bef, 0x3b83, 0x3ab5, 0x3062, 0x38bc, 0x399c, 0x2ce4, 0x2f2c, 0x39bf, 0x2ed1, 0x385f, 0x37e0, 0x35ee, 0x397d, 0x3b0c, 0x3049, 
+0x39d5, 0x322e, 0x3936, 0x3747, 0x2e15, 0x3b41, 0x3874, 0x3bd0, 0x2c04, 0x3800, 0x375b, 0x3b2d, 0x38d8, 0x3a51, 0x3406, 0x38da, 0x38ba, 0x3497, 0x382e, 0x35fc, 0x39d4, 0x3775, 0x3b1e, 0x3813, 0x3649, 0x31af, 0x37bb, 0x334a, 0x3a6e, 0x3284, 0x26e0, 0x2e01, 
+0x2ebb, 0x344b, 0x3821, 0x381a, 0x385a, 0x2534, 0x3635, 0x2a92, 0x3b8c, 0x31f0, 0x3947, 0x3ac7, 0x3743, 0x3924, 0x39e4, 0x358f, 0x2b62, 0x392c, 0x3955, 0x3341, 0x3676, 0x38ac, 0x3957, 0x335b, 0x2ca2, 0x39ff, 0x37cb, 0x341f, 0x3ac9, 0x3b6c, 0x2f14, 0x34c3, 
+0x3018, 0x3169, 0x355b, 0x3624, 0x31ed, 0x379e, 0x3268, 0x309b, 0x35db, 0x3872, 0x3bdb, 0x34c7, 0x3408, 0x3359, 0x3920, 0x331f, 0x3866, 0x3af0, 0x2a1a, 0x39e0, 0x3b14, 0x34fa, 0x2d18, 0x3963, 0x35e8, 0x2539, 0x38f5, 0x37b3, 0x378f, 0x31b5, 0x3a6c, 0x3685, 
+0x3a06, 0x318a, 0x2934, 0x33c1, 0x3be8, 0x375b, 0x3860, 0x3543, 0x3702, 0x3951, 0x3677, 0x37ff, 0x2e27, 0x2e3a, 0x340f, 0x3817, 0x2f04, 0x357e, 0x3a1d, 0x2dd6, 0x252a, 0x3945, 0x162a, 0x3b19, 0x3a53, 0x35d2, 0x3a5d, 0x3474, 0x38e9, 0x374b, 0x387c, 0x1f1a, 
+0x38ac, 0x3291, 0x3393, 0x3b53, 0x3169, 0x3bca, 0x2f1a, 0x3551, 0x38a3, 0x28e3, 0x369d, 0x34a1, 0x38a8, 0x34c3, 0x3841, 0x390d, 0x3b13, 0x3282, 0x3a29, 0x3a78, 0x2df3, 0x3a37, 0x35f4, 0x35a6, 0x38e8, 0x3328, 0x3beb, 0x390b, 0x32dc, 0x34dc, 0x396d, 0x3a78, 
+0x39ba, 0x3a06, 0x2cdd, 0x3bc3, 0x2d43, 0x2992, 0x3663, 0x3a68, 0x2c3e, 0x394e, 0x2c9f, 0x380e, 0x37f5, 0x3557, 0x2873, 0x390f, 0x39e7, 0x3939, 0x3669, 0x385c, 0x3a68, 0x32c4, 0x2b04, 0x2d6d, 0x39d3, 0x3895, 0x331d, 0x3b59, 0x3463, 0x2b6a, 0x31de, 0x3296, 
+0x3aae, 0x3bcd, 0x345a, 0x3897, 0x374b, 0x3bd4, 0x38a2, 0x357f, 0x3402, 0x3a0c, 0x3507, 0x3865, 0x3a54, 0x3878, 0x3859, 0x383e, 0x32b5, 0x34ea, 0x328d, 0x38b6, 0x3464, 0x2f5b, 0x35ff, 0x3817, 0x2f24, 0x3533, 0x3b21, 0x37ba, 0x3837, 0x2e34, 0x3bad, 0x34bc 
+};
\ No newline at end of file
diff --git a/hwpe/redmule_softclear/inc/x_input.h b/hwpe/redmule_softclear/inc/x_input.h
new file mode 100644
index 0000000..1e38d23
--- /dev/null
+++ b/hwpe/redmule_softclear/inc/x_input.h
@@ -0,0 +1,27 @@
+ /* Header file generated by RedMulE Golden Model */
+uint16_t x_inp [768] = {
+0x2153, 0x3bb5, 0x3896, 0x365f, 0x2483, 0x3518, 0x2dd1, 0x3bca, 0x397b, 0x29b1, 0x3705, 0x36c8, 0x398b, 0x3661, 0x2f05, 0x365a, 0x3bf9, 0x34df, 0x363b, 0x38d9, 0x39c6, 0x3abb, 0x3952, 0x38f2, 0x392d, 0x3b3e, 0x2afb, 0x3a9d, 0x353b, 0x3b73, 0x3a01, 0x3679, 
+0x3934, 0x397d, 0x2904, 0x3822, 0x3462, 0x3b44, 0x39e9, 0x28be, 0x331e, 0x3a1d, 0x39e5, 0x34da, 0x3a19, 0x3906, 0x1d35, 0x3871, 0x31e7, 0x3b29, 0x325d, 0x3797, 0x2b2f, 0x38b4, 0x232f, 0x38aa, 0x3aca, 0x316f, 0x3811, 0x3950, 0x32ea, 0x3bc7, 0x382c, 0x38a2, 
+0x29ce, 0x3afa, 0x3a39, 0x2ccc, 0x39fd, 0x3b3d, 0x384a, 0x3a35, 0x3802, 0x366a, 0x37ec, 0x3598, 0x3bf8, 0x3a85, 0x3a1b, 0x386e, 0x3b4c, 0x39de, 0x38c2, 0x2f93, 0x3b4c, 0x39c4, 0x3b9e, 0x3844, 0x346d, 0x3bff, 0x32ce, 0x296d, 0x3130, 0x3b3d, 0x3b44, 0x369d, 
+0x3b13, 0x31ed, 0x330a, 0x3831, 0x34e7, 0x37b3, 0x331a, 0x3918, 0x32d3, 0x3995, 0x3991, 0x3919, 0x3a26, 0x385b, 0x2b76, 0x3a3b, 0x37f2, 0x26a7, 0x3225, 0x3b64, 0x28f0, 0x3456, 0x3822, 0x341e, 0x381a, 0x38d8, 0x2c11, 0x33be, 0x33ac, 0x353f, 0x3476, 0x3abc, 
+0x36ec, 0x3a1d, 0x39d3, 0x3821, 0x36ac, 0x3bce, 0x3ad2, 0x3616, 0x36a1, 0x2cb3, 0x38d2, 0x314f, 0x385c, 0x3b63, 0x3bb6, 0x2951, 0x372d, 0x2c42, 0x3823, 0x3883, 0x3872, 0x31ee, 0x36c5, 0x399a, 0x31b0, 0x3887, 0x3884, 0x3865, 0x3896, 0x36c3, 0x32e3, 0x346c, 
+0x3935, 0x3b50, 0x2b6d, 0x38cd, 0x388f, 0x3389, 0x395d, 0x31cd, 0x2efd, 0x3154, 0x2f35, 0x3444, 0x3293, 0x3b6b, 0x1bec, 0x3b69, 0x3bf3, 0x3611, 0x3508, 0x3742, 0x3a50, 0x3ab7, 0x3457, 0x38d3, 0x3344, 0x38e8, 0x33c0, 0x3668, 0x3bee, 0x3b21, 0x3727, 0x3121, 
+0x316c, 0x3288, 0x2d50, 0x2e74, 0x35d5, 0x37e2, 0x303d, 0x36af, 0x341f, 0x3436, 0x2df7, 0x399d, 0x30f4, 0x3aaf, 0x34e4, 0x2c2a, 0x3116, 0x34d3, 0x36ac, 0x35e3, 0x3760, 0x36e1, 0x3ad2, 0x3547, 0x38f4, 0x369c, 0x3ba9, 0x34f0, 0x3a39, 0x3b19, 0x36e6, 0x395d, 
+0x3be8, 0x3293, 0x3bfc, 0x3435, 0x2eb3, 0x3360, 0x3919, 0x3bed, 0x396a, 0x37fc, 0x3242, 0x384b, 0x38cb, 0x3b2c, 0x3b28, 0x28cf, 0x3828, 0x3855, 0x3ba9, 0x2fa7, 0x340b, 0x32f1, 0x3ada, 0x36fa, 0x31f5, 0x3436, 0x29d0, 0x33e6, 0x3232, 0x3bec, 0x3904, 0x2797, 
+0x3b81, 0x3bac, 0x38d2, 0x343d, 0x31af, 0x3b1e, 0x33fc, 0x3864, 0x3624, 0x3905, 0x2945, 0x3b52, 0x2d08, 0x3a17, 0x3b84, 0x3804, 0x3a24, 0x38a3, 0x3562, 0x3ae6, 0x3bba, 0x3a45, 0x3679, 0x31fa, 0x3994, 0x2c3d, 0x383f, 0x399d, 0x34f7, 0x360e, 0x35f3, 0x38f0, 
+0x38d4, 0x399a, 0x3a48, 0x3987, 0x3b54, 0x382c, 0x3210, 0x35ef, 0x36ca, 0x31b4, 0x3625, 0x371f, 0x37bd, 0x3680, 0x3a3a, 0x3ac0, 0x3bbf, 0x3bf5, 0x39f2, 0x29c2, 0x363e, 0x3a4e, 0x3596, 0x3b1b, 0x3459, 0x3669, 0x3aa1, 0x39c3, 0x3376, 0x390d, 0x2456, 0x39b5, 
+0x3a66, 0x3ad8, 0x3b51, 0x36aa, 0x32be, 0x3ac8, 0x392b, 0x3740, 0x3a48, 0x38f5, 0x3b2d, 0x3a5f, 0x2ff3, 0x366f, 0x39d3, 0x35e5, 0x3822, 0x38db, 0x3b8a, 0x34be, 0x2d33, 0x36dd, 0x3578, 0x3bdf, 0x2c7e, 0x39cf, 0x32ff, 0x35c9, 0x3970, 0x3bcb, 0x351e, 0x3956, 
+0x2c42, 0x3308, 0x377a, 0x361c, 0x39a0, 0x36c9, 0x2dcb, 0x3bf2, 0x3b5f, 0x33ee, 0x24c1, 0x2ce9, 0x3927, 0x305d, 0x3702, 0x3119, 0x35f9, 0x3855, 0x3374, 0x349b, 0x3bcf, 0x2dea, 0x34f0, 0x363f, 0x37da, 0x3a74, 0x35fc, 0x35fa, 0x316b, 0x3804, 0x37a7, 0x3986, 
+0x3073, 0x3aed, 0x31c7, 0x3844, 0x34a4, 0x387d, 0x3a20, 0x3037, 0x3a00, 0x3b70, 0x377f, 0x3686, 0x3b7e, 0x38b3, 0x32e3, 0x3323, 0x391e, 0x3228, 0x3930, 0x3997, 0x3a5e, 0x398b, 0x3512, 0x35b0, 0x365c, 0x325d, 0x3b61, 0x38b8, 0x39a4, 0x3423, 0x3bd7, 0x38af, 
+0x2d3d, 0x382d, 0x38ac, 0x26ca, 0x395e, 0x21a8, 0x3520, 0x386f, 0x3b95, 0x32c0, 0x3b84, 0x3a51, 0x3b4b, 0x31d2, 0x3747, 0x3b96, 0x3b40, 0x3535, 0x38d1, 0x3899, 0x3b00, 0x3827, 0x3ae3, 0x38c8, 0x3a07, 0x338d, 0x2e96, 0x3a46, 0x394a, 0x39de, 0x2951, 0x3a02, 
+0x3838, 0x2d45, 0x28c0, 0x3958, 0x3070, 0x2aa2, 0x3510, 0x38ce, 0x271c, 0x3440, 0x3954, 0x30bc, 0x3b35, 0x2f1d, 0x3afb, 0x2dae, 0x356f, 0x2e13, 0x3981, 0x326d, 0x3a28, 0x3a36, 0x3a95, 0x38cb, 0x38db, 0x3150, 0x2c9e, 0x34c5, 0x3adb, 0x3bdf, 0x38f2, 0x3994, 
+0x36f8, 0x31c0, 0x3a4f, 0x3825, 0x394b, 0x3a8b, 0x38ac, 0x3167, 0x2e2d, 0x3a93, 0x34f3, 0x37bd, 0x3b63, 0x2f2f, 0x3ae0, 0x3ad8, 0x34a8, 0x2e1c, 0x3890, 0x3705, 0x3b69, 0x3bc1, 0x28af, 0x3b36, 0x348b, 0x3111, 0x3a8d, 0x389c, 0x3916, 0x36dc, 0x3bae, 0x3874, 
+0x3593, 0x3638, 0x3018, 0x3a56, 0x38a3, 0x2ad4, 0x3a25, 0x38d7, 0x3864, 0x31c1, 0x28d1, 0x39c8, 0x37d6, 0x2c7f, 0x3ba5, 0x34b8, 0x3bef, 0x3b83, 0x3ab5, 0x3062, 0x38bc, 0x399c, 0x2ce4, 0x2f2c, 0x39bf, 0x2ed1, 0x385f, 0x37e0, 0x35ee, 0x397d, 0x3b0c, 0x3049, 
+0x39d5, 0x322e, 0x3936, 0x3747, 0x2e15, 0x3b41, 0x3874, 0x3bd0, 0x2c04, 0x3800, 0x375b, 0x3b2d, 0x38d8, 0x3a51, 0x3406, 0x38da, 0x38ba, 0x3497, 0x382e, 0x35fc, 0x39d4, 0x3775, 0x3b1e, 0x3813, 0x3649, 0x31af, 0x37bb, 0x334a, 0x3a6e, 0x3284, 0x26e0, 0x2e01, 
+0x2ebb, 0x344b, 0x3821, 0x381a, 0x385a, 0x2534, 0x3635, 0x2a92, 0x3b8c, 0x31f0, 0x3947, 0x3ac7, 0x3743, 0x3924, 0x39e4, 0x358f, 0x2b62, 0x392c, 0x3955, 0x3341, 0x3676, 0x38ac, 0x3957, 0x335b, 0x2ca2, 0x39ff, 0x37cb, 0x341f, 0x3ac9, 0x3b6c, 0x2f14, 0x34c3, 
+0x3018, 0x3169, 0x355b, 0x3624, 0x31ed, 0x379e, 0x3268, 0x309b, 0x35db, 0x3872, 0x3bdb, 0x34c7, 0x3408, 0x3359, 0x3920, 0x331f, 0x3866, 0x3af0, 0x2a1a, 0x39e0, 0x3b14, 0x34fa, 0x2d18, 0x3963, 0x35e8, 0x2539, 0x38f5, 0x37b3, 0x378f, 0x31b5, 0x3a6c, 0x3685, 
+0x3a06, 0x318a, 0x2934, 0x33c1, 0x3be8, 0x375b, 0x3860, 0x3543, 0x3702, 0x3951, 0x3677, 0x37ff, 0x2e27, 0x2e3a, 0x340f, 0x3817, 0x2f04, 0x357e, 0x3a1d, 0x2dd6, 0x252a, 0x3945, 0x162a, 0x3b19, 0x3a53, 0x35d2, 0x3a5d, 0x3474, 0x38e9, 0x374b, 0x387c, 0x1f1a, 
+0x38ac, 0x3291, 0x3393, 0x3b53, 0x3169, 0x3bca, 0x2f1a, 0x3551, 0x38a3, 0x28e3, 0x369d, 0x34a1, 0x38a8, 0x34c3, 0x3841, 0x390d, 0x3b13, 0x3282, 0x3a29, 0x3a78, 0x2df3, 0x3a37, 0x35f4, 0x35a6, 0x38e8, 0x3328, 0x3beb, 0x390b, 0x32dc, 0x34dc, 0x396d, 0x3a78, 
+0x39ba, 0x3a06, 0x2cdd, 0x3bc3, 0x2d43, 0x2992, 0x3663, 0x3a68, 0x2c3e, 0x394e, 0x2c9f, 0x380e, 0x37f5, 0x3557, 0x2873, 0x390f, 0x39e7, 0x3939, 0x3669, 0x385c, 0x3a68, 0x32c4, 0x2b04, 0x2d6d, 0x39d3, 0x3895, 0x331d, 0x3b59, 0x3463, 0x2b6a, 0x31de, 0x3296, 
+0x3aae, 0x3bcd, 0x345a, 0x3897, 0x374b, 0x3bd4, 0x38a2, 0x357f, 0x3402, 0x3a0c, 0x3507, 0x3865, 0x3a54, 0x3878, 0x3859, 0x383e, 0x32b5, 0x34ea, 0x328d, 0x38b6, 0x3464, 0x2f5b, 0x35ff, 0x3817, 0x2f24, 0x3533, 0x3b21, 0x37ba, 0x3837, 0x2e34, 0x3bad, 0x34bc 
+};
\ No newline at end of file
diff --git a/hwpe/redmule_softclear/inc/y_2D.h b/hwpe/redmule_softclear/inc/y_2D.h
new file mode 100644
index 0000000..9484a10
--- /dev/null
+++ b/hwpe/redmule_softclear/inc/y_2D.h
@@ -0,0 +1,27 @@
+ /* Header file generated by RedMulE Golden Model */
+uint16_t y_inp_2D [32][32] = {
+0x3150, 0x2dc1, 0x3033, 0x31f5, 0x3bb6, 0x3bff, 0x39f9, 0x3662, 0x3720, 0x351d, 0x384b, 0x3093, 0x3b9d, 0x35ad, 0x3695, 0x3466, 0x2300, 0x3445, 0x33ae, 0x3586, 0x38a3, 0x3bdb, 0x33a2, 0x379b, 0x3a0e, 0x38b0, 0x39ba, 0x379b, 0x39d3, 0x3a51, 0x3b30, 0x3794, 
+0x3b76, 0x3042, 0x38cc, 0x2dfc, 0x3b1a, 0x37fb, 0x38f7, 0x3824, 0x386f, 0x38c7, 0x36ee, 0x3a9c, 0x38d3, 0x2c67, 0x3a80, 0x2f30, 0x3328, 0x3721, 0x3790, 0x34e5, 0x3a6c, 0x3643, 0x3934, 0x3034, 0x38d4, 0x362e, 0x3b4b, 0x3408, 0x30c2, 0x370e, 0x3b31, 0x3b16, 
+0x3b6b, 0x39d4, 0x339c, 0x381e, 0x313e, 0x3671, 0x3ae2, 0x3479, 0x3940, 0x342d, 0x3925, 0x370a, 0x35d8, 0x2dad, 0x3888, 0x24b9, 0x375d, 0x34bd, 0x3243, 0x2ebb, 0x3970, 0x3a21, 0x3a07, 0x3877, 0x3888, 0x3569, 0x372d, 0x2ac1, 0x331e, 0x384d, 0x3996, 0x34a4, 
+0x35c1, 0x33a9, 0x21ed, 0x3a42, 0x388d, 0x34e4, 0x33c3, 0x34f9, 0x3a7b, 0x33fb, 0x2cdd, 0x3b0e, 0x333b, 0x3973, 0x34fc, 0x3771, 0x32ea, 0x2de4, 0x31a8, 0x3946, 0x3657, 0x3a4e, 0x36f6, 0x2829, 0x3ba2, 0x3bdc, 0x3bb3, 0x306c, 0x398d, 0x3a1f, 0x3991, 0x3846, 
+0x3547, 0x3292, 0x2e85, 0x31ed, 0x3979, 0x3a90, 0x28a4, 0x3bed, 0x36d8, 0x340e, 0x3b6a, 0x3ab6, 0x3824, 0x382b, 0x3ac3, 0x3811, 0x36d7, 0x3519, 0x3a92, 0x3a42, 0x29d1, 0x383a, 0x3a9b, 0x300e, 0x2cd3, 0x39cd, 0x3874, 0x3a07, 0x2eb1, 0x3b86, 0x3ad8, 0x3a5d, 
+0x3712, 0x284a, 0x38c1, 0x3bec, 0x39c0, 0x32cd, 0x3ad8, 0x3bce, 0x3817, 0x3896, 0x3aa7, 0x3870, 0x3996, 0x32cc, 0x3a4c, 0x3757, 0x3814, 0x3b65, 0x3acb, 0x376e, 0x34c0, 0x3609, 0x3bf0, 0x3b24, 0x3b29, 0x3848, 0x34b7, 0x398a, 0x220c, 0x3498, 0x3a8c, 0x3883, 
+0x38c4, 0x3af6, 0x3a42, 0x2dd6, 0x3147, 0x3717, 0x3a8e, 0x3af9, 0x3296, 0x38ef, 0x34fa, 0x3555, 0x3b29, 0x38de, 0x315e, 0x3773, 0x3b67, 0x3116, 0x38ec, 0x357c, 0x35d0, 0x2518, 0x3958, 0x2a03, 0x37d9, 0x3699, 0x3a1e, 0x3230, 0x3b13, 0x36d4, 0x3b2a, 0x39ad, 
+0x3b10, 0x351a, 0x3b97, 0x3326, 0x2b54, 0x3b7d, 0x386f, 0x373e, 0x37fa, 0x389b, 0x3b90, 0x3292, 0x3975, 0x38f3, 0x37f1, 0x3590, 0x3810, 0x2fd7, 0x3bf7, 0x3a5a, 0x3a1c, 0x34dd, 0x354c, 0x32f8, 0x3095, 0x321e, 0x39e0, 0x395c, 0x3717, 0x357f, 0x394a, 0x34b1, 
+0x3ba4, 0x380c, 0x3604, 0x2f50, 0x348d, 0x3828, 0x3a9f, 0x39ce, 0x32ca, 0x3906, 0x3ab2, 0x2ca5, 0x38c9, 0x362a, 0x34b2, 0x29dc, 0x3a36, 0x3052, 0x31b7, 0x3589, 0x387c, 0x3401, 0x3b22, 0x3ad6, 0x3ae8, 0x3238, 0x3494, 0x3502, 0x3717, 0x3a6c, 0x3229, 0x368c, 
+0x3056, 0x3a56, 0x3498, 0x39eb, 0x2864, 0x342d, 0x39e0, 0x34a1, 0x2b99, 0x3a04, 0x38ff, 0x328c, 0x34d9, 0x387d, 0x3a3c, 0x32e5, 0x39eb, 0x3984, 0x34dd, 0x38a7, 0x373f, 0x39b4, 0x3235, 0x2f58, 0x2f39, 0x3800, 0x3758, 0x3939, 0x39fc, 0x3a4b, 0x38bf, 0x30ee, 
+0x345e, 0x39c8, 0x3a6d, 0x3262, 0x3b81, 0x31dc, 0x3a15, 0x3bd0, 0x36af, 0x36de, 0x37d5, 0x39d7, 0x3ad3, 0x3ac1, 0x3109, 0x35ea, 0x31c6, 0x398d, 0x3987, 0x3a4a, 0x34d2, 0x2ed2, 0x35e6, 0x352c, 0x39eb, 0x3bd6, 0x3a5b, 0x39d1, 0x34aa, 0x3ade, 0x394b, 0x38a1, 
+0x2bed, 0x38de, 0x3811, 0x3813, 0x391a, 0x374b, 0x3829, 0x3725, 0x38f0, 0x3583, 0x3966, 0x3a7d, 0x375a, 0x38fe, 0x3696, 0x361c, 0x39a8, 0x35f0, 0x38e1, 0x3003, 0x3595, 0x316e, 0x3862, 0x3af8, 0x3af2, 0x34c8, 0x381d, 0x37d8, 0x3893, 0x3a9c, 0x3989, 0x308c, 
+0x30cc, 0x2538, 0x399d, 0x3919, 0x399e, 0x21cc, 0x38e9, 0x30f8, 0x3a20, 0x3b3c, 0x3990, 0x259c, 0x3143, 0x3080, 0x3967, 0x3afb, 0x3a1b, 0x3779, 0x2eeb, 0x39f3, 0x379a, 0x369c, 0x3985, 0x3a1b, 0x3ba6, 0x3a53, 0x28d5, 0x3881, 0x31d9, 0x3a34, 0x3bd9, 0x393a, 
+0x3601, 0x2c6e, 0x3636, 0x3298, 0x39bb, 0x3a08, 0x38db, 0x35ad, 0x3a09, 0x36a6, 0x3bc7, 0x3bac, 0x34ae, 0x3291, 0x290b, 0x3250, 0x2648, 0x333d, 0x2bf3, 0x34b1, 0x30e0, 0x351f, 0x3a74, 0x38dc, 0x3883, 0x2841, 0x35e1, 0x390d, 0x3a50, 0x3abd, 0x386d, 0x3bb7, 
+0x3b94, 0x36b7, 0x3a49, 0x332f, 0x3a1d, 0x354b, 0x3bab, 0x3346, 0x3417, 0x351e, 0x3b6d, 0x391a, 0x2db3, 0x3b1c, 0x3a4a, 0x37b7, 0x36cf, 0x3a56, 0x39c4, 0x3be9, 0x34f0, 0x39be, 0x3691, 0x1ba5, 0x3888, 0x3040, 0x3ae1, 0x3b9b, 0x398f, 0x3a49, 0x3a16, 0x38c0, 
+0x386c, 0x39ab, 0x37fa, 0x382c, 0x3a6f, 0x393f, 0x340d, 0x38ef, 0x39d1, 0x3845, 0x398f, 0x363e, 0x3687, 0x3052, 0x3a2b, 0x392c, 0x2f5c, 0x3412, 0x3a1f, 0x3b2f, 0x3bcc, 0x3a63, 0x3a89, 0x36e9, 0x3921, 0x3b80, 0x2dc0, 0x3a03, 0x3beb, 0x38d3, 0x36cb, 0x39a3, 
+0x3978, 0x3a88, 0x3ba4, 0x3561, 0x28c5, 0x33a0, 0x37be, 0x2c39, 0x30ee, 0x3782, 0x2c07, 0x354e, 0x3491, 0x3a92, 0x331a, 0x3b15, 0x32e1, 0x3839, 0x3afb, 0x36c2, 0x2fd0, 0x29ad, 0x3b2e, 0x39c1, 0x2a8c, 0x341a, 0x2f90, 0x395a, 0x3969, 0x37ea, 0x3a5c, 0x3b6d, 
+0x3971, 0x3a93, 0x304e, 0x3623, 0x3a22, 0x31ee, 0x29df, 0x2c93, 0x3a01, 0x3a62, 0x366c, 0x371d, 0x3af3, 0x2e08, 0x3ac0, 0x3642, 0x3a28, 0x368d, 0x2d3d, 0x36d9, 0x32c3, 0x373f, 0x36fe, 0x3487, 0x2c81, 0x3623, 0x3b59, 0x3a91, 0x350a, 0x34f4, 0x3b09, 0x2c25, 
+0x3b13, 0x325a, 0x379e, 0x3a7d, 0x34b1, 0x39d5, 0x2ba8, 0x322b, 0x3b5e, 0x37ab, 0x2e24, 0x3ba9, 0x3a3d, 0x34f7, 0x3ba1, 0x3877, 0x3071, 0x39fb, 0x3bbd, 0x3633, 0x3b36, 0x2daa, 0x3b9b, 0x3aa0, 0x395c, 0x3b8f, 0x38d5, 0x3ab0, 0x3a8f, 0x36c2, 0x3b1f, 0x3489, 
+0x2acc, 0x3845, 0x3715, 0x37d8, 0x3992, 0x3bff, 0x350e, 0x3ad7, 0x39b0, 0x35ac, 0x3287, 0x385f, 0x3bd4, 0x37a3, 0x3438, 0x39a5, 0x3bcf, 0x38c3, 0x34f6, 0x3ae3, 0x3b57, 0x39af, 0x35eb, 0x3bed, 0x34d4, 0x2a95, 0x3b13, 0x384e, 0x3a3b, 0x33da, 0x3bce, 0x3b99, 
+0x3559, 0x3335, 0x3a2e, 0x3123, 0x38db, 0x33d0, 0x3638, 0x3b17, 0x3a72, 0x3afc, 0x3936, 0x3838, 0x2b69, 0x3895, 0x3a1a, 0x3192, 0x39d5, 0x37a5, 0x2eb0, 0x2e8b, 0x329a, 0x3b90, 0x390a, 0x3a1e, 0x3847, 0x375d, 0x3873, 0x35e2, 0x3771, 0x30f5, 0x3231, 0x3bd7, 
+0x2bbc, 0x3ace, 0x31ad, 0x3a6b, 0x28a4, 0x3b48, 0x3ba3, 0x3a84, 0x3353, 0x39f6, 0x381f, 0x2dd6, 0x314c, 0x34af, 0x3929, 0x3921, 0x383b, 0x34b0, 0x3923, 0x32c9, 0x3ae7, 0x318f, 0x3480, 0x2ad8, 0x3042, 0x3a4c, 0x349d, 0x2c12, 0x3abb, 0x3a57, 0x3b0d, 0x3111, 
+0x3359, 0x3a84, 0x38f2, 0x368d, 0x2f4b, 0x3ba0, 0x395c, 0x3026, 0x3a15, 0x2a04, 0x326e, 0x3522, 0x31a2, 0x382f, 0x2ada, 0x3b7c, 0x2f80, 0x3af5, 0x2d35, 0x38fa, 0x39ab, 0x2c6d, 0x2e7a, 0x39f6, 0x31a4, 0x3a53, 0x358c, 0x3951, 0x3a4e, 0x3916, 0x2a3f, 0x3ae9, 
+0x3b03, 0x39f8, 0x39fe, 0x3a61, 0x39fb, 0x3704, 0x360d, 0x39a7, 0x37a9, 0x348f, 0x3a30, 0x3af5, 0x366f, 0x3b29, 0x3a6a, 0x33d5, 0x370a, 0x39cd, 0x3444, 0x3bea, 0x3b2b, 0x312e, 0x3b8e, 0x32cf, 0x3b79, 0x3302, 0x3bba, 0x3962, 0x3413, 0x37a1, 0x39e0, 0x3805 
+};
\ No newline at end of file
diff --git a/hwpe/redmule_softclear/inc/y_input.h b/hwpe/redmule_softclear/inc/y_input.h
new file mode 100644
index 0000000..45a2375
--- /dev/null
+++ b/hwpe/redmule_softclear/inc/y_input.h
@@ -0,0 +1,27 @@
+ /* Header file generated by RedMulE Golden Model */
+uint16_t y_inp [768] = {
+0x3150, 0x2dc1, 0x3033, 0x31f5, 0x3bb6, 0x3bff, 0x39f9, 0x3662, 0x3720, 0x351d, 0x384b, 0x3093, 0x3b9d, 0x35ad, 0x3695, 0x3466, 0x2300, 0x3445, 0x33ae, 0x3586, 0x38a3, 0x3bdb, 0x33a2, 0x379b, 0x3a0e, 0x38b0, 0x39ba, 0x379b, 0x39d3, 0x3a51, 0x3b30, 0x3794, 
+0x3b76, 0x3042, 0x38cc, 0x2dfc, 0x3b1a, 0x37fb, 0x38f7, 0x3824, 0x386f, 0x38c7, 0x36ee, 0x3a9c, 0x38d3, 0x2c67, 0x3a80, 0x2f30, 0x3328, 0x3721, 0x3790, 0x34e5, 0x3a6c, 0x3643, 0x3934, 0x3034, 0x38d4, 0x362e, 0x3b4b, 0x3408, 0x30c2, 0x370e, 0x3b31, 0x3b16, 
+0x3b6b, 0x39d4, 0x339c, 0x381e, 0x313e, 0x3671, 0x3ae2, 0x3479, 0x3940, 0x342d, 0x3925, 0x370a, 0x35d8, 0x2dad, 0x3888, 0x24b9, 0x375d, 0x34bd, 0x3243, 0x2ebb, 0x3970, 0x3a21, 0x3a07, 0x3877, 0x3888, 0x3569, 0x372d, 0x2ac1, 0x331e, 0x384d, 0x3996, 0x34a4, 
+0x35c1, 0x33a9, 0x21ed, 0x3a42, 0x388d, 0x34e4, 0x33c3, 0x34f9, 0x3a7b, 0x33fb, 0x2cdd, 0x3b0e, 0x333b, 0x3973, 0x34fc, 0x3771, 0x32ea, 0x2de4, 0x31a8, 0x3946, 0x3657, 0x3a4e, 0x36f6, 0x2829, 0x3ba2, 0x3bdc, 0x3bb3, 0x306c, 0x398d, 0x3a1f, 0x3991, 0x3846, 
+0x3547, 0x3292, 0x2e85, 0x31ed, 0x3979, 0x3a90, 0x28a4, 0x3bed, 0x36d8, 0x340e, 0x3b6a, 0x3ab6, 0x3824, 0x382b, 0x3ac3, 0x3811, 0x36d7, 0x3519, 0x3a92, 0x3a42, 0x29d1, 0x383a, 0x3a9b, 0x300e, 0x2cd3, 0x39cd, 0x3874, 0x3a07, 0x2eb1, 0x3b86, 0x3ad8, 0x3a5d, 
+0x3712, 0x284a, 0x38c1, 0x3bec, 0x39c0, 0x32cd, 0x3ad8, 0x3bce, 0x3817, 0x3896, 0x3aa7, 0x3870, 0x3996, 0x32cc, 0x3a4c, 0x3757, 0x3814, 0x3b65, 0x3acb, 0x376e, 0x34c0, 0x3609, 0x3bf0, 0x3b24, 0x3b29, 0x3848, 0x34b7, 0x398a, 0x220c, 0x3498, 0x3a8c, 0x3883, 
+0x38c4, 0x3af6, 0x3a42, 0x2dd6, 0x3147, 0x3717, 0x3a8e, 0x3af9, 0x3296, 0x38ef, 0x34fa, 0x3555, 0x3b29, 0x38de, 0x315e, 0x3773, 0x3b67, 0x3116, 0x38ec, 0x357c, 0x35d0, 0x2518, 0x3958, 0x2a03, 0x37d9, 0x3699, 0x3a1e, 0x3230, 0x3b13, 0x36d4, 0x3b2a, 0x39ad, 
+0x3b10, 0x351a, 0x3b97, 0x3326, 0x2b54, 0x3b7d, 0x386f, 0x373e, 0x37fa, 0x389b, 0x3b90, 0x3292, 0x3975, 0x38f3, 0x37f1, 0x3590, 0x3810, 0x2fd7, 0x3bf7, 0x3a5a, 0x3a1c, 0x34dd, 0x354c, 0x32f8, 0x3095, 0x321e, 0x39e0, 0x395c, 0x3717, 0x357f, 0x394a, 0x34b1, 
+0x3ba4, 0x380c, 0x3604, 0x2f50, 0x348d, 0x3828, 0x3a9f, 0x39ce, 0x32ca, 0x3906, 0x3ab2, 0x2ca5, 0x38c9, 0x362a, 0x34b2, 0x29dc, 0x3a36, 0x3052, 0x31b7, 0x3589, 0x387c, 0x3401, 0x3b22, 0x3ad6, 0x3ae8, 0x3238, 0x3494, 0x3502, 0x3717, 0x3a6c, 0x3229, 0x368c, 
+0x3056, 0x3a56, 0x3498, 0x39eb, 0x2864, 0x342d, 0x39e0, 0x34a1, 0x2b99, 0x3a04, 0x38ff, 0x328c, 0x34d9, 0x387d, 0x3a3c, 0x32e5, 0x39eb, 0x3984, 0x34dd, 0x38a7, 0x373f, 0x39b4, 0x3235, 0x2f58, 0x2f39, 0x3800, 0x3758, 0x3939, 0x39fc, 0x3a4b, 0x38bf, 0x30ee, 
+0x345e, 0x39c8, 0x3a6d, 0x3262, 0x3b81, 0x31dc, 0x3a15, 0x3bd0, 0x36af, 0x36de, 0x37d5, 0x39d7, 0x3ad3, 0x3ac1, 0x3109, 0x35ea, 0x31c6, 0x398d, 0x3987, 0x3a4a, 0x34d2, 0x2ed2, 0x35e6, 0x352c, 0x39eb, 0x3bd6, 0x3a5b, 0x39d1, 0x34aa, 0x3ade, 0x394b, 0x38a1, 
+0x2bed, 0x38de, 0x3811, 0x3813, 0x391a, 0x374b, 0x3829, 0x3725, 0x38f0, 0x3583, 0x3966, 0x3a7d, 0x375a, 0x38fe, 0x3696, 0x361c, 0x39a8, 0x35f0, 0x38e1, 0x3003, 0x3595, 0x316e, 0x3862, 0x3af8, 0x3af2, 0x34c8, 0x381d, 0x37d8, 0x3893, 0x3a9c, 0x3989, 0x308c, 
+0x30cc, 0x2538, 0x399d, 0x3919, 0x399e, 0x21cc, 0x38e9, 0x30f8, 0x3a20, 0x3b3c, 0x3990, 0x259c, 0x3143, 0x3080, 0x3967, 0x3afb, 0x3a1b, 0x3779, 0x2eeb, 0x39f3, 0x379a, 0x369c, 0x3985, 0x3a1b, 0x3ba6, 0x3a53, 0x28d5, 0x3881, 0x31d9, 0x3a34, 0x3bd9, 0x393a, 
+0x3601, 0x2c6e, 0x3636, 0x3298, 0x39bb, 0x3a08, 0x38db, 0x35ad, 0x3a09, 0x36a6, 0x3bc7, 0x3bac, 0x34ae, 0x3291, 0x290b, 0x3250, 0x2648, 0x333d, 0x2bf3, 0x34b1, 0x30e0, 0x351f, 0x3a74, 0x38dc, 0x3883, 0x2841, 0x35e1, 0x390d, 0x3a50, 0x3abd, 0x386d, 0x3bb7, 
+0x3b94, 0x36b7, 0x3a49, 0x332f, 0x3a1d, 0x354b, 0x3bab, 0x3346, 0x3417, 0x351e, 0x3b6d, 0x391a, 0x2db3, 0x3b1c, 0x3a4a, 0x37b7, 0x36cf, 0x3a56, 0x39c4, 0x3be9, 0x34f0, 0x39be, 0x3691, 0x1ba5, 0x3888, 0x3040, 0x3ae1, 0x3b9b, 0x398f, 0x3a49, 0x3a16, 0x38c0, 
+0x386c, 0x39ab, 0x37fa, 0x382c, 0x3a6f, 0x393f, 0x340d, 0x38ef, 0x39d1, 0x3845, 0x398f, 0x363e, 0x3687, 0x3052, 0x3a2b, 0x392c, 0x2f5c, 0x3412, 0x3a1f, 0x3b2f, 0x3bcc, 0x3a63, 0x3a89, 0x36e9, 0x3921, 0x3b80, 0x2dc0, 0x3a03, 0x3beb, 0x38d3, 0x36cb, 0x39a3, 
+0x3978, 0x3a88, 0x3ba4, 0x3561, 0x28c5, 0x33a0, 0x37be, 0x2c39, 0x30ee, 0x3782, 0x2c07, 0x354e, 0x3491, 0x3a92, 0x331a, 0x3b15, 0x32e1, 0x3839, 0x3afb, 0x36c2, 0x2fd0, 0x29ad, 0x3b2e, 0x39c1, 0x2a8c, 0x341a, 0x2f90, 0x395a, 0x3969, 0x37ea, 0x3a5c, 0x3b6d, 
+0x3971, 0x3a93, 0x304e, 0x3623, 0x3a22, 0x31ee, 0x29df, 0x2c93, 0x3a01, 0x3a62, 0x366c, 0x371d, 0x3af3, 0x2e08, 0x3ac0, 0x3642, 0x3a28, 0x368d, 0x2d3d, 0x36d9, 0x32c3, 0x373f, 0x36fe, 0x3487, 0x2c81, 0x3623, 0x3b59, 0x3a91, 0x350a, 0x34f4, 0x3b09, 0x2c25, 
+0x3b13, 0x325a, 0x379e, 0x3a7d, 0x34b1, 0x39d5, 0x2ba8, 0x322b, 0x3b5e, 0x37ab, 0x2e24, 0x3ba9, 0x3a3d, 0x34f7, 0x3ba1, 0x3877, 0x3071, 0x39fb, 0x3bbd, 0x3633, 0x3b36, 0x2daa, 0x3b9b, 0x3aa0, 0x395c, 0x3b8f, 0x38d5, 0x3ab0, 0x3a8f, 0x36c2, 0x3b1f, 0x3489, 
+0x2acc, 0x3845, 0x3715, 0x37d8, 0x3992, 0x3bff, 0x350e, 0x3ad7, 0x39b0, 0x35ac, 0x3287, 0x385f, 0x3bd4, 0x37a3, 0x3438, 0x39a5, 0x3bcf, 0x38c3, 0x34f6, 0x3ae3, 0x3b57, 0x39af, 0x35eb, 0x3bed, 0x34d4, 0x2a95, 0x3b13, 0x384e, 0x3a3b, 0x33da, 0x3bce, 0x3b99, 
+0x3559, 0x3335, 0x3a2e, 0x3123, 0x38db, 0x33d0, 0x3638, 0x3b17, 0x3a72, 0x3afc, 0x3936, 0x3838, 0x2b69, 0x3895, 0x3a1a, 0x3192, 0x39d5, 0x37a5, 0x2eb0, 0x2e8b, 0x329a, 0x3b90, 0x390a, 0x3a1e, 0x3847, 0x375d, 0x3873, 0x35e2, 0x3771, 0x30f5, 0x3231, 0x3bd7, 
+0x2bbc, 0x3ace, 0x31ad, 0x3a6b, 0x28a4, 0x3b48, 0x3ba3, 0x3a84, 0x3353, 0x39f6, 0x381f, 0x2dd6, 0x314c, 0x34af, 0x3929, 0x3921, 0x383b, 0x34b0, 0x3923, 0x32c9, 0x3ae7, 0x318f, 0x3480, 0x2ad8, 0x3042, 0x3a4c, 0x349d, 0x2c12, 0x3abb, 0x3a57, 0x3b0d, 0x3111, 
+0x3359, 0x3a84, 0x38f2, 0x368d, 0x2f4b, 0x3ba0, 0x395c, 0x3026, 0x3a15, 0x2a04, 0x326e, 0x3522, 0x31a2, 0x382f, 0x2ada, 0x3b7c, 0x2f80, 0x3af5, 0x2d35, 0x38fa, 0x39ab, 0x2c6d, 0x2e7a, 0x39f6, 0x31a4, 0x3a53, 0x358c, 0x3951, 0x3a4e, 0x3916, 0x2a3f, 0x3ae9, 
+0x3b03, 0x39f8, 0x39fe, 0x3a61, 0x39fb, 0x3704, 0x360d, 0x39a7, 0x37a9, 0x348f, 0x3a30, 0x3af5, 0x366f, 0x3b29, 0x3a6a, 0x33d5, 0x370a, 0x39cd, 0x3444, 0x3bea, 0x3b2b, 0x312e, 0x3b8e, 0x32cf, 0x3b79, 0x3302, 0x3bba, 0x3962, 0x3413, 0x37a1, 0x39e0, 0x3805 
+};
\ No newline at end of file
diff --git a/hwpe/redmule_softclear/inc/z_2D.h b/hwpe/redmule_softclear/inc/z_2D.h
new file mode 100644
index 0000000..aff808a
--- /dev/null
+++ b/hwpe/redmule_softclear/inc/z_2D.h
@@ -0,0 +1,27 @@
+ /* Header file generated by RedMulE Golden Model */
+uint16_t z_oup_2D [24][32] = {
+0x4845, 0x4897, 0x4608, 0x4838, 0x4855, 0x487b, 0x4869, 0x4880, 0x46d1, 0x48b0, 0x48db, 0x483f, 0x48c9, 0x485f, 0x4881, 0x483a, 0x484b, 0x472c, 0x4762, 0x492b, 0x4822, 0x48fd, 0x488e, 0x492e, 0x483e, 0x484f, 0x49e8, 0x46d7, 0x484b, 0x489d, 0x490b, 0x47e9, 
+0x484f, 0x47d2, 0x44be, 0x4747, 0x47c7, 0x46c0, 0x4727, 0x48af, 0x46c5, 0x482d, 0x483d, 0x482e, 0x4897, 0x479f, 0x488b, 0x4749, 0x489a, 0x46a8, 0x46f2, 0x488b, 0x4891, 0x47e8, 0x4872, 0x483d, 0x4716, 0x46fd, 0x49b5, 0x46a0, 0x46e7, 0x47a4, 0x48a1, 0x4767, 
+0x4939, 0x4935, 0x4703, 0x48c1, 0x4863, 0x48bd, 0x4913, 0x48cf, 0x48b6, 0x48b8, 0x4946, 0x4920, 0x495e, 0x48e1, 0x4938, 0x48b2, 0x493a, 0x4882, 0x483b, 0x49d5, 0x4911, 0x4972, 0x496b, 0x49df, 0x48f2, 0x4888, 0x4a46, 0x4821, 0x48c1, 0x490c, 0x49b2, 0x48a3, 
+0x463a, 0x47b0, 0x44cb, 0x4762, 0x4765, 0x46b9, 0x466a, 0x4814, 0x4631, 0x4796, 0x4666, 0x474b, 0x4798, 0x4704, 0x4838, 0x4761, 0x47d3, 0x4590, 0x45ea, 0x48a2, 0x47f1, 0x4844, 0x484b, 0x4776, 0x47d6, 0x46d8, 0x48f3, 0x44d3, 0x46fa, 0x478d, 0x481e, 0x466e, 
+0x4827, 0x481e, 0x45a2, 0x4794, 0x4727, 0x4806, 0x475d, 0x48d5, 0x4708, 0x4828, 0x4862, 0x480d, 0x4895, 0x4832, 0x48bd, 0x47f1, 0x482a, 0x46a7, 0x47b1, 0x492d, 0x484d, 0x4884, 0x48dc, 0x485f, 0x476d, 0x480c, 0x48e9, 0x46d3, 0x4728, 0x4884, 0x48a0, 0x480e, 
+0x4862, 0x4813, 0x4675, 0x485a, 0x47e8, 0x4738, 0x4836, 0x4823, 0x46e7, 0x4821, 0x4822, 0x47b3, 0x4846, 0x4855, 0x4863, 0x4717, 0x4872, 0x47c1, 0x46d5, 0x488e, 0x47e2, 0x485f, 0x487c, 0x48b8, 0x481e, 0x4788, 0x48bd, 0x4677, 0x46c9, 0x47f8, 0x48fe, 0x47fc, 
+0x47a0, 0x47b2, 0x4588, 0x467e, 0x4662, 0x46c7, 0x46e8, 0x4812, 0x4536, 0x474e, 0x46c0, 0x468f, 0x481f, 0x4679, 0x46a1, 0x46e2, 0x4809, 0x4560, 0x4630, 0x47eb, 0x46b5, 0x4757, 0x4848, 0x477f, 0x46a6, 0x46d8, 0x4870, 0x459a, 0x4670, 0x4678, 0x47d2, 0x468c, 
+0x4762, 0x48c4, 0x46e3, 0x4791, 0x46b1, 0x486d, 0x47d0, 0x4867, 0x468d, 0x47f6, 0x48a5, 0x4756, 0x4857, 0x4854, 0x4866, 0x4838, 0x484d, 0x46ec, 0x47d2, 0x48f6, 0x484a, 0x4879, 0x4848, 0x483c, 0x471d, 0x4806, 0x48fa, 0x4730, 0x4768, 0x47b8, 0x4865, 0x46f9, 
+0x48a8, 0x4918, 0x46ca, 0x4867, 0x4800, 0x4862, 0x48d3, 0x4910, 0x474e, 0x4849, 0x48eb, 0x486b, 0x4966, 0x48c5, 0x48f4, 0x4830, 0x48f9, 0x4778, 0x481e, 0x499e, 0x48cf, 0x48f1, 0x4982, 0x4923, 0x487c, 0x47cf, 0x49ea, 0x4649, 0x4773, 0x495e, 0x48b2, 0x483f, 
+0x48a7, 0x4975, 0x4616, 0x481e, 0x481f, 0x4866, 0x48b6, 0x4864, 0x47dc, 0x4873, 0x485c, 0x487f, 0x4938, 0x491f, 0x490d, 0x48b6, 0x48f8, 0x48a1, 0x4859, 0x492d, 0x489c, 0x4915, 0x4899, 0x4887, 0x486c, 0x4859, 0x49ca, 0x471e, 0x4867, 0x4918, 0x48d3, 0x4827, 
+0x488b, 0x4998, 0x4704, 0x481d, 0x48b8, 0x4880, 0x4876, 0x4944, 0x470c, 0x48f2, 0x48b9, 0x489b, 0x4956, 0x48e5, 0x48d6, 0x48a5, 0x48dc, 0x4856, 0x484e, 0x49ab, 0x48e0, 0x490e, 0x48dd, 0x4945, 0x488b, 0x48dd, 0x4a32, 0x47ea, 0x4835, 0x4911, 0x4965, 0x4819, 
+0x460e, 0x481e, 0x452c, 0x4673, 0x475c, 0x4717, 0x46f6, 0x46d0, 0x4696, 0x46bc, 0x4726, 0x481e, 0x4763, 0x46ea, 0x46fe, 0x4758, 0x478b, 0x4627, 0x4704, 0x483f, 0x46ad, 0x47b1, 0x4792, 0x4816, 0x46f2, 0x4684, 0x4827, 0x45a8, 0x472f, 0x47a4, 0x4797, 0x462b, 
+0x483f, 0x48ab, 0x468f, 0x4863, 0x485a, 0x4766, 0x481d, 0x48cb, 0x47dc, 0x4903, 0x48fc, 0x4830, 0x48cc, 0x483e, 0x48ab, 0x4864, 0x4966, 0x4763, 0x4794, 0x499d, 0x488e, 0x488b, 0x48dc, 0x4960, 0x4854, 0x484c, 0x499c, 0x474c, 0x4826, 0x48bc, 0x4949, 0x4883, 
+0x489d, 0x4905, 0x4718, 0x481e, 0x48e3, 0x48f4, 0x48c1, 0x4904, 0x47e8, 0x48b3, 0x4892, 0x48d4, 0x48ff, 0x4894, 0x48d5, 0x4886, 0x48fa, 0x4803, 0x47d2, 0x492e, 0x4870, 0x48b2, 0x48e5, 0x492b, 0x487b, 0x4785, 0x49e3, 0x471d, 0x4837, 0x48bf, 0x489b, 0x48c4, 
+0x475c, 0x4871, 0x464a, 0x4811, 0x47af, 0x471c, 0x4817, 0x4817, 0x463b, 0x484e, 0x477f, 0x464f, 0x4704, 0x487c, 0x47a3, 0x4725, 0x4853, 0x462a, 0x465a, 0x4860, 0x4736, 0x4880, 0x47e1, 0x482b, 0x4811, 0x46c0, 0x48dc, 0x475d, 0x4668, 0x4806, 0x4893, 0x46f4, 
+0x4858, 0x4959, 0x463d, 0x487b, 0x480f, 0x484e, 0x48c0, 0x48a6, 0x4847, 0x4894, 0x48a0, 0x484a, 0x491e, 0x48f4, 0x48fc, 0x48b5, 0x48ce, 0x47d2, 0x47db, 0x497f, 0x4955, 0x4939, 0x48a7, 0x48ce, 0x4890, 0x4884, 0x49d6, 0x4763, 0x486e, 0x4922, 0x48f4, 0x48c3, 
+0x47ec, 0x491c, 0x4698, 0x4783, 0x4715, 0x4754, 0x4745, 0x4752, 0x472f, 0x4832, 0x4817, 0x4809, 0x47f8, 0x48c3, 0x47e6, 0x4800, 0x48b6, 0x4730, 0x480a, 0x48cb, 0x479e, 0x488e, 0x47c2, 0x488e, 0x472f, 0x47ee, 0x489d, 0x4744, 0x4755, 0x4851, 0x4846, 0x47d3, 
+0x4838, 0x48a0, 0x4634, 0x4762, 0x4786, 0x4806, 0x47e3, 0x482d, 0x4726, 0x486c, 0x47b7, 0x4803, 0x48ac, 0x4814, 0x48e0, 0x4839, 0x4827, 0x4750, 0x46f2, 0x48c5, 0x483f, 0x4886, 0x48ad, 0x4856, 0x47e8, 0x47a9, 0x4937, 0x4743, 0x46d0, 0x481f, 0x484c, 0x4804, 
+0x47fd, 0x481f, 0x456d, 0x4813, 0x474d, 0x4807, 0x4688, 0x480e, 0x46e8, 0x4810, 0x469f, 0x4799, 0x4853, 0x478f, 0x47f2, 0x4824, 0x47d0, 0x471f, 0x46da, 0x485f, 0x4813, 0x481c, 0x482e, 0x4863, 0x4786, 0x480b, 0x48c9, 0x46b8, 0x475a, 0x46e2, 0x4852, 0x46c5, 
+0x45af, 0x4802, 0x4466, 0x46c2, 0x465d, 0x4743, 0x46b7, 0x47ba, 0x4636, 0x46c3, 0x4677, 0x4784, 0x485a, 0x47c2, 0x46dc, 0x46ac, 0x47de, 0x460e, 0x465f, 0x4834, 0x47f4, 0x4769, 0x46fc, 0x4810, 0x45fd, 0x45ea, 0x48d0, 0x45b5, 0x4704, 0x4783, 0x4830, 0x46c4, 
+0x4759, 0x47c7, 0x453d, 0x45b0, 0x4741, 0x4702, 0x4736, 0x4793, 0x461b, 0x47ba, 0x470b, 0x46dd, 0x4657, 0x470b, 0x470d, 0x4710, 0x486c, 0x468f, 0x45c3, 0x46ba, 0x479d, 0x483b, 0x46c9, 0x4774, 0x46a9, 0x46a7, 0x4833, 0x4606, 0x4690, 0x46a9, 0x46f5, 0x46a7, 
+0x47ac, 0x48bb, 0x452c, 0x4803, 0x470f, 0x4824, 0x47d5, 0x48cb, 0x4707, 0x484a, 0x4832, 0x4797, 0x4851, 0x482c, 0x487a, 0x4877, 0x4891, 0x465d, 0x47f4, 0x48ce, 0x4898, 0x4899, 0x484e, 0x486a, 0x47ac, 0x47f0, 0x493e, 0x4611, 0x47e2, 0x489e, 0x488c, 0x46af, 
+0x4665, 0x4836, 0x45e4, 0x46b6, 0x46a1, 0x46b9, 0x46c8, 0x46dd, 0x4658, 0x474b, 0x467b, 0x4777, 0x4769, 0x4798, 0x4785, 0x475e, 0x472a, 0x4656, 0x45fb, 0x4881, 0x46fc, 0x472d, 0x476e, 0x47a3, 0x465d, 0x46ca, 0x4855, 0x4500, 0x464f, 0x479a, 0x46c3, 0x4738, 
+0x481e, 0x486c, 0x4659, 0x4801, 0x4756, 0x477a, 0x47d5, 0x487b, 0x4706, 0x4808, 0x484f, 0x4838, 0x4870, 0x4863, 0x48d3, 0x4806, 0x4865, 0x4771, 0x46be, 0x494c, 0x4915, 0x484c, 0x4900, 0x4862, 0x481a, 0x46e8, 0x4974, 0x46a0, 0x4775, 0x483d, 0x487c, 0x480e 
+};
\ No newline at end of file
diff --git a/hwpe/redmule_softclear/inc/z_output.h b/hwpe/redmule_softclear/inc/z_output.h
new file mode 100644
index 0000000..96c7e5f
--- /dev/null
+++ b/hwpe/redmule_softclear/inc/z_output.h
@@ -0,0 +1,27 @@
+ /* Header file generated by RedMulE Golden Model */
+uint16_t z_oup [768] = {
+0x4845, 0x4897, 0x4608, 0x4838, 0x4855, 0x487b, 0x4869, 0x4880, 0x46d1, 0x48b0, 0x48db, 0x483f, 0x48c9, 0x485f, 0x4881, 0x483a, 0x484b, 0x472c, 0x4762, 0x492b, 0x4822, 0x48fd, 0x488e, 0x492e, 0x483e, 0x484f, 0x49e8, 0x46d7, 0x484b, 0x489d, 0x490b, 0x47e9, 
+0x484f, 0x47d2, 0x44be, 0x4747, 0x47c7, 0x46c0, 0x4727, 0x48af, 0x46c5, 0x482d, 0x483d, 0x482e, 0x4897, 0x479f, 0x488b, 0x4749, 0x489a, 0x46a8, 0x46f2, 0x488b, 0x4891, 0x47e8, 0x4872, 0x483d, 0x4716, 0x46fd, 0x49b5, 0x46a0, 0x46e7, 0x47a4, 0x48a1, 0x4767, 
+0x4939, 0x4935, 0x4703, 0x48c1, 0x4863, 0x48bd, 0x4913, 0x48cf, 0x48b6, 0x48b8, 0x4946, 0x4920, 0x495e, 0x48e1, 0x4938, 0x48b2, 0x493a, 0x4882, 0x483b, 0x49d5, 0x4911, 0x4972, 0x496b, 0x49df, 0x48f2, 0x4888, 0x4a46, 0x4821, 0x48c1, 0x490c, 0x49b2, 0x48a3, 
+0x463a, 0x47b0, 0x44cb, 0x4762, 0x4765, 0x46b9, 0x466a, 0x4814, 0x4631, 0x4796, 0x4666, 0x474b, 0x4798, 0x4704, 0x4838, 0x4761, 0x47d3, 0x4590, 0x45ea, 0x48a2, 0x47f1, 0x4844, 0x484b, 0x4776, 0x47d6, 0x46d8, 0x48f3, 0x44d3, 0x46fa, 0x478d, 0x481e, 0x466e, 
+0x4827, 0x481e, 0x45a2, 0x4794, 0x4727, 0x4806, 0x475d, 0x48d5, 0x4708, 0x4828, 0x4862, 0x480d, 0x4895, 0x4832, 0x48bd, 0x47f1, 0x482a, 0x46a7, 0x47b1, 0x492d, 0x484d, 0x4884, 0x48dc, 0x485f, 0x476d, 0x480c, 0x48e9, 0x46d3, 0x4728, 0x4884, 0x48a0, 0x480e, 
+0x4862, 0x4813, 0x4675, 0x485a, 0x47e8, 0x4738, 0x4836, 0x4823, 0x46e7, 0x4821, 0x4822, 0x47b3, 0x4846, 0x4855, 0x4863, 0x4717, 0x4872, 0x47c1, 0x46d5, 0x488e, 0x47e2, 0x485f, 0x487c, 0x48b8, 0x481e, 0x4788, 0x48bd, 0x4677, 0x46c9, 0x47f8, 0x48fe, 0x47fc, 
+0x47a0, 0x47b2, 0x4588, 0x467e, 0x4662, 0x46c7, 0x46e8, 0x4812, 0x4536, 0x474e, 0x46c0, 0x468f, 0x481f, 0x4679, 0x46a1, 0x46e2, 0x4809, 0x4560, 0x4630, 0x47eb, 0x46b5, 0x4757, 0x4848, 0x477f, 0x46a6, 0x46d8, 0x4870, 0x459a, 0x4670, 0x4678, 0x47d2, 0x468c, 
+0x4762, 0x48c4, 0x46e3, 0x4791, 0x46b1, 0x486d, 0x47d0, 0x4867, 0x468d, 0x47f6, 0x48a5, 0x4756, 0x4857, 0x4854, 0x4866, 0x4838, 0x484d, 0x46ec, 0x47d2, 0x48f6, 0x484a, 0x4879, 0x4848, 0x483c, 0x471d, 0x4806, 0x48fa, 0x4730, 0x4768, 0x47b8, 0x4865, 0x46f9, 
+0x48a8, 0x4918, 0x46ca, 0x4867, 0x4800, 0x4862, 0x48d3, 0x4910, 0x474e, 0x4849, 0x48eb, 0x486b, 0x4966, 0x48c5, 0x48f4, 0x4830, 0x48f9, 0x4778, 0x481e, 0x499e, 0x48cf, 0x48f1, 0x4982, 0x4923, 0x487c, 0x47cf, 0x49ea, 0x4649, 0x4773, 0x495e, 0x48b2, 0x483f, 
+0x48a7, 0x4975, 0x4616, 0x481e, 0x481f, 0x4866, 0x48b6, 0x4864, 0x47dc, 0x4873, 0x485c, 0x487f, 0x4938, 0x491f, 0x490d, 0x48b6, 0x48f8, 0x48a1, 0x4859, 0x492d, 0x489c, 0x4915, 0x4899, 0x4887, 0x486c, 0x4859, 0x49ca, 0x471e, 0x4867, 0x4918, 0x48d3, 0x4827, 
+0x488b, 0x4998, 0x4704, 0x481d, 0x48b8, 0x4880, 0x4876, 0x4944, 0x470c, 0x48f2, 0x48b9, 0x489b, 0x4956, 0x48e5, 0x48d6, 0x48a5, 0x48dc, 0x4856, 0x484e, 0x49ab, 0x48e0, 0x490e, 0x48dd, 0x4945, 0x488b, 0x48dd, 0x4a32, 0x47ea, 0x4835, 0x4911, 0x4965, 0x4819, 
+0x460e, 0x481e, 0x452c, 0x4673, 0x475c, 0x4717, 0x46f6, 0x46d0, 0x4696, 0x46bc, 0x4726, 0x481e, 0x4763, 0x46ea, 0x46fe, 0x4758, 0x478b, 0x4627, 0x4704, 0x483f, 0x46ad, 0x47b1, 0x4792, 0x4816, 0x46f2, 0x4684, 0x4827, 0x45a8, 0x472f, 0x47a4, 0x4797, 0x462b, 
+0x483f, 0x48ab, 0x468f, 0x4863, 0x485a, 0x4766, 0x481d, 0x48cb, 0x47dc, 0x4903, 0x48fc, 0x4830, 0x48cc, 0x483e, 0x48ab, 0x4864, 0x4966, 0x4763, 0x4794, 0x499d, 0x488e, 0x488b, 0x48dc, 0x4960, 0x4854, 0x484c, 0x499c, 0x474c, 0x4826, 0x48bc, 0x4949, 0x4883, 
+0x489d, 0x4905, 0x4718, 0x481e, 0x48e3, 0x48f4, 0x48c1, 0x4904, 0x47e8, 0x48b3, 0x4892, 0x48d4, 0x48ff, 0x4894, 0x48d5, 0x4886, 0x48fa, 0x4803, 0x47d2, 0x492e, 0x4870, 0x48b2, 0x48e5, 0x492b, 0x487b, 0x4785, 0x49e3, 0x471d, 0x4837, 0x48bf, 0x489b, 0x48c4, 
+0x475c, 0x4871, 0x464a, 0x4811, 0x47af, 0x471c, 0x4817, 0x4817, 0x463b, 0x484e, 0x477f, 0x464f, 0x4704, 0x487c, 0x47a3, 0x4725, 0x4853, 0x462a, 0x465a, 0x4860, 0x4736, 0x4880, 0x47e1, 0x482b, 0x4811, 0x46c0, 0x48dc, 0x475d, 0x4668, 0x4806, 0x4893, 0x46f4, 
+0x4858, 0x4959, 0x463d, 0x487b, 0x480f, 0x484e, 0x48c0, 0x48a6, 0x4847, 0x4894, 0x48a0, 0x484a, 0x491e, 0x48f4, 0x48fc, 0x48b5, 0x48ce, 0x47d2, 0x47db, 0x497f, 0x4955, 0x4939, 0x48a7, 0x48ce, 0x4890, 0x4884, 0x49d6, 0x4763, 0x486e, 0x4922, 0x48f4, 0x48c3, 
+0x47ec, 0x491c, 0x4698, 0x4783, 0x4715, 0x4754, 0x4745, 0x4752, 0x472f, 0x4832, 0x4817, 0x4809, 0x47f8, 0x48c3, 0x47e6, 0x4800, 0x48b6, 0x4730, 0x480a, 0x48cb, 0x479e, 0x488e, 0x47c2, 0x488e, 0x472f, 0x47ee, 0x489d, 0x4744, 0x4755, 0x4851, 0x4846, 0x47d3, 
+0x4838, 0x48a0, 0x4634, 0x4762, 0x4786, 0x4806, 0x47e3, 0x482d, 0x4726, 0x486c, 0x47b7, 0x4803, 0x48ac, 0x4814, 0x48e0, 0x4839, 0x4827, 0x4750, 0x46f2, 0x48c5, 0x483f, 0x4886, 0x48ad, 0x4856, 0x47e8, 0x47a9, 0x4937, 0x4743, 0x46d0, 0x481f, 0x484c, 0x4804, 
+0x47fd, 0x481f, 0x456d, 0x4813, 0x474d, 0x4807, 0x4688, 0x480e, 0x46e8, 0x4810, 0x469f, 0x4799, 0x4853, 0x478f, 0x47f2, 0x4824, 0x47d0, 0x471f, 0x46da, 0x485f, 0x4813, 0x481c, 0x482e, 0x4863, 0x4786, 0x480b, 0x48c9, 0x46b8, 0x475a, 0x46e2, 0x4852, 0x46c5, 
+0x45af, 0x4802, 0x4466, 0x46c2, 0x465d, 0x4743, 0x46b7, 0x47ba, 0x4636, 0x46c3, 0x4677, 0x4784, 0x485a, 0x47c2, 0x46dc, 0x46ac, 0x47de, 0x460e, 0x465f, 0x4834, 0x47f4, 0x4769, 0x46fc, 0x4810, 0x45fd, 0x45ea, 0x48d0, 0x45b5, 0x4704, 0x4783, 0x4830, 0x46c4, 
+0x4759, 0x47c7, 0x453d, 0x45b0, 0x4741, 0x4702, 0x4736, 0x4793, 0x461b, 0x47ba, 0x470b, 0x46dd, 0x4657, 0x470b, 0x470d, 0x4710, 0x486c, 0x468f, 0x45c3, 0x46ba, 0x479d, 0x483b, 0x46c9, 0x4774, 0x46a9, 0x46a7, 0x4833, 0x4606, 0x4690, 0x46a9, 0x46f5, 0x46a7, 
+0x47ac, 0x48bb, 0x452c, 0x4803, 0x470f, 0x4824, 0x47d5, 0x48cb, 0x4707, 0x484a, 0x4832, 0x4797, 0x4851, 0x482c, 0x487a, 0x4877, 0x4891, 0x465d, 0x47f4, 0x48ce, 0x4898, 0x4899, 0x484e, 0x486a, 0x47ac, 0x47f0, 0x493e, 0x4611, 0x47e2, 0x489e, 0x488c, 0x46af, 
+0x4665, 0x4836, 0x45e4, 0x46b6, 0x46a1, 0x46b9, 0x46c8, 0x46dd, 0x4658, 0x474b, 0x467b, 0x4777, 0x4769, 0x4798, 0x4785, 0x475e, 0x472a, 0x4656, 0x45fb, 0x4881, 0x46fc, 0x472d, 0x476e, 0x47a3, 0x465d, 0x46ca, 0x4855, 0x4500, 0x464f, 0x479a, 0x46c3, 0x4738, 
+0x481e, 0x486c, 0x4659, 0x4801, 0x4756, 0x477a, 0x47d5, 0x487b, 0x4706, 0x4808, 0x484f, 0x4838, 0x4870, 0x4863, 0x48d3, 0x4806, 0x4865, 0x4771, 0x46be, 0x494c, 0x4915, 0x484c, 0x4900, 0x4862, 0x481a, 0x46e8, 0x4974, 0x46a0, 0x4775, 0x483d, 0x487c, 0x480e 
+};
\ No newline at end of file
diff --git a/hwpe/redmule_softclear/pulp_inject_fault.tcl b/hwpe/redmule_softclear/pulp_inject_fault.tcl
new file mode 100644
index 0000000..61ccadf
--- /dev/null
+++ b/hwpe/redmule_softclear/pulp_inject_fault.tcl
@@ -0,0 +1,53 @@
+# Copyright 2023 ETH Zurich and University of Bologna.
+# Solderpad Hardware License, Version 0.51, see LICENSE for details.
+# SPDX-License-Identifier: SHL-0.51
+#
+# Author: Michael Rogenmoser (michaero@iis.ee.ethz.ch)
+
+transcript quietly
+if {! [info exists ::env(VSIM_PATH)]} {error "Define VSIM_PATH"}
+set utils_base_path  [file join $::env(VSIM_PATH) scripts fault_injection_utils]
+set script_base_path [file join $::env(VSIM_PATH) fault_injection_sim scripts]
+
+set verbosity            2
+set log_injections       1
+# Easy way to generate a variable seed
+# set seed                 [clock seconds]
+# Default value
+set seed                 12345
+set print_statistics     1
+
+set inject_start_time 550000000000ps
+set inject_stop_time 750000000000ps
+set injection_clock "pulp_cluster_tb/cluster_i/clk_i"
+set injection_clock_trigger 0
+set fault_period 150
+set rand_initial_injection_phase 0
+# max_num set to 0 means until stop_time
+set max_num_fault_inject 0
+set signal_fault_duration 20ns
+set register_fault_duration 0ns
+
+set allow_multi_bit_upset $::env(MULTI_BIT_UPSET)
+set use_bitwidth_as_weight 0
+set check_core_output_modification 0
+set check_core_next_state_modification 0
+set reg_to_sig_ratio 1
+
+source [file join $utils_base_path pulp_extract_nets.tcl]
+
+set inject_signals_netlist []
+set inject_register_netlist []
+set output_netlist []
+set next_state_netlist []
+set assertion_disable_list []
+
+# for {set idx 0} {$idx < 12} {incr idx} {
+#     set inject_signals_netlist [list {*}$inject_signals_netlist {*}[get_all_core_nets $idx]]
+#     set output_netlist [list {*}$output_netlist {*}[get_core_output_nets $idx]]
+# }
+
+set inject_register_netlist [list {*}$inject_register_netlist {*}[get_memory_slice {0 16} {256 336}]]
+
+source [file join $script_base_path inject_fault.tcl]
+
diff --git a/hwpe/redmule_softclear/redmule.c b/hwpe/redmule_softclear/redmule.c
new file mode 100644
index 0000000..a80171c
--- /dev/null
+++ b/hwpe/redmule_softclear/redmule.c
@@ -0,0 +1,185 @@
+/*
+ * Copyright (C) 2022-2023 ETH Zurich and University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ * SPDX-License-Identifier: Apache-2.0
+ * 
+ * Author: Yvan Tortorella  <yvan.tortorella@unibo.it>
+ *
+ * RedMulE SW test
+ */
+
+#include <stdint.h>
+#include "stdio.h"
+#include "archi_redmule.h"
+#include "hal_redmule.h"
+#include "pulp.h"
+
+static inline void wait_cycles(const unsigned cycles)
+{
+  /**
+   * Each iteration of the loop below will take four cycles on RI5CY (one for
+   * `addi` and three for the taken `bnez`; if the instructions hit in the
+   * I$).  Thus, we let `i` count the number of remaining loop iterations and
+   * initialize it to a fourth of the number of clock cyles.  With this
+   * initialization, we must not enter the loop if the number of clock cycles
+   * is less than four, because this will cause an underflow on the first
+   * subtraction.
+   */
+  register unsigned threshold;
+  asm volatile("li %[threshold], 4" : [threshold] "=r" (threshold));
+  asm volatile goto("ble %[cycles], %[threshold], %l2"
+          : /* no output */
+          : [cycles] "r" (cycles), [threshold] "r" (threshold)
+          : /* no clobbers */
+          : __wait_cycles_end);
+  register unsigned i = cycles >> 2;
+__wait_cycles_start:
+  // Decrement `i` and loop if it is not yet zero.
+  asm volatile("addi %0, %0, -1" : "+r" (i));
+  asm volatile goto("bnez %0, %l1"
+          : /* no output */
+          : "r" (i)
+          : /* no clobbers */
+          : __wait_cycles_start);
+__wait_cycles_end:
+  return;
+}
+
+int main() {
+
+  volatile int errors = 0;
+  unsigned int cluster_id = rt_cluster_id();
+  unsigned int intc_data_correctable_cnt, redmule_data_correctable_cnt = 0;
+  unsigned int intc_meta_correctable_cnt = 0;
+  unsigned int intc_data_uncorrectable_cnt, redmule_data_uncorrectable_cnt = 0;
+  unsigned int intc_meta_uncorrectable_cnt = 0;
+
+  if(get_core_id() == 0){
+
+    uint16_t m_size = M_SIZE;
+    uint16_t n_size = N_SIZE;
+    uint16_t k_size = K_SIZE;
+
+    uint8_t *x_ext = x_inp;
+    uint8_t *w_ext = w_inp;
+    uint8_t *y_ext = y_inp;
+    uint8_t *z_ext = z_oup;
+
+    uint8_t volatile *x = (uint8_t volatile *) pi_l1_malloc(0, (2*m_size*n_size));
+    uint8_t volatile *w = (uint8_t volatile *) pi_l1_malloc(0, (2*n_size*k_size));
+    uint8_t volatile *y = (uint8_t volatile *) pi_l1_malloc(0, (2*m_size*k_size));
+    uint8_t volatile *z = (uint8_t volatile *) pi_l1_malloc(0, (2*m_size*k_size));
+
+    #ifdef USE_DMA
+      volatile unsigned int dma_id = 0;
+      dma_id = mchan_alloc();
+      mchan_transfer((unsigned int) 2*(2*m_size*n_size),
+                     (unsigned int) x_ext,
+                     (unsigned int) x    );
+      mchan_barrier(dma_id);
+      mchan_free(dma_id);
+    
+      dma_id = mchan_alloc();
+      mchan_transfer((unsigned int) 2*(2*n_size*k_size),
+                     (unsigned int) w_ext,
+                     (unsigned int) w    );
+      mchan_barrier(dma_id);
+      mchan_free(dma_id);
+    
+      dma_id = mchan_alloc();
+      mchan_transfer((unsigned int) 2*(2*m_size*k_size),
+                     (unsigned int) y_ext,
+                     (unsigned int) y    );
+      mchan_barrier(dma_id);
+    #else
+      generate_test_data16((int) x, (int) w, (int) y, (int) m_size, (int) n_size, (int) k_size);
+    #endif
+
+    int gold_sum = 0, check_sum = 0;
+    int i,j;
+
+    int offload_id_tmp, offload_id;
+
+    // Enable RedMulE
+    hwpe_cg_enable();
+    asm volatile("": : :"memory");
+
+    hwpe_soft_clear();
+    asm volatile("": : :"memory");
+
+    volatile int job_id = -1;
+
+    // job 0
+    job_id = hwpe_wait_acquire();
+    asm volatile("": : :"memory");
+    redmule_x_add_set ((unsigned int) x);
+    redmule_w_add_set ((unsigned int) w);
+    redmule_y_add_set ((unsigned int) y);
+    redmule_z_add_set ((unsigned int) z);
+    redmule_cfg (m_size, n_size, k_size, gemm_ops);
+    asm volatile("": : :"memory");
+    hwpe_trigger_job();
+    asm volatile("": : :"memory");
+
+    // job 1
+    job_id = hwpe_wait_acquire();
+    asm volatile("": : :"memory");
+    redmule_x_add_set ((unsigned int) x);
+    redmule_w_add_set ((unsigned int) w);
+    redmule_y_add_set ((unsigned int) y);
+    redmule_z_add_set ((unsigned int) z);
+    redmule_cfg (m_size, n_size, k_size, gemm_ops);
+    asm volatile("": : :"memory");
+    hwpe_trigger_job();
+    asm volatile("": : :"memory");
+
+    // soft-clear execution
+    wait_cycles(20);
+    hwpe_soft_clear();
+    wait_cycles(100);
+
+    // job 0
+    job_id = hwpe_wait_acquire();
+    asm volatile("": : :"memory");
+    redmule_x_add_set ((unsigned int) x);
+    redmule_w_add_set ((unsigned int) w);
+    redmule_y_add_set ((unsigned int) y);
+    redmule_z_add_set ((unsigned int) z);
+    redmule_cfg (m_size, n_size, k_size, gemm_ops);
+    asm volatile("": : :"memory");
+    hwpe_trigger_job();
+    asm volatile("": : :"memory");
+
+    // Wait for end of computation
+    redmule_evt_wait();
+
+    // Disable RedMulE
+    hwpe_cg_disable();
+
+    errors = redmule_compare16((int) z, (int) m_size, (int) k_size);
+
+    *(int *) 0x1A1040A0 = errors;
+
+    if(job_id != 0) {
+      printf ("Terminated test with wrong job id!!! and %d errors. See you!\n", errors);
+      errors++;
+    }
+    else {
+      printf ("Terminated test with OK job id and %d errors. See you!\n", errors);
+    }
+
+  }
+  synch_barrier();
+  return (errors != 0);
+}
diff --git a/hwpe/softex/Makefile b/hwpe/softex/Makefile
new file mode 100644
index 0000000..a15e398
--- /dev/null
+++ b/hwpe/softex/Makefile
@@ -0,0 +1,9 @@
+PULP_APP = test
+PULP_APP_SRCS = softex.c
+PULP_CFLAGS = -O3
+
+ifeq ($(use_dma),1)
+ PULP_CFLAGS += -DUSE_DMA
+endif
+
+include $(PULP_SDK_HOME)/install/rules/pulp_rt.mk
diff --git a/hwpe/softex/archi_softex.h b/hwpe/softex/archi_softex.h
new file mode 100644
index 0000000..8d08543
--- /dev/null
+++ b/hwpe/softex/archi_softex.h
@@ -0,0 +1,81 @@
+/*
+ * Andrea Belano <andrea.belano@studio.unibo.it>
+ *
+ * Copyright 2024 ETH Zurich and University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ *
+ * SPDX-License-Identifier: Apache-2.0
+ */
+
+#ifndef __ARCHI_SOFTEX__
+#define __ARCHI_SOFTEX__
+
+#define DATA_WIDTH      256
+
+/* PULP Cluster Archi defines */
+#define ARCHI_CLUST_CTRL_BASE ARCHI_CLUSTER_CTRL_ADDR
+#define ARCHI_CLUST_HWPE_BASE ARCHI_HWCE_ADDR
+#define DMA_COMMAND_QUEUE     ARCHI_MCHAN_DEMUX_ADDR
+#define DMA_STATUS_REGISTER   (ARCHI_MCHAN_DEMUX_ADDR + 4)
+#define ARCHI_CL_HWPE_EVT0 12
+#define ARCHI_CL_HWPE_EVT1 13
+#define FC_DMA_EVENT 8
+#define CL_DMA_EVENT 22
+#define CLUST_CTRL_HWPE_EN 0x18
+#define CLUST_CTRL_HWPE_EN_MASK 0x800
+#define __builtin_bitinsert(a,b,c,d) (a | (((b << (32-c)) >> (32-c)) << d))
+
+#define SOFTEX_BASE_ADD    ARCHI_CLUST_HWPE_BASE
+#define SOFTEX_CG_EN_MSK   0x4000
+// Commands
+#define SOFTEX_TRIGGER     0x00
+#define SOFTEX_ACQUIRE     0x04
+#define SOFTEX_FINISHED    0x08
+#define SOFTEX_STATUS      0x0C
+#define SOFTEX_RUNNING_JOB 0x10
+#define SOFTEX_SOFT_CLEAR  0x14
+
+#define SOFTEX_REG_OFFS    0x20
+
+#define SOFTEX_IN_ADDR         SOFTEX_REG_OFFS + 0x00
+#define SOFTEX_OUT_ADDR        SOFTEX_REG_OFFS + 0x04
+#define SOFTEX_TOT_LEN         SOFTEX_REG_OFFS + 0x08
+#define SOFTEX_COMMANDS        SOFTEX_REG_OFFS + 0x0C
+#define SOFTEX_CACHE_BASE_ADDR SOFTEX_REG_OFFS + 0x10
+#define SOFTEX_CAST_CTRL       SOFTEX_REG_OFFS + 0x14
+
+
+#define SOFTEX_CMD_ACC_ONLY        0x00000001
+#define SOFTEX_CMD_DIV_ONLY        0x00000002
+#define SOFTEX_CMD_ACQUIRE_SLOT    0x00000004
+#define SOFTEX_CMD_LAST            0x00000008
+#define SOFTEX_CMD_SET_CACHE_ADDR  0x00000010
+#define SOFTEX_CMD_NO_OP           0x00000020
+#define SOFTEX_CMD_INT_INPUT       0x00000040
+#define SOFTEX_CMD_INT_OUTPUT      0x00000080
+
+/* DMA Archi */
+#define DMA_TX  0
+#define DMA_RX  1
+#define DMA_INC 1
+
+#define PLP_DMA_TYPE_BIT    0x00000011
+#define PLP_DMA_INCR_BIT    0x00000012
+#define PLP_DMA_2D_BIT      0x00000013
+#define PLP_DMA_ELE_BIT     0x00000014
+#define PLP_DMA_ILE_BIT     0x00000015
+#define PLP_DMA_BLE_BIT     0x00000016
+#define PLP_DMA_2D_TCDM_BIT 0x0000017
+
+#endif
\ No newline at end of file
diff --git a/hwpe/softex/hal_softex.h b/hwpe/softex/hal_softex.h
new file mode 100644
index 0000000..199a972
--- /dev/null
+++ b/hwpe/softex/hal_softex.h
@@ -0,0 +1,91 @@
+/*
+ * Andrea Belano <andrea.belano@studio.unibo.it>
+ *
+ * Copyright 2024 ETH Zurich and University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ *
+ * SPDX-License-Identifier: Apache-2.0
+ */
+
+#ifndef __HAL_SOFTEX__
+#define __HAL_SOFTEX__
+
+#include <pulp.h>
+#include "archi_softex.h"
+
+#define HWPE_WRITE(value, offset) *(volatile int *)(SOFTEX_BASE_ADD + offset) = value
+#define HWPE_READ(offset) *(volatile int *)(SOFTEX_BASE_ADD + offset)
+
+static inline void hwpe_trigger_job() {
+    HWPE_WRITE(0, SOFTEX_TRIGGER);
+}
+
+static inline int hwpe_acquire_job() {
+    return HWPE_READ(SOFTEX_ACQUIRE);
+}
+
+static inline unsigned int hwpe_get_status() {
+    return HWPE_READ(SOFTEX_STATUS);
+}
+
+static inline void hwpe_soft_clear() {
+  HWPE_WRITE(0, SOFTEX_SOFT_CLEAR);
+}
+
+static inline void hwpe_cg_enable() {
+  *(volatile int*) (ARCHI_CLUST_CTRL_BASE + CLUST_CTRL_HWPE_EN) |= (CLUST_CTRL_HWPE_EN_MASK | SOFTEX_CG_EN_MSK);
+}
+
+static inline void hwpe_cg_disable() {
+  *(volatile int*) (ARCHI_CLUST_CTRL_BASE + CLUST_CTRL_HWPE_EN) &= ~(CLUST_CTRL_HWPE_EN_MASK | SOFTEX_CG_EN_MSK);
+}
+
+static inline void softex_evt_wait() {
+  do {
+    eu_evt_maskWaitAndClr (1 << ARCHI_CL_HWPE_EVT0);
+  } while((*(int volatile *)(ARCHI_CLUST_HWPE_BASE + SOFTEX_STATUS)) != 0);
+}
+
+/* DMA APIs */
+static inline int mchan_alloc(){
+  return *(volatile int*) DMA_COMMAND_QUEUE;
+}
+
+static inline void mchan_transfer(unsigned int len,
+                                  unsigned int ext_addr,
+                                  unsigned int tcdm_addr) {
+
+  *(volatile int*) DMA_COMMAND_QUEUE = len |
+                                       (DMA_RX << PLP_DMA_TYPE_BIT)  |
+                                       (DMA_INC << PLP_DMA_INCR_BIT) |
+                                       (0 << PLP_DMA_2D_BIT)         |
+                                       (1 << PLP_DMA_ELE_BIT)        |
+                                       (1 << PLP_DMA_ILE_BIT)        |
+                                       (0 << PLP_DMA_BLE_BIT)        |
+                                       (0 << PLP_DMA_2D_TCDM_BIT);
+  *(volatile int*) DMA_COMMAND_QUEUE = tcdm_addr;
+  *(volatile int*) DMA_COMMAND_QUEUE = ext_addr;
+}
+
+static inline void mchan_barrier(int id) {
+  while(((*(volatile int*)(DMA_STATUS_REGISTER)) >> id ) & 0x1 ) {
+    eu_evt_maskWaitAndClr(1 << FC_DMA_EVENT);
+ }
+}
+
+static inline void mchan_free(int id) {
+  *(volatile int*) DMA_STATUS_REGISTER = 0x1 << id;
+}
+
+#endif
\ No newline at end of file
diff --git a/hwpe/softex/inc/golden.h b/hwpe/softex/inc/golden.h
new file mode 100644
index 0000000..12f0012
--- /dev/null
+++ b/hwpe/softex/inc/golden.h
@@ -0,0 +1,1031 @@
+#ifndef __SOFTEX_GOLDEN__
+#define __SOFTEX_GOLDEN__
+
+#define GOLDEN {    \
+   0x31c8,    \
+   0x323b,    \
+   0x2cba,    \
+   0x2bf2,    \
+   0x3c72,    \
+   0x2c91,    \
+   0x39a1,    \
+   0x308b,    \
+   0x3c56,    \
+   0x395d,    \
+   0x308b,    \
+   0x2f8c,    \
+   0x2d45,    \
+   0x3693,    \
+   0x2613,    \
+   0x3a16,    \
+   0x32fe,    \
+   0x326f,    \
+   0x3310,    \
+   0x2b82,    \
+   0x2fb4,    \
+   0x2c91,    \
+   0x2f5b,    \
+   0x28dd,    \
+   0x3730,    \
+   0x3b28,    \
+   0x3b58,    \
+   0x364a,    \
+   0x2c1b,    \
+   0x3890,    \
+   0x2c80,    \
+   0x3c89,    \
+   0x38d1,    \
+   0x2cef,    \
+   0x2f2a,    \
+   0x29db,    \
+   0x2cba,    \
+   0x294e,    \
+   0x3906,    \
+   0x38b8,    \
+   0x39cf,    \
+   0x3a8c,    \
+   0x38d1,    \
+   0x3028,    \
+   0x3a78,    \
+   0x2e7b,    \
+   0x2d90,    \
+   0x2f5b,    \
+   0x31c8,    \
+   0x2f9f,    \
+   0x29f8,    \
+   0x364a,    \
+   0x34db,    \
+   0x33fb,    \
+   0x3ae8,    \
+   0x381a,    \
+   0x3028,    \
+   0x35bf,    \
+   0x2689,    \
+   0x364a,    \
+   0x3211,    \
+   0x3791,    \
+   0x33c3,    \
+   0x3748,    \
+   0x3a5b,    \
+   0x39a1,    \
+   0x2672,    \
+   0x276f,    \
+   0x34db,    \
+   0x2656,    \
+   0x2cef,    \
+   0x3c13,    \
+   0x33ac,    \
+   0x3386,    \
+   0x39a1,    \
+   0x3058,    \
+   0x3253,    \
+   0x3b75,    \
+   0x3a9f,    \
+   0x2753,    \
+   0x360b,    \
+   0x309d,    \
+   0x3568,    \
+   0x3c89,    \
+   0x276f,    \
+   0x28ac,    \
+   0x35bf,    \
+   0x38a3,    \
+   0x344f,    \
+   0x2ea1,    \
+   0x3003,    \
+   0x3c3d,    \
+   0x3a16,    \
+   0x2626,    \
+   0x2d2e,    \
+   0x3436,    \
+   0x37f0,    \
+   0x31c8,    \
+   0x36bd,    \
+   0x2db8,    \
+   0x3386,    \
+   0x3a41,    \
+   0x33dd,    \
+   0x3200,    \
+   0x2823,    \
+   0x3c02,    \
+   0x2dd1,    \
+   0x3225,    \
+   0x2898,    \
+   0x340e,    \
+   0x387e,    \
+   0x3808,    \
+   0x30b2,    \
+   0x381a,    \
+   0x371c,    \
+   0x2689,    \
+   0x27ae,    \
+   0x3be5,    \
+   0x2b32,    \
+   0x33ac,    \
+   0x3c56,    \
+   0x2656,    \
+   0x2c48,    \
+   0x2cd3,    \
+   0x2672,    \
+   0x3a5b,    \
+   0x3583,    \
+   0x3943,    \
+   0x2bd6,    \
+   0x2a67,    \
+   0x387e,    \
+   0x3c3d,    \
+   0x3709,    \
+   0x3ab4,    \
+   0x3534,    \
+   0x3c13,    \
+   0x29aa,    \
+   0x2a95,    \
+   0x269b,    \
+   0x346a,    \
+   0x3211,    \
+   0x3943,    \
+   0x2921,    \
+   0x3b9d,    \
+   0x2cd3,    \
+   0x3a41,    \
+   0x3890,    \
+   0x323b,    \
+   0x2cef,    \
+   0x2851,    \
+   0x3225,    \
+   0x2b4a,    \
+   0x2bbd,    \
+   0x2a67,    \
+   0x3351,    \
+   0x3351,    \
+   0x2823,    \
+   0x3028,    \
+   0x382e,    \
+   0x3808,    \
+   0x2cba,    \
+   0x3665,    \
+   0x2d2e,    \
+   0x2ca5,    \
+   0x2db8,    \
+   0x3cb0,    \
+   0x2fb4,    \
+   0x2838,    \
+   0x36a7,    \
+   0x3253,    \
+   0x2810,    \
+   0x2753,    \
+   0x395d,    \
+   0x2f41,    \
+   0x36bd,    \
+   0x279a,    \
+   0x3126,    \
+   0x3172,    \
+   0x336d,    \
+   0x2f05,    \
+   0x2f9f,    \
+   0x35a8,    \
+   0x392c,    \
+   0x3015,    \
+   0x3780,    \
+   0x303f,    \
+   0x2e8e,    \
+   0x36a7,    \
+   0x39cf,    \
+   0x34db,    \
+   0x2f41,    \
+   0x39a1,    \
+   0x3ce2,    \
+   0x28fa,    \
+   0x2a4c,    \
+   0x38ed,    \
+   0x2936,    \
+   0x3ab4,    \
+   0x3102,    \
+   0x30ca,    \
+   0x3be5,    \
+   0x31b0,    \
+   0x2dd1,    \
+   0x2e18,    \
+   0x36a7,    \
+   0x361d,    \
+   0x27fd,    \
+   0x2eb6,    \
+   0x3a16,    \
+   0x2eb6,    \
+   0x2a67,    \
+   0x37d3,    \
+   0x3436,    \
+   0x33c3,    \
+   0x336d,    \
+   0x27c5,    \
+   0x2f5b,    \
+   0x360b,    \
+   0x2cd3,    \
+   0x387e,    \
+   0x3568,    \
+   0x3890,    \
+   0x3b9d,    \
+   0x3be5,    \
+   0x2788,    \
+   0x27fd,    \
+   0x2886,    \
+   0x3126,    \
+   0x2eea,    \
+   0x2753,    \
+   0x2f8c,    \
+   0x34db,    \
+   0x2d2e,    \
+   0x3075,    \
+   0x360b,    \
+   0x2a1f,    \
+   0x3595,    \
+   0x3682,    \
+   0x2eb6,    \
+   0x29db,    \
+   0x395d,    \
+   0x2f2a,    \
+   0x2bbd,    \
+   0x3906,    \
+   0x263c,    \
+   0x2e06,    \
+   0x3acc,    \
+   0x2700,    \
+   0x28dd,    \
+   0x2f16,    \
+   0x2a67,    \
+   0x309d,    \
+   0x2b1d,    \
+   0x2aa8,    \
+   0x3709,    \
+   0x28fa,    \
+   0x3436,    \
+   0x3534,    \
+   0x3b03,    \
+   0x2fe7,    \
+   0x2aa8,    \
+   0x2823,    \
+   0x326f,    \
+   0x32ae,    \
+   0x3c72,    \
+   0x29db,    \
+   0x3253,    \
+   0x29db,    \
+   0x3225,    \
+   0x2c91,    \
+   0x2700,    \
+   0x2bf2,    \
+   0x3bca,    \
+   0x34f8,    \
+   0x29db,    \
+   0x3568,    \
+   0x27c5,    \
+   0x3730,    \
+   0x2672,    \
+   0x3730,    \
+   0x3a2a,    \
+   0x303f,    \
+   0x32c5,    \
+   0x33c3,    \
+   0x3693,    \
+   0x37f0,    \
+   0x33ac,    \
+   0x3906,    \
+   0x313d,    \
+   0x3780,    \
+   0x2cba,    \
+   0x2c30,    \
+   0x3c72,    \
+   0x2cd3,    \
+   0x2810,    \
+   0x263c,    \
+   0x382e,    \
+   0x28fa,    \
+   0x382e,    \
+   0x395d,    \
+   0x28ac,    \
+   0x38a3,    \
+   0x32e0,    \
+   0x2724,    \
+   0x31b0,    \
+   0x2613,    \
+   0x273a,    \
+   0x2cef,    \
+   0x29c1,    \
+   0x2b93,    \
+   0x2f78,    \
+   0x2e2c,    \
+   0x344f,    \
+   0x2b93,    \
+   0x2996,    \
+   0x2d60,    \
+   0x3ce2,    \
+   0x3be5,    \
+   0x2672,    \
+   0x28c3,    \
+   0x2ded,    \
+   0x34db,    \
+   0x2b4a,    \
+   0x38d1,    \
+   0x27ae,    \
+   0x38ed,    \
+   0x2c1b,    \
+   0x3172,    \
+   0x3288,    \
+   0x2810,    \
+   0x2bbd,    \
+   0x3b03,    \
+   0x326f,    \
+   0x3b15,    \
+   0x3b9d,    \
+   0x2a0c,    \
+   0x3ae8,    \
+   0x3846,    \
+   0x2e8e,    \
+   0x2ecf,    \
+   0x286d,    \
+   0x3b28,    \
+   0x2eea,    \
+   0x3156,    \
+   0x2c48,    \
+   0x3ab4,    \
+   0x35bf,    \
+   0x2c48,    \
+   0x32c5,    \
+   0x2d90,    \
+   0x3015,    \
+   0x3860,    \
+   0x3be5,    \
+   0x2ded,    \
+   0x3a2a,    \
+   0x3b03,    \
+   0x32ae,    \
+   0x346a,    \
+   0x3890,    \
+   0x34f8,    \
+   0x3a41,    \
+   0x3b58,    \
+   0x36bd,    \
+   0x3bb2,    \
+   0x2b4a,    \
+   0x39a1,    \
+   0x3693,    \
+   0x2a83,    \
+   0x2d1a,    \
+   0x3890,    \
+   0x3918,    \
+   0x3762,    \
+   0x3a2a,    \
+   0x3860,    \
+   0x2f9f,    \
+   0x3ce2,    \
+   0x3730,    \
+   0x34c1,    \
+   0x279a,    \
+   0x35a8,    \
+   0x2ba6,    \
+   0x392c,    \
+   0x33ac,    \
+   0x3943,    \
+   0x2da3,    \
+   0x2eea,    \
+   0x344f,    \
+   0x35d8,    \
+   0x2ea1,    \
+   0x3c13,    \
+   0x2a95,    \
+   0x2bf2,    \
+   0x2823,    \
+   0x36f2,    \
+   0x381a,    \
+   0x3a05,    \
+   0x3846,    \
+   0x308b,    \
+   0x3ae8,    \
+   0x3b8b,    \
+   0x354c,    \
+   0x3b9d,    \
+   0x2e2c,    \
+   0x273a,    \
+   0x3323,    \
+   0x28fa,    \
+   0x2b82,    \
+   0x34aa,    \
+   0x2cef,    \
+   0x26e2,    \
+   0x392c,    \
+   0x2cba,    \
+   0x2a67,    \
+   0x2b4a,    \
+   0x2a67,    \
+   0x3748,    \
+   0x371c,    \
+   0x3288,    \
+   0x32c5,    \
+   0x3c3d,    \
+   0x2e2c,    \
+   0x3791,    \
+   0x2e5d,    \
+   0x3075,    \
+   0x3a05,    \
+   0x2e8e,    \
+   0x32fe,    \
+   0x3c02,    \
+   0x27fd,    \
+   0x2cba,    \
+   0x3c89,    \
+   0x2838,    \
+   0x2700,    \
+   0x2e43,    \
+   0x3386,    \
+   0x37f0,    \
+   0x3421,    \
+   0x354c,    \
+   0x31e2,    \
+   0x2996,    \
+   0x30b2,    \
+   0x2f8c,    \
+   0x2a0c,    \
+   0x3c02,    \
+   0x32e0,    \
+   0x2a83,    \
+   0x2886,    \
+   0x3568,    \
+   0x326f,    \
+   0x2788,    \
+   0x398e,    \
+   0x371c,    \
+   0x2851,    \
+   0x269b,    \
+   0x2613,    \
+   0x2672,    \
+   0x364a,    \
+   0x2bf2,    \
+   0x387e,    \
+   0x2f78,    \
+   0x3398,    \
+   0x2851,    \
+   0x2838,    \
+   0x39b6,    \
+   0x2e18,    \
+   0x3c3d,    \
+   0x39b6,    \
+   0x3943,    \
+   0x28fa,    \
+   0x2a83,    \
+   0x3cb0,    \
+   0x2f2a,    \
+   0x2e7b,    \
+   0x2f78,    \
+   0x3bca,    \
+   0x37bb,    \
+   0x279a,    \
+   0x2e7b,    \
+   0x3a9f,    \
+   0x392c,    \
+   0x27c5,    \
+   0x3156,    \
+   0x37a5,    \
+   0x3c56,    \
+   0x3172,    \
+   0x2d2e,    \
+   0x3665,    \
+   0x2fb4,    \
+   0x2d08,    \
+   0x361d,    \
+   0x2abf,    \
+   0x36f2,    \
+   0x2cd3,    \
+   0x30e5,    \
+   0x2fb4,    \
+   0x2a1f,    \
+   0x3cc8,    \
+   0x38a3,    \
+   0x397b,    \
+   0x2f9f,    \
+   0x39b6,    \
+   0x3200,    \
+   0x3253,    \
+   0x3156,    \
+   0x2724,    \
+   0x296a,    \
+   0x3a5b,    \
+   0x2c30,    \
+   0x3398,    \
+   0x2f5b,    \
+   0x2cef,    \
+   0x39ea,    \
+   0x3906,    \
+   0x3b3f,    \
+   0x2724,    \
+   0x2753,    \
+   0x3200,    \
+   0x2a1f,    \
+   0x329a,    \
+   0x32ae,    \
+   0x2b32,    \
+   0x33c3,    \
+   0x3075,    \
+   0x3126,    \
+   0x37a5,    \
+   0x361d,    \
+   0x32fe,    \
+   0x3665,    \
+   0x263c,    \
+   0x2c09,    \
+   0x3709,    \
+   0x29db,    \
+   0x3682,    \
+   0x39cf,    \
+   0x2ecf,    \
+   0x3d00,    \
+   0x30ca,    \
+   0x3421,    \
+   0x3890,    \
+   0x2f2a,    \
+   0x3172,    \
+   0x263c,    \
+   0x2689,    \
+   0x39ea,    \
+   0x2cef,    \
+   0x26c8,    \
+   0x3015,    \
+   0x2936,    \
+   0x2689,    \
+   0x3943,    \
+   0x2db8,    \
+   0x37d3,    \
+   0x3534,    \
+   0x3682,    \
+   0x32fe,    \
+   0x2b32,    \
+   0x2f05,    \
+   0x3421,    \
+   0x2613,    \
+   0x2b4a,    \
+   0x3189,    \
+   0x3310,    \
+   0x2bf2,    \
+   0x2f41,    \
+   0x3780,    \
+   0x3386,    \
+   0x3102,    \
+   0x326f,    \
+   0x2cd3,    \
+   0x3c3d,    \
+   0x2eb6,    \
+   0x33c3,    \
+   0x2f5b,    \
+   0x2bbd,    \
+   0x3b03,    \
+   0x2cba,    \
+   0x3780,    \
+   0x346a,    \
+   0x398e,    \
+   0x2c30,    \
+   0x28fa,    \
+   0x2a95,    \
+   0x3ae8,    \
+   0x2c91,    \
+   0x381a,    \
+   0x2788,    \
+   0x26b0,    \
+   0x34f8,    \
+   0x3c27,    \
+   0x34db,    \
+   0x3780,    \
+   0x3398,    \
+   0x3b58,    \
+   0x3172,    \
+   0x31e2,    \
+   0x350c,    \
+   0x3bb2,    \
+   0x3748,    \
+   0x350c,    \
+   0x3b75,    \
+   0x34aa,    \
+   0x2eb6,    \
+   0x381a,    \
+   0x2cba,    \
+   0x2b1d,    \
+   0x2613,    \
+   0x3496,    \
+   0x2abf,    \
+   0x3102,    \
+   0x2936,    \
+   0x3351,    \
+   0x2bbd,    \
+   0x3496,    \
+   0x3398,    \
+   0x3015,    \
+   0x2a34,    \
+   0x2613,    \
+   0x2e43,    \
+   0x2ded,    \
+   0x29db,    \
+   0x37a5,    \
+   0x296a,    \
+   0x2bbd,    \
+   0x3bca,    \
+   0x37a5,    \
+   0x3a2a,    \
+   0x2f9f,    \
+   0x3709,    \
+   0x29c1,    \
+   0x3bca,    \
+   0x398e,    \
+   0x3595,    \
+   0x29f8,    \
+   0x2da3,    \
+   0x3568,    \
+   0x3485,    \
+   0x2da3,    \
+   0x38a3,    \
+   0x3323,    \
+   0x2d90,    \
+   0x344f,    \
+   0x26b0,    \
+   0x2a0c,    \
+   0x3534,    \
+   0x3b03,    \
+   0x3b28,    \
+   0x3c72,    \
+   0x2f5b,    \
+   0x2a34,    \
+   0x3436,    \
+   0x30ca,    \
+   0x3730,    \
+   0x3211,    \
+   0x32e0,    \
+   0x3126,    \
+   0x3003,    \
+   0x395d,    \
+   0x294e,    \
+   0x38ed,    \
+   0x3c9c,    \
+   0x2996,    \
+   0x2b32,    \
+   0x313d,    \
+   0x350c,    \
+   0x2d90,    \
+   0x3d11,    \
+   0x2eb6,    \
+   0x2f78,    \
+   0x39a1,    \
+   0x3351,    \
+   0x2da3,    \
+   0x2985,    \
+   0x3225,    \
+   0x2c62,    \
+   0x3ce2,    \
+   0x3682,    \
+   0x39cf,    \
+   0x2aa8,    \
+   0x2ea1,    \
+   0x2d1a,    \
+   0x344f,    \
+   0x290e,    \
+   0x2700,    \
+   0x2af5,    \
+   0x3808,    \
+   0x3398,    \
+   0x3211,    \
+   0x2bbd,    \
+   0x3780,    \
+   0x3323,    \
+   0x3172,    \
+   0x350c,    \
+   0x3acc,    \
+   0x3058,    \
+   0x3ab4,    \
+   0x3200,    \
+   0x2fb4,    \
+   0x3cb0,    \
+   0x2cba,    \
+   0x3015,    \
+   0x354c,    \
+   0x26e2,    \
+   0x2cef,    \
+   0x3b28,    \
+   0x2c30,    \
+   0x3c02,    \
+   0x2700,    \
+   0x2788,    \
+   0x2fb4,    \
+   0x28ac,    \
+   0x3b03,    \
+   0x33c3,    \
+   0x34db,    \
+   0x3b28,    \
+   0x3780,    \
+   0x3b58,    \
+   0x3d00,    \
+   0x3075,    \
+   0x28fa,    \
+   0x34db,    \
+   0x34db,    \
+   0x279a,    \
+   0x2ea1,    \
+   0x3386,    \
+   0x37bb,    \
+   0x35bf,    \
+   0x2eb6,    \
+   0x371c,    \
+   0x2e7b,    \
+   0x2d90,    \
+   0x2f78,    \
+   0x32ae,    \
+   0x3b03,    \
+   0x269b,    \
+   0x2ba6,    \
+   0x2ba6,    \
+   0x364a,    \
+   0x2c91,    \
+   0x279a,    \
+   0x28fa,    \
+   0x3a9f,    \
+   0x33c3,    \
+   0x2abf,    \
+   0x3a2a,    \
+   0x3846,    \
+   0x2f2a,    \
+   0x2b93,    \
+   0x397b,    \
+   0x2b4a,    \
+   0x371c,    \
+   0x2fcc,    \
+   0x2a1f,    \
+   0x37f0,    \
+   0x3943,    \
+   0x3730,    \
+   0x3d00,    \
+   0x2c80,    \
+   0x346a,    \
+   0x39a1,    \
+   0x32ae,    \
+   0x308b,    \
+   0x39cf,    \
+   0x32e0,    \
+   0x3c02,    \
+   0x2aa8,    \
+   0x3568,    \
+   0x3c89,    \
+   0x2724,    \
+   0x296a,    \
+   0x3acc,    \
+   0x296a,    \
+   0x2e7b,    \
+   0x2e5d,    \
+   0x2936,    \
+   0x3c27,    \
+   0x3a9f,    \
+   0x2ca5,    \
+   0x2a4c,    \
+   0x2f41,    \
+   0x2b82,    \
+   0x28c3,    \
+   0x39a1,    \
+   0x286d,    \
+   0x308b,    \
+   0x3253,    \
+   0x395d,    \
+   0x2e8e,    \
+   0x2db8,    \
+   0x3906,    \
+   0x2c48,    \
+   0x3310,    \
+   0x326f,    \
+   0x2810,    \
+   0x3172,    \
+   0x3632,    \
+   0x2b93,    \
+   0x3a5b,    \
+   0x34db,    \
+   0x290e,    \
+   0x294e,    \
+   0x3c3d,    \
+   0x3534,    \
+   0x29c1,    \
+   0x37a5,    \
+   0x39b6,    \
+   0x3c13,    \
+   0x3acc,    \
+   0x269b,    \
+   0x33dd,    \
+   0x29db,    \
+   0x2851,    \
+   0x323b,    \
+   0x3436,    \
+   0x3709,    \
+   0x26c8,    \
+   0x2656,    \
+   0x354c,    \
+   0x2810,    \
+   0x3ab4,    \
+   0x3b9d,    \
+   0x371c,    \
+   0x2838,    \
+   0x3386,    \
+   0x3b15,    \
+   0x3bca,    \
+   0x32e0,    \
+   0x2a67,    \
+   0x2838,    \
+   0x2c91,    \
+   0x3189,    \
+   0x279a,    \
+   0x3225,    \
+   0x27c5,    \
+   0x3730,    \
+   0x2cd3,    \
+   0x269b,    \
+   0x3225,    \
+   0x309d,    \
+   0x26e2,    \
+   0x2a83,    \
+   0x36a7,    \
+   0x34aa,    \
+   0x2bf2,    \
+   0x387e,    \
+   0x3c3d,    \
+   0x28dd,    \
+   0x3730,    \
+   0x3253,    \
+   0x2ca5,    \
+   0x3338,    \
+   0x2a67,    \
+   0x2ea1,    \
+   0x2d7d,    \
+   0x3496,    \
+   0x2cba,    \
+   0x2e2c,    \
+   0x2985,    \
+   0x3126,    \
+   0x2ecf,    \
+   0x2e43,    \
+   0x32c5,    \
+   0x3be5,    \
+   0x2e18,    \
+   0x3386,    \
+   0x3906,    \
+   0x2689,    \
+   0x33fb,    \
+   0x31e2,    \
+   0x3762,    \
+   0x2b65,    \
+   0x3a5b,    \
+   0x3a8c,    \
+   0x263c,    \
+   0x27ae,    \
+   0x2a4c,    \
+   0x2d60,    \
+   0x3860,    \
+   0x303f,    \
+   0x2724,    \
+   0x344f,    \
+   0x2d08,    \
+   0x351f,    \
+   0x361d,    \
+   0x30ca,    \
+   0x37bb,    \
+   0x3ce2,    \
+   0x2724,    \
+   0x3ab4,    \
+   0x2b65,    \
+   0x3860,    \
+   0x33c3,    \
+   0x3015,    \
+   0x2c1b,    \
+   0x3189,    \
+   0x2886,    \
+   0x30b2,    \
+   0x2996,    \
+   0x3288,    \
+   0x3918,    \
+   0x2b93,    \
+   0x2f9f,    \
+   0x2613,    \
+   0x33fb,    \
+   0x3791,    \
+   0x3c13,    \
+   0x3113,    \
+   0x2af5,    \
+   0x3791,    \
+   0x2e43,    \
+   0x2ca5,    \
+   0x27ae,    \
+   0x3b15,    \
+   0x34aa,    \
+   0x3338,    \
+   0x39a1,    \
+   0x387e,    \
+   0x3172,    \
+   0x2b0b,    \
+   0x34aa,    \
+   0x2ded,    \
+   0x3be5,    \
+   0x3323,    \
+   0x3568,    \
+   0x2f41,    \
+   0x2bd6,    \
+   0x3568,    \
+   0x34c1,    \
+   0x2689,    \
+   0x2672,    \
+   0x2eb6,    \
+   0x3b8b,    \
+   0x3058,    \
+   0x3113,    \
+   0x3583,    \
+   0x2a34,    \
+   0x3906,    \
+   0x3bb2,    \
+   0x2eea,    \
+   0x2ad8,    \
+   0x319c,    \
+   0x3693,    \
+   0x33ac,    \
+   0x38b8,    \
+   0x2689,    \
+   0x31e2,    \
+   0x33ac,    \
+   0x27ae,    \
+   0x2bbd,    \
+   0x360b,    \
+   0x2985,    \
+   0x28ac,    \
+   0x2b1d,    \
+   0x336d,    \
+   0x35bf,    \
+   0x323b,    \
+   0x2e8e,    \
+   0x351f,    \
+   0x2b82,    \
+   0x313d,    \
+   0x3421,    \
+   0x2f9f,    \
+   0x3b8b,    \
+   0x2fe7,    \
+   0x360b,    \
+   0x2b32,    \
+   0x2b32,    \
+   0x27c5,    \
+   0x273a,    \
+   0x3b75,    \
+   0x2626,    \
+   0x2f2a,    \
+   0x296a,    \
+   0x3730,    \
+   0x37d3,    \
+   0x2700,    \
+   0x395d,    \
+   0x30ca,    \
+   0x2921,    \
+   0x3156,    \
+   0x3c9c,    \
+   0x2fe7,    \
+   0x2788,    \
+   0x2bd6,    \
+   0x35bf,    \
+   0x3a16,    \
+   0x36a7,    \
+   0x351f,    \
+   0x395d,    \
+   0x3288,    \
+   0x3a2a,    \
+   0x37a5,    \
+   0x2bbd,    \
+   0x2d1a,    \
+   0x34db,    \
+   0x3126,    \
+   0x2fe7,    \
+   0x33fb,    \
+   0x294e,    \
+   0x2bbd,    \
+   0x3156,    \
+   0x2e7b,    \
+   0x2a67,    \
+   0x38ed,    \
+   0x397b,    \
+   0x30ca,    \
+   0x2f78,    \
+   0x2a0c,    \
+   0x3075,    \
+   0x2d2e,    \
+   0x2656,    \
+   0x3a16,    \
+   0x2a83,    \
+   0x2c80,    \
+   0x3693,    \
+   0x354c,    \
+   0x2788,    \
+   0x36f2,    \
+   0x37f0,    \
+   0x27ae,    \
+   0x34f8,    \
+   0x2b82,    \
+   0x35a8,    \
+   0x3632,    \
+   0x2f16,    \
+}
+
+#endif
\ No newline at end of file
diff --git a/hwpe/softex/inc/scores.h b/hwpe/softex/inc/scores.h
new file mode 100644
index 0000000..d60ba7a
--- /dev/null
+++ b/hwpe/softex/inc/scores.h
@@ -0,0 +1,1037 @@
+#ifndef __SOFTEX_SCORES__
+#define __SOFTEX_SCORES__
+
+#define LENGTH  1024
+
+#define FMT_WIDTH  2
+
+#define N_VECTORS  1
+
+#define SCORES {  \
+   0x4182,    \
+   0x4187,    \
+   0x4114,    \
+   0x4102,    \
+   0x41f8,    \
+   0x4110,    \
+   0x41d9,    \
+   0x4168,    \
+   0x41f7,    \
+   0x41d6,    \
+   0x4168,    \
+   0x4152,    \
+   0x4120,    \
+   0x41b7,    \
+   0x0000,    \
+   0x41de,    \
+   0x418f,    \
+   0x4189,    \
+   0x4190,    \
+   0x40f0,    \
+   0x4156,    \
+   0x4110,    \
+   0x414e,    \
+   0x4078,    \
+   0x41be,    \
+   0x41ea,    \
+   0x41ec,    \
+   0x41b4,    \
+   0x4106,    \
+   0x41cd,    \
+   0x410e,    \
+   0x41f9,    \
+   0x41d0,    \
+   0x4118,    \
+   0x414a,    \
+   0x40a8,    \
+   0x4114,    \
+   0x4090,    \
+   0x41d2,    \
+   0x41cf,    \
+   0x41db,    \
+   0x41e3,    \
+   0x41d0,    \
+   0x4160,    \
+   0x41e2,    \
+   0x413a,    \
+   0x4126,    \
+   0x414e,    \
+   0x4182,    \
+   0x4154,    \
+   0x40ac,    \
+   0x41b4,    \
+   0x41a4,    \
+   0x419a,    \
+   0x41e7,    \
+   0x41c8,    \
+   0x4160,    \
+   0x41ae,    \
+   0x3f20,    \
+   0x41b4,    \
+   0x4185,    \
+   0x41c2,    \
+   0x4198,    \
+   0x41bf,    \
+   0x41e1,    \
+   0x41d9,    \
+   0x3f00,    \
+   0x3ff0,    \
+   0x41a4,    \
+   0x3ec0,    \
+   0x4118,    \
+   0x41f4,    \
+   0x4197,    \
+   0x4195,    \
+   0x41d9,    \
+   0x4164,    \
+   0x4188,    \
+   0x41ed,    \
+   0x41e4,    \
+   0x3fe0,    \
+   0x41b1,    \
+   0x416a,    \
+   0x41aa,    \
+   0x41f9,    \
+   0x3ff0,    \
+   0x4068,    \
+   0x41ae,    \
+   0x41ce,    \
+   0x419e,    \
+   0x413e,    \
+   0x415c,    \
+   0x41f6,    \
+   0x41de,    \
+   0x3e00,    \
+   0x411e,    \
+   0x419d,    \
+   0x41c6,    \
+   0x4182,    \
+   0x41b9,    \
+   0x412a,    \
+   0x4195,    \
+   0x41e0,    \
+   0x4199,    \
+   0x4184,    \
+   0x4038,    \
+   0x41f3,    \
+   0x412c,    \
+   0x4186,    \
+   0x4060,    \
+   0x419b,    \
+   0x41cc,    \
+   0x41c7,    \
+   0x416c,    \
+   0x41c8,    \
+   0x41bd,    \
+   0x3f20,    \
+   0x4010,    \
+   0x41f2,    \
+   0x40e4,    \
+   0x4197,    \
+   0x41f7,    \
+   0x3ec0,    \
+   0x410a,    \
+   0x4116,    \
+   0x3f00,    \
+   0x41e1,    \
+   0x41ab,    \
+   0x41d5,    \
+   0x4100,    \
+   0x40c0,    \
+   0x41cc,    \
+   0x41f6,    \
+   0x41bc,    \
+   0x41e5,    \
+   0x41a8,    \
+   0x41f4,    \
+   0x40a0,    \
+   0x40c8,    \
+   0x3f40,    \
+   0x419f,    \
+   0x4185,    \
+   0x41d5,    \
+   0x4088,    \
+   0x41ef,    \
+   0x4116,    \
+   0x41e0,    \
+   0x41cd,    \
+   0x4187,    \
+   0x4118,    \
+   0x4048,    \
+   0x4186,    \
+   0x40e8,    \
+   0x40fc,    \
+   0x40c0,    \
+   0x4193,    \
+   0x4193,    \
+   0x4038,    \
+   0x4160,    \
+   0x41c9,    \
+   0x41c7,    \
+   0x4114,    \
+   0x41b5,    \
+   0x411e,    \
+   0x4112,    \
+   0x412a,    \
+   0x41fb,    \
+   0x4156,    \
+   0x4040,    \
+   0x41b8,    \
+   0x4188,    \
+   0x4030,    \
+   0x3fe0,    \
+   0x41d6,    \
+   0x414c,    \
+   0x41b9,    \
+   0x4008,    \
+   0x4176,    \
+   0x417c,    \
+   0x4194,    \
+   0x4146,    \
+   0x4154,    \
+   0x41ad,    \
+   0x41d4,    \
+   0x415e,    \
+   0x41c1,    \
+   0x4162,    \
+   0x413c,    \
+   0x41b8,    \
+   0x41db,    \
+   0x41a4,    \
+   0x414c,    \
+   0x41d9,    \
+   0x41fd,    \
+   0x4080,    \
+   0x40bc,    \
+   0x41d1,    \
+   0x408c,    \
+   0x41e5,    \
+   0x4172,    \
+   0x416e,    \
+   0x41f2,    \
+   0x4181,    \
+   0x412c,    \
+   0x4132,    \
+   0x41b8,    \
+   0x41b2,    \
+   0x4028,    \
+   0x4140,    \
+   0x41de,    \
+   0x4140,    \
+   0x40c0,    \
+   0x41c5,    \
+   0x419d,    \
+   0x4198,    \
+   0x4194,    \
+   0x4018,    \
+   0x414e,    \
+   0x41b1,    \
+   0x4116,    \
+   0x41cc,    \
+   0x41aa,    \
+   0x41cd,    \
+   0x41ef,    \
+   0x41f2,    \
+   0x4000,    \
+   0x4028,    \
+   0x4058,    \
+   0x4176,    \
+   0x4144,    \
+   0x3fe0,    \
+   0x4152,    \
+   0x41a4,    \
+   0x411e,    \
+   0x4166,    \
+   0x41b1,    \
+   0x40b4,    \
+   0x41ac,    \
+   0x41b6,    \
+   0x4140,    \
+   0x40a8,    \
+   0x41d6,    \
+   0x414a,    \
+   0x40fc,    \
+   0x41d2,    \
+   0x3e80,    \
+   0x4130,    \
+   0x41e6,    \
+   0x3fa0,    \
+   0x4078,    \
+   0x4148,    \
+   0x40c0,    \
+   0x416a,    \
+   0x40e0,    \
+   0x40cc,    \
+   0x41bc,    \
+   0x4080,    \
+   0x419d,    \
+   0x41a8,    \
+   0x41e8,    \
+   0x415a,    \
+   0x40cc,    \
+   0x4038,    \
+   0x4189,    \
+   0x418c,    \
+   0x41f8,    \
+   0x40a8,    \
+   0x4188,    \
+   0x40a8,    \
+   0x4186,    \
+   0x4110,    \
+   0x3fa0,    \
+   0x4102,    \
+   0x41f1,    \
+   0x41a5,    \
+   0x40a8,    \
+   0x41aa,    \
+   0x4018,    \
+   0x41be,    \
+   0x3f00,    \
+   0x41be,    \
+   0x41df,    \
+   0x4162,    \
+   0x418d,    \
+   0x4198,    \
+   0x41b7,    \
+   0x41c6,    \
+   0x4197,    \
+   0x41d2,    \
+   0x4178,    \
+   0x41c1,    \
+   0x4114,    \
+   0x4108,    \
+   0x41f8,    \
+   0x4116,    \
+   0x4030,    \
+   0x3e80,    \
+   0x41c9,    \
+   0x4080,    \
+   0x41c9,    \
+   0x41d6,    \
+   0x4068,    \
+   0x41ce,    \
+   0x418e,    \
+   0x3fc0,    \
+   0x4181,    \
+   0x0000,    \
+   0x3fd0,    \
+   0x4118,    \
+   0x40a4,    \
+   0x40f4,    \
+   0x4150,    \
+   0x4134,    \
+   0x419e,    \
+   0x40f4,    \
+   0x409c,    \
+   0x4122,    \
+   0x41fd,    \
+   0x41f2,    \
+   0x3f00,    \
+   0x4070,    \
+   0x412e,    \
+   0x41a4,    \
+   0x40e8,    \
+   0x41d0,    \
+   0x4010,    \
+   0x41d1,    \
+   0x4106,    \
+   0x417c,    \
+   0x418a,    \
+   0x4030,    \
+   0x40fc,    \
+   0x41e8,    \
+   0x4189,    \
+   0x41e9,    \
+   0x41ef,    \
+   0x40b0,    \
+   0x41e7,    \
+   0x41ca,    \
+   0x413c,    \
+   0x4142,    \
+   0x4050,    \
+   0x41ea,    \
+   0x4144,    \
+   0x417a,    \
+   0x410a,    \
+   0x41e5,    \
+   0x41ae,    \
+   0x410a,    \
+   0x418d,    \
+   0x4126,    \
+   0x415e,    \
+   0x41cb,    \
+   0x41f2,    \
+   0x412e,    \
+   0x41df,    \
+   0x41e8,    \
+   0x418c,    \
+   0x419f,    \
+   0x41cd,    \
+   0x41a5,    \
+   0x41e0,    \
+   0x41ec,    \
+   0x41b9,    \
+   0x41f0,    \
+   0x40e8,    \
+   0x41d9,    \
+   0x41b7,    \
+   0x40c4,    \
+   0x411c,    \
+   0x41cd,    \
+   0x41d3,    \
+   0x41c0,    \
+   0x41df,    \
+   0x41cb,    \
+   0x4154,    \
+   0x41fd,    \
+   0x41be,    \
+   0x41a3,    \
+   0x4008,    \
+   0x41ad,    \
+   0x40f8,    \
+   0x41d4,    \
+   0x4197,    \
+   0x41d5,    \
+   0x4128,    \
+   0x4144,    \
+   0x419e,    \
+   0x41af,    \
+   0x413e,    \
+   0x41f4,    \
+   0x40c8,    \
+   0x4102,    \
+   0x4038,    \
+   0x41bb,    \
+   0x41c8,    \
+   0x41dd,    \
+   0x41ca,    \
+   0x4168,    \
+   0x41e7,    \
+   0x41ee,    \
+   0x41a9,    \
+   0x41ef,    \
+   0x4134,    \
+   0x3fd0,    \
+   0x4191,    \
+   0x4080,    \
+   0x40f0,    \
+   0x41a2,    \
+   0x4118,    \
+   0x3f90,    \
+   0x41d4,    \
+   0x4114,    \
+   0x40c0,    \
+   0x40e8,    \
+   0x40c0,    \
+   0x41bf,    \
+   0x41bd,    \
+   0x418a,    \
+   0x418d,    \
+   0x41f6,    \
+   0x4134,    \
+   0x41c2,    \
+   0x4138,    \
+   0x4166,    \
+   0x41dd,    \
+   0x413c,    \
+   0x418f,    \
+   0x41f3,    \
+   0x4028,    \
+   0x4114,    \
+   0x41f9,    \
+   0x4040,    \
+   0x3fa0,    \
+   0x4136,    \
+   0x4195,    \
+   0x41c6,    \
+   0x419c,    \
+   0x41a9,    \
+   0x4183,    \
+   0x409c,    \
+   0x416c,    \
+   0x4152,    \
+   0x40b0,    \
+   0x41f3,    \
+   0x418e,    \
+   0x40c4,    \
+   0x4058,    \
+   0x41aa,    \
+   0x4189,    \
+   0x4000,    \
+   0x41d8,    \
+   0x41bd,    \
+   0x4048,    \
+   0x3f40,    \
+   0x0000,    \
+   0x3f00,    \
+   0x41b4,    \
+   0x4102,    \
+   0x41cc,    \
+   0x4150,    \
+   0x4196,    \
+   0x4048,    \
+   0x4040,    \
+   0x41da,    \
+   0x4132,    \
+   0x41f6,    \
+   0x41da,    \
+   0x41d5,    \
+   0x4080,    \
+   0x40c4,    \
+   0x41fb,    \
+   0x414a,    \
+   0x413a,    \
+   0x4150,    \
+   0x41f1,    \
+   0x41c4,    \
+   0x4008,    \
+   0x413a,    \
+   0x41e4,    \
+   0x41d4,    \
+   0x4018,    \
+   0x417a,    \
+   0x41c3,    \
+   0x41f7,    \
+   0x417c,    \
+   0x411e,    \
+   0x41b5,    \
+   0x4156,    \
+   0x411a,    \
+   0x41b2,    \
+   0x40d0,    \
+   0x41bb,    \
+   0x4116,    \
+   0x4170,    \
+   0x4156,    \
+   0x40b4,    \
+   0x41fc,    \
+   0x41ce,    \
+   0x41d7,    \
+   0x4154,    \
+   0x41da,    \
+   0x4184,    \
+   0x4188,    \
+   0x417a,    \
+   0x3fc0,    \
+   0x4094,    \
+   0x41e1,    \
+   0x4108,    \
+   0x4196,    \
+   0x414e,    \
+   0x4118,    \
+   0x41dc,    \
+   0x41d2,    \
+   0x41eb,    \
+   0x3fc0,    \
+   0x3fe0,    \
+   0x4184,    \
+   0x40b4,    \
+   0x418b,    \
+   0x418c,    \
+   0x40e4,    \
+   0x4198,    \
+   0x4166,    \
+   0x4176,    \
+   0x41c3,    \
+   0x41b2,    \
+   0x418f,    \
+   0x41b5,    \
+   0x3e80,    \
+   0x4104,    \
+   0x41bc,    \
+   0x40a8,    \
+   0x41b6,    \
+   0x41db,    \
+   0x4142,    \
+   0x41fe,    \
+   0x416e,    \
+   0x419c,    \
+   0x41cd,    \
+   0x414a,    \
+   0x417c,    \
+   0x3e80,    \
+   0x3f20,    \
+   0x41dc,    \
+   0x4118,    \
+   0x3f80,    \
+   0x415e,    \
+   0x408c,    \
+   0x3f20,    \
+   0x41d5,    \
+   0x412a,    \
+   0x41c5,    \
+   0x41a8,    \
+   0x41b6,    \
+   0x418f,    \
+   0x40e4,    \
+   0x4146,    \
+   0x419c,    \
+   0x0000,    \
+   0x40e8,    \
+   0x417e,    \
+   0x4190,    \
+   0x4102,    \
+   0x414c,    \
+   0x41c1,    \
+   0x4195,    \
+   0x4172,    \
+   0x4189,    \
+   0x4116,    \
+   0x41f6,    \
+   0x4140,    \
+   0x4198,    \
+   0x414e,    \
+   0x40fc,    \
+   0x41e8,    \
+   0x4114,    \
+   0x41c1,    \
+   0x419f,    \
+   0x41d8,    \
+   0x4108,    \
+   0x4080,    \
+   0x40c8,    \
+   0x41e7,    \
+   0x4110,    \
+   0x41c8,    \
+   0x4000,    \
+   0x3f60,    \
+   0x41a5,    \
+   0x41f5,    \
+   0x41a4,    \
+   0x41c1,    \
+   0x4196,    \
+   0x41ec,    \
+   0x417c,    \
+   0x4183,    \
+   0x41a6,    \
+   0x41f0,    \
+   0x41bf,    \
+   0x41a6,    \
+   0x41ed,    \
+   0x41a2,    \
+   0x4140,    \
+   0x41c8,    \
+   0x4114,    \
+   0x40e0,    \
+   0x0000,    \
+   0x41a1,    \
+   0x40d0,    \
+   0x4172,    \
+   0x408c,    \
+   0x4193,    \
+   0x40fc,    \
+   0x41a1,    \
+   0x4196,    \
+   0x415e,    \
+   0x40b8,    \
+   0x0000,    \
+   0x4136,    \
+   0x412e,    \
+   0x40a8,    \
+   0x41c3,    \
+   0x4094,    \
+   0x40fc,    \
+   0x41f1,    \
+   0x41c3,    \
+   0x41df,    \
+   0x4154,    \
+   0x41bc,    \
+   0x40a4,    \
+   0x41f1,    \
+   0x41d8,    \
+   0x41ac,    \
+   0x40ac,    \
+   0x4128,    \
+   0x41aa,    \
+   0x41a0,    \
+   0x4128,    \
+   0x41ce,    \
+   0x4191,    \
+   0x4126,    \
+   0x419e,    \
+   0x3f60,    \
+   0x40b0,    \
+   0x41a8,    \
+   0x41e8,    \
+   0x41ea,    \
+   0x41f8,    \
+   0x414e,    \
+   0x40b8,    \
+   0x419d,    \
+   0x416e,    \
+   0x41be,    \
+   0x4185,    \
+   0x418e,    \
+   0x4176,    \
+   0x415c,    \
+   0x41d6,    \
+   0x4090,    \
+   0x41d1,    \
+   0x41fa,    \
+   0x409c,    \
+   0x40e4,    \
+   0x4178,    \
+   0x41a6,    \
+   0x4126,    \
+   0x41ff,    \
+   0x4140,    \
+   0x4150,    \
+   0x41d9,    \
+   0x4193,    \
+   0x4128,    \
+   0x4098,    \
+   0x4186,    \
+   0x410c,    \
+   0x41fd,    \
+   0x41b6,    \
+   0x41db,    \
+   0x40cc,    \
+   0x413e,    \
+   0x411c,    \
+   0x419e,    \
+   0x4084,    \
+   0x3fa0,    \
+   0x40d8,    \
+   0x41c7,    \
+   0x4196,    \
+   0x4185,    \
+   0x40fc,    \
+   0x41c1,    \
+   0x4191,    \
+   0x417c,    \
+   0x41a6,    \
+   0x41e6,    \
+   0x4164,    \
+   0x41e5,    \
+   0x4184,    \
+   0x4156,    \
+   0x41fb,    \
+   0x4114,    \
+   0x415e,    \
+   0x41a9,    \
+   0x3f90,    \
+   0x4118,    \
+   0x41ea,    \
+   0x4108,    \
+   0x41f3,    \
+   0x3fa0,    \
+   0x4000,    \
+   0x4156,    \
+   0x4068,    \
+   0x41e8,    \
+   0x4198,    \
+   0x41a4,    \
+   0x41ea,    \
+   0x41c1,    \
+   0x41ec,    \
+   0x41fe,    \
+   0x4166,    \
+   0x4080,    \
+   0x41a4,    \
+   0x41a4,    \
+   0x4008,    \
+   0x413e,    \
+   0x4195,    \
+   0x41c4,    \
+   0x41ae,    \
+   0x4140,    \
+   0x41bd,    \
+   0x413a,    \
+   0x4126,    \
+   0x4150,    \
+   0x418c,    \
+   0x41e8,    \
+   0x3f40,    \
+   0x40f8,    \
+   0x40f8,    \
+   0x41b4,    \
+   0x4110,    \
+   0x4008,    \
+   0x4080,    \
+   0x41e4,    \
+   0x4198,    \
+   0x40d0,    \
+   0x41df,    \
+   0x41ca,    \
+   0x414a,    \
+   0x40f4,    \
+   0x41d7,    \
+   0x40e8,    \
+   0x41bd,    \
+   0x4158,    \
+   0x40b4,    \
+   0x41c6,    \
+   0x41d5,    \
+   0x41be,    \
+   0x41fe,    \
+   0x410e,    \
+   0x419f,    \
+   0x41d9,    \
+   0x418c,    \
+   0x4168,    \
+   0x41db,    \
+   0x418e,    \
+   0x41f3,    \
+   0x40cc,    \
+   0x41aa,    \
+   0x41f9,    \
+   0x3fc0,    \
+   0x4094,    \
+   0x41e6,    \
+   0x4094,    \
+   0x413a,    \
+   0x4138,    \
+   0x408c,    \
+   0x41f5,    \
+   0x41e4,    \
+   0x4112,    \
+   0x40bc,    \
+   0x414c,    \
+   0x40f0,    \
+   0x4070,    \
+   0x41d9,    \
+   0x4050,    \
+   0x4168,    \
+   0x4188,    \
+   0x41d6,    \
+   0x413c,    \
+   0x412a,    \
+   0x41d2,    \
+   0x410a,    \
+   0x4190,    \
+   0x4189,    \
+   0x4030,    \
+   0x417c,    \
+   0x41b3,    \
+   0x40f4,    \
+   0x41e1,    \
+   0x41a4,    \
+   0x4084,    \
+   0x4090,    \
+   0x41f6,    \
+   0x41a8,    \
+   0x40a4,    \
+   0x41c3,    \
+   0x41da,    \
+   0x41f4,    \
+   0x41e6,    \
+   0x3f40,    \
+   0x4199,    \
+   0x40a8,    \
+   0x4048,    \
+   0x4187,    \
+   0x419d,    \
+   0x41bc,    \
+   0x3f80,    \
+   0x3ec0,    \
+   0x41a9,    \
+   0x4030,    \
+   0x41e5,    \
+   0x41ef,    \
+   0x41bd,    \
+   0x4040,    \
+   0x4195,    \
+   0x41e9,    \
+   0x41f1,    \
+   0x418e,    \
+   0x40c0,    \
+   0x4040,    \
+   0x4110,    \
+   0x417e,    \
+   0x4008,    \
+   0x4186,    \
+   0x4018,    \
+   0x41be,    \
+   0x4116,    \
+   0x3f40,    \
+   0x4186,    \
+   0x416a,    \
+   0x3f90,    \
+   0x40c4,    \
+   0x41b8,    \
+   0x41a2,    \
+   0x4102,    \
+   0x41cc,    \
+   0x41f6,    \
+   0x4078,    \
+   0x41be,    \
+   0x4188,    \
+   0x4112,    \
+   0x4192,    \
+   0x40c0,    \
+   0x413e,    \
+   0x4124,    \
+   0x41a1,    \
+   0x4114,    \
+   0x4134,    \
+   0x4098,    \
+   0x4176,    \
+   0x4142,    \
+   0x4136,    \
+   0x418d,    \
+   0x41f2,    \
+   0x4132,    \
+   0x4195,    \
+   0x41d2,    \
+   0x3f20,    \
+   0x419a,    \
+   0x4183,    \
+   0x41c0,    \
+   0x40ec,    \
+   0x41e1,    \
+   0x41e3,    \
+   0x3e80,    \
+   0x4010,    \
+   0x40bc,    \
+   0x4122,    \
+   0x41cb,    \
+   0x4162,    \
+   0x3fc0,    \
+   0x419e,    \
+   0x411a,    \
+   0x41a7,    \
+   0x41b2,    \
+   0x416e,    \
+   0x41c4,    \
+   0x41fd,    \
+   0x3fc0,    \
+   0x41e5,    \
+   0x40ec,    \
+   0x41cb,    \
+   0x4198,    \
+   0x415e,    \
+   0x4106,    \
+   0x417e,    \
+   0x4058,    \
+   0x416c,    \
+   0x409c,    \
+   0x418a,    \
+   0x41d3,    \
+   0x40f4,    \
+   0x4154,    \
+   0x0000,    \
+   0x419a,    \
+   0x41c2,    \
+   0x41f4,    \
+   0x4174,    \
+   0x40d8,    \
+   0x41c2,    \
+   0x4136,    \
+   0x4112,    \
+   0x4010,    \
+   0x41e9,    \
+   0x41a2,    \
+   0x4192,    \
+   0x41d9,    \
+   0x41cc,    \
+   0x417c,    \
+   0x40dc,    \
+   0x41a2,    \
+   0x412e,    \
+   0x41f2,    \
+   0x4191,    \
+   0x41aa,    \
+   0x414c,    \
+   0x4100,    \
+   0x41aa,    \
+   0x41a3,    \
+   0x3f20,    \
+   0x3f00,    \
+   0x4140,    \
+   0x41ee,    \
+   0x4164,    \
+   0x4174,    \
+   0x41ab,    \
+   0x40b8,    \
+   0x41d2,    \
+   0x41f0,    \
+   0x4144,    \
+   0x40d4,    \
+   0x4180,    \
+   0x41b7,    \
+   0x4197,    \
+   0x41cf,    \
+   0x3f20,    \
+   0x4183,    \
+   0x4197,    \
+   0x4010,    \
+   0x40fc,    \
+   0x41b1,    \
+   0x4098,    \
+   0x4068,    \
+   0x40e0,    \
+   0x4194,    \
+   0x41ae,    \
+   0x4187,    \
+   0x413c,    \
+   0x41a7,    \
+   0x40f0,    \
+   0x4178,    \
+   0x419c,    \
+   0x4154,    \
+   0x41ee,    \
+   0x415a,    \
+   0x41b1,    \
+   0x40e4,    \
+   0x40e4,    \
+   0x4018,    \
+   0x3fd0,    \
+   0x41ed,    \
+   0x3e00,    \
+   0x414a,    \
+   0x4094,    \
+   0x41be,    \
+   0x41c5,    \
+   0x3fa0,    \
+   0x41d6,    \
+   0x416e,    \
+   0x4088,    \
+   0x417a,    \
+   0x41fa,    \
+   0x415a,    \
+   0x4000,    \
+   0x4100,    \
+   0x41ae,    \
+   0x41de,    \
+   0x41b8,    \
+   0x41a7,    \
+   0x41d6,    \
+   0x418a,    \
+   0x41df,    \
+   0x41c3,    \
+   0x40fc,    \
+   0x411c,    \
+   0x41a4,    \
+   0x4176,    \
+   0x415a,    \
+   0x419a,    \
+   0x4090,    \
+   0x40fc,    \
+   0x417a,    \
+   0x413a,    \
+   0x40c0,    \
+   0x41d1,    \
+   0x41d7,    \
+   0x416e,    \
+   0x4150,    \
+   0x40b0,    \
+   0x4166,    \
+   0x411e,    \
+   0x3ec0,    \
+   0x41de,    \
+   0x40c4,    \
+   0x410e,    \
+   0x41b7,    \
+   0x41a9,    \
+   0x4000,    \
+   0x41bb,    \
+   0x41c6,    \
+   0x4010,    \
+   0x41a5,    \
+   0x40f0,    \
+   0x41ad,    \
+   0x41b3,    \
+   0x4148     \
+}
+
+#endif
\ No newline at end of file
diff --git a/hwpe/softex/softex.c b/hwpe/softex/softex.c
new file mode 100644
index 0000000..3126f1a
--- /dev/null
+++ b/hwpe/softex/softex.c
@@ -0,0 +1,95 @@
+/*
+ * Andrea Belano <andrea.belano@studio.unibo.it>
+ *
+ * Copyright 2024 ETH Zurich and University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ *
+ * SPDX-License-Identifier: Apache-2.0
+ */
+
+#include <stdint.h>
+#include <memory.h>
+#include "stdio.h"
+#include "archi_softex.h"
+#include "hal_softex.h"
+#include "pulp.h"
+
+#include "inc/golden.h"
+#include "inc/scores.h"
+
+#define TOLERANCE 0x2
+
+uint16_t scores_ext [LENGTH] = SCORES;
+uint16_t golden [LENGTH]     = GOLDEN;
+
+int main() {
+  volatile int errors = 0;
+
+  uint16_t volatile *scores = (uint16_t volatile *) pi_l1_malloc(0, (FMT_WIDTH*LENGTH));
+
+  if(get_core_id() == 0){
+    #ifdef USE_DMA
+      volatile unsigned int dma_id = 0;
+      dma_id = mchan_alloc();
+
+      mchan_transfer((unsigned int) FMT_WIDTH*LENGTH  ,
+                     (unsigned int) scores_ext        ,
+                     (unsigned int) scores            
+      );
+
+      mchan_barrier(dma_id);
+      mchan_free(dma_id);
+    #else
+      memcpy(scores, scores_ext, LENGTH*FMT_WIDTH);
+    #endif
+
+    // Enable softex
+    hwpe_cg_enable();
+
+    hwpe_soft_clear();
+
+    HWPE_WRITE(scores, SOFTEX_IN_ADDR);
+    HWPE_WRITE(LENGTH * FMT_WIDTH, SOFTEX_TOT_LEN);
+    HWPE_WRITE(scores, SOFTEX_OUT_ADDR);
+
+    hwpe_trigger_job();
+
+    softex_evt_wait();
+
+    // Disable softex
+    hwpe_cg_disable();
+
+    for (int i = 0; i < LENGTH; i++) {
+      uint16_t diff;
+
+      if (golden [i] >= scores[i]) {
+        diff = golden [i] - scores [i];
+      } else {
+        diff = scores [i] - golden [i];
+      }
+
+      if (diff > TOLERANCE) {
+        errors += 1;
+        
+        printf ("Mismatch!!!\tIndex: %d\tExpected: 0x%04x\tWas: 0x%04x\tDifference: 0x%x\n", i, golden [i], scores [i], diff);
+      }
+    }
+
+    *(int *) 0x1A1040A0 = errors;
+
+    printf("Test completed with %d errors\n", errors);
+  }
+  synch_barrier();
+  return errors;
+}
diff --git a/idma-tests.yaml b/idma-tests.yaml
new file mode 100644
index 0000000..6bef5c3
--- /dev/null
+++ b/idma-tests.yaml
@@ -0,0 +1,8 @@
+idma_tests:
+  idma_simple:
+    path: ./idma_tests/idma_simple
+    command: make clean all run
+  idma_zeromem:
+    path: ./idma_tests/idma_zeromem
+    command: make clean all run
+
diff --git a/idma_tests/README.md b/idma_tests/README.md
new file mode 100644
index 0000000..74096cd
--- /dev/null
+++ b/idma_tests/README.md
@@ -0,0 +1,20 @@
+## IDMA Tests
+
+This folder contains basic tests for the iDMA IP. 
+Currently, the following are supported:
+- **1D transfers**
+- **2D transfers**
+- **3D transfers**
+
+To launch each test:
+1. Move into the related folder.
+2. Launch the following:
+    1. `make stimuli` : this will generate the randomized stimuli for the test. The randomized stimuli consist of transfer sizes, number of transfers to be executed, n-dimensional strides, etc ...
+    2. `make all` : this will compile the C code. Few choices are available:
+        - **Single Core mode**: no flags are needed. All transfers specified in the stimuli will be executed by Core 0 only.
+        - **Sequential Multi-Core Mode**: specify **MULTI_CORE_S=1** in the command line when compiling the code. In this mode, all cores will execute the transfers specified in the stimuli in a sequential manner.
+        - **Parallel Multi-Core Mode**: specify **MULTI_CORE_P=1** in the command line when compiling the code. In this mode, all cores will execute the transfers specified in the stimulin in a parallel manner.
+    3. `make run` : this will launch the simulation in bash mode (use gui=1 for Modelsim gui).
+4. All transfers will be executed in the three different directions that are currently supported: **L1->L2, L2->L1, L1->L1**.
+
+Updated drivers for the iDMA can be found at **pulp_cluster/pulp-runtime/include/hal/dma/idma_v2.h** and **pulp_cluster/pulp-runtime/include/archi/dma/idma_v2.h**. 
diff --git a/idma_tests/idma_2D_FC_TCDM/Makefile b/idma_tests/idma_2D_FC_TCDM/Makefile
new file mode 100755
index 0000000..8a5210c
--- /dev/null
+++ b/idma_tests/idma_2D_FC_TCDM/Makefile
@@ -0,0 +1,22 @@
+PULP_APP = test
+
+TEST_SRCS ?= idma_2D_FC_TCDM.c
+PULP_APP_SRCS = $(TEST_SRCS)
+PULP_APP_FC_SRCS = $(TEST_FC_SRCS)
+ifdef TEST_FC_SRCS
+pulpFc=1
+endif
+
+space :=
+space +=
+
+#BUILD_DIR = $(subst $(space),_,$(CURDIR)/build/$(TEST_SRCS))
+
+ifdef VERBOSE
+PULP_CFLAGS += -DVERBOSE
+endif
+
+PULP_CFLAGS += -O3
+stackSize = 4096
+
+include $(PULP_SDK_HOME)/install/rules/pulp.mk
diff --git a/idma_tests/idma_2D_FC_TCDM/idma_2D_FC_TCDM.c b/idma_tests/idma_2D_FC_TCDM/idma_2D_FC_TCDM.c
new file mode 100644
index 0000000..8e41e26
--- /dev/null
+++ b/idma_tests/idma_2D_FC_TCDM/idma_2D_FC_TCDM.c
@@ -0,0 +1,101 @@
+#include <math.h>
+#include <stdint.h>
+#include <stdio.h>
+#include "pulp.h"
+
+// #define VERBOSE
+
+#define MAX_BUFFER_SIZE 0x2000
+
+L2_DATA static uint8_t ext[MAX_BUFFER_SIZE];
+L1_DATA static uint8_t loc[MAX_BUFFER_SIZE];
+
+// Transfer types
+typedef enum {L2_TO_L1, L1_TO_L2} test_type_t;
+
+#ifdef VERBOSE
+static void print_memory(uint8_t *mem, unsigned int size, const char *name) {
+    printf("Memory dump of %s:\n", name);
+    for (unsigned int i = 0; i < size; i++) {
+        printf("0x%02X ", mem[i]);
+        if ((i + 1) % 16 == 0) {
+            printf("\n");
+        }
+    }
+    printf("\n");
+}
+#endif
+
+int test_idma_2D(uint32_t size, test_type_t type, uint32_t ext_addr, uint32_t tcdm_addr,
+                 unsigned int length, unsigned int src_stride, unsigned int dst_stride, unsigned int num_reps);
+
+int main(void) {
+    if (rt_cluster_id() != 0) {
+        return bench_cluster_forward(0);
+    }
+
+    int error_count = 0;
+
+    if (get_core_id() == 0) {
+        unsigned int sizes[] = {16, 64, 256, 1024};     // Total number of bytes (for initialization of the values in memory)
+        unsigned int lengths[] = {4, 8, 16, 64};        // Number of bytes to transfer per 1D transaction
+        unsigned int num_reps[] = {4, 8, 16, 16};       // Number of repetitions of 1D transactions
+        unsigned int src_strides[] = {4, 8, 24, 56};    // Number of strides in source memory
+        unsigned int dst_strides[] = {4, 12, 24, 64};   // Number of strides in destination memory
+
+        for (unsigned int i = 0; i < 4; i++) {
+            error_count += test_idma_2D(sizes[i], L2_TO_L1, (uint32_t)ext, (uint32_t)loc,
+                                        lengths[i], src_strides[i], dst_strides[i], num_reps[i]);
+        }
+
+        for (unsigned int i = 0; i < 4; i++) {
+            error_count += test_idma_2D(sizes[i], L1_TO_L2, (uint32_t)ext, (uint32_t)loc,
+                                        lengths[i], src_strides[i], dst_strides[i], num_reps[i]);
+        }
+    }
+
+    return error_count;
+}
+
+int test_idma_2D(uint32_t size, test_type_t type, uint32_t ext_addr, uint32_t tcdm_addr,
+                 unsigned int length, unsigned int src_stride, unsigned int dst_stride, unsigned int num_reps) {
+    volatile uint8_t *src_ptr, *dst_ptr;
+    volatile int error = 0;
+    volatile unsigned int id;
+
+    if (type == L2_TO_L1) {
+        src_ptr = (uint8_t *)ext_addr;
+        dst_ptr = (uint8_t *)tcdm_addr;
+        for (uint32_t i = 0; i < size; i++) src_ptr[i] = (uint8_t)(i & 0xFF);
+        id = pulp_cl_idma_L2ToL1_2d((unsigned int)src_ptr, (unsigned int)dst_ptr, length, src_stride, dst_stride, num_reps);
+        plp_cl_dma_wait_toL1(id);
+    } else {
+        src_ptr = (uint8_t *)tcdm_addr;
+        dst_ptr = (uint8_t *)ext_addr;
+        for (uint32_t i = 0; i < size; i++) src_ptr[i] = (uint8_t)(i & 0xFF);
+        id = pulp_cl_idma_L1ToL2_2d((unsigned int)src_ptr, (unsigned int)dst_ptr, length, src_stride, dst_stride, num_reps);
+        plp_cl_dma_wait_toL2(id);
+    }
+
+    // plp_dma_barrier();
+
+    for (unsigned int rep = 0; rep < num_reps; rep++) {
+        unsigned int src_offset = rep * src_stride;
+        unsigned int dst_offset = rep * dst_stride;
+        for (unsigned int i = 0; i < length; i++) {
+            uint8_t expected = src_ptr[src_offset + i];
+            uint8_t actual = dst_ptr[dst_offset + i];
+            if (expected != actual) error++;
+        }
+    }
+
+#ifdef VERBOSE
+    if (error == 0) {
+        printf("Test passed for %s with size %d, length %d, src_stride %d, dst_stride %d, num_reps %d\n", type == L2_TO_L1 ? "L2_TO_L1" : "L1_TO_L2", size, length, src_stride, dst_stride, num_reps);
+    } else {
+        printf("Test failed for %s with size %d, length %d, src_stride %d, dst_stride %d, num_reps %d\n", type == L2_TO_L1 ? "L2_TO_L1" : "L1_TO_L2", size, length, src_stride, dst_stride, num_reps);
+    }
+#endif
+
+    return error;
+}
\ No newline at end of file
diff --git a/idma_tests/idma_3D_FC_TCDM/Makefile b/idma_tests/idma_3D_FC_TCDM/Makefile
new file mode 100755
index 0000000..02ba50d
--- /dev/null
+++ b/idma_tests/idma_3D_FC_TCDM/Makefile
@@ -0,0 +1,22 @@
+PULP_APP = test
+
+TEST_SRCS ?= idma_3D_FC_TCDM.c
+PULP_APP_SRCS = $(TEST_SRCS)
+PULP_APP_FC_SRCS = $(TEST_FC_SRCS)
+ifdef TEST_FC_SRCS
+pulpFc=1
+endif
+
+space :=
+space +=
+
+#BUILD_DIR = $(subst $(space),_,$(CURDIR)/build/$(TEST_SRCS))
+
+ifdef VERBOSE
+PULP_CFLAGS += -DVERBOSE
+endif
+
+PULP_CFLAGS += -O3
+stackSize = 4096
+
+include $(PULP_SDK_HOME)/install/rules/pulp.mk
diff --git a/idma_tests/idma_3D_FC_TCDM/idma_3D_FC_TCDM.c b/idma_tests/idma_3D_FC_TCDM/idma_3D_FC_TCDM.c
new file mode 100644
index 0000000..ae5131b
--- /dev/null
+++ b/idma_tests/idma_3D_FC_TCDM/idma_3D_FC_TCDM.c
@@ -0,0 +1,166 @@
+#include <math.h>
+#include <stdint.h>
+#include <stdio.h>
+#include "pulp.h"
+
+// #define VERBOSE
+
+#define MAX_BUFFER_SIZE 0x2000
+
+L2_DATA static uint8_t ext[MAX_BUFFER_SIZE];
+L1_DATA static uint8_t loc[MAX_BUFFER_SIZE];
+
+// Transfer types
+typedef enum {L2_TO_L1, L1_TO_L2} test_type_t;
+
+#ifdef VERBOSE
+static void print_memory(uint8_t *mem, unsigned int size, const char *name) {
+    printf("Memory dump of %s:\n", name);
+    for (unsigned int i = 0; i < size; i++) {
+        printf("0x%02X ", mem[i]);
+        if ((i + 1) % 16 == 0) {
+            printf("\n");
+        }
+    }
+    printf("\n");
+}
+#endif
+
+int test_idma_3D(uint32_t size, test_type_t type, uint32_t ext_addr, uint32_t tcdm_addr,
+                 unsigned int length, unsigned int src_stride_2d, unsigned int dst_stride_2d, unsigned int num_reps,
+                 unsigned int src_stride_3d, unsigned int dst_stride_3d, unsigned int num_reps_3d);
+
+int main(void) {
+    if (rt_cluster_id() != 0) {
+        return bench_cluster_forward(0);
+    }
+
+    int error_count = 0;
+
+    if (get_core_id() == 0) {
+        unsigned int sizes[] = {64, 256, 1024, 2048};     // Total number of bytes
+        unsigned int lengths[] = {4, 8, 16, 32};          // Number of bytes per 1D transaction
+        unsigned int src_strides_2d[] = {8, 16, 32, 64};  // 2D stride in source
+        unsigned int dst_strides_2d[] = {8, 16, 32, 64};  // 2D stride in destination
+        unsigned int num_reps_3d[] = {2, 2, 2, 2};        // Number of 2D pages
+        unsigned int src_strides_3d[] = {32, 128, 512, 1024}; // 3D stride in source
+        unsigned int dst_strides_3d[] = {32, 128, 512, 1024}; // 3D stride in destination
+
+        for (unsigned int i = 0; i < 4; i++) {
+            unsigned int num_reps = sizes[i] / lengths[i]; // Calculate num_reps like in multi_core test
+            
+            printf("Transfer: %d\n", i);
+            printf("Size: %d | Length: %d | Src_stride_2d: %d | Dst_stride_2d: %d | Num_reps_2d: %d\n", 
+                   sizes[i], lengths[i], src_strides_2d[i], dst_strides_2d[i], num_reps);
+            printf("Src_stride_3d: %d | Dst_stride_3d: %d | Num_reps_3d: %d\n", 
+                   src_strides_3d[i], dst_strides_3d[i], num_reps_3d[i]);
+            
+            printf("L2 to L1 transfer\n");
+            error_count += test_idma_3D(sizes[i], L2_TO_L1, (uint32_t)ext, (uint32_t)loc,
+                                        lengths[i], src_strides_2d[i], dst_strides_2d[i], num_reps,
+                                        src_strides_3d[i], dst_strides_3d[i], num_reps_3d[i]);
+
+            printf("L1 to L2 transfer\n");
+            error_count += test_idma_3D(sizes[i], L1_TO_L2, (uint32_t)ext, (uint32_t)loc,
+                                        lengths[i], src_strides_2d[i], dst_strides_2d[i], num_reps,
+                                        src_strides_3d[i], dst_strides_3d[i], num_reps_3d[i]);
+        }
+    }
+
+    return error_count;
+}
+
+int test_idma_3D(uint32_t size, test_type_t type, uint32_t ext_addr, uint32_t tcdm_addr,
+                 unsigned int length, unsigned int src_stride_2d, unsigned int dst_stride_2d, unsigned int num_reps,
+                 unsigned int src_stride_3d, unsigned int dst_stride_3d, unsigned int num_reps_3d) {
+    volatile uint8_t *src_ptr, *dst_ptr;
+    unsigned int offset_2d, offset_3d;
+    int error = 0;
+
+    if (type == L2_TO_L1) {
+        // L2 to L1 transfer
+        src_ptr = (uint8_t*)ext_addr;
+        dst_ptr = (uint8_t*)tcdm_addr;
+
+        // Fill source region with test data
+        offset_2d = 0;
+        offset_3d = 0;
+        for (int j = 0; j < num_reps_3d; j++) {
+            for (int q = 0; q < num_reps; q++) {
+                for (int i = 0; i < length; i++) {
+                    src_ptr[i + offset_2d + offset_3d] = (uint8_t)(i & 0xFF);
+                }
+                offset_2d += src_stride_2d;
+            }
+            offset_2d = 0;
+            offset_3d += (num_reps - 1) * src_stride_2d + src_stride_3d;
+        }
+
+        plp_cl_dma_wait_toL1(pulp_cl_idma_L2ToL1_3d((unsigned int)src_ptr, (unsigned int)dst_ptr, 
+                                                     length, src_stride_2d, dst_stride_2d, num_reps, 
+                                                     src_stride_3d, dst_stride_3d, num_reps_3d));
+    } else {
+        // L1 to L2 transfer
+        src_ptr = (uint8_t*)tcdm_addr;
+        dst_ptr = (uint8_t*)ext_addr;
+
+        // Fill source region with test data
+        offset_2d = 0;
+        offset_3d = 0;
+        for (int j = 0; j < num_reps_3d; j++) {
+            for (int q = 0; q < num_reps; q++) {
+                for (int i = 0; i < length; i++) {
+                    src_ptr[i + offset_2d + offset_3d] = (uint8_t)(i & 0xFF);
+                }
+                offset_2d += src_stride_2d;
+            }
+            offset_2d = 0;
+            offset_3d += (num_reps - 1) * src_stride_2d + src_stride_3d;
+        }
+
+        plp_cl_dma_wait_toL2(pulp_cl_idma_L1ToL2_3d((unsigned int)src_ptr, (unsigned int)dst_ptr, 
+                                                     length, src_stride_2d, dst_stride_2d, num_reps, 
+                                                     src_stride_3d, dst_stride_3d, num_reps_3d));
+    }
+
+    // Check the results - exactly like multi_core test
+    int src_offset_2d = 0;
+    int dst_offset_2d = 0;
+    int src_offset_3d = 0;
+    int dst_offset_3d = 0;
+
+    for (int rep_3d = 0; rep_3d < num_reps_3d; rep_3d++) {
+        for (unsigned int rep = 0; rep < num_reps; rep++) {
+            for (unsigned int i = 0; i < length; i++) {
+                uint8_t expected = src_ptr[src_offset_2d + src_offset_3d + i];
+                uint8_t actual = dst_ptr[dst_offset_2d + dst_offset_3d + i];
+
+                if (expected != actual) {
+#ifdef VERBOSE
+                    printf("ERROR: expected @%8x[%d] = %8x vs actual @%8x[%d] = %8x\n", 
+                           &src_ptr[src_offset_2d + src_offset_3d + i], src_offset_2d + src_offset_3d + i, expected, 
+                           &dst_ptr[dst_offset_2d + dst_offset_3d + i], dst_offset_2d + dst_offset_3d + i, actual);
+#endif
+                    error++;
+                }
+            }
+            src_offset_2d += src_stride_2d;
+            dst_offset_2d += dst_stride_2d;
+        }
+        src_offset_2d = 0;
+        dst_offset_2d = 0;
+        src_offset_3d += (num_reps - 1) * src_stride_2d + src_stride_3d;
+        dst_offset_3d += (num_reps - 1) * dst_stride_2d + dst_stride_3d;
+    }
+
+#ifdef VERBOSE
+    if (error == 0) {
+        printf("Test passed for %s 3D transfer\n", type == L2_TO_L1 ? "L2_TO_L1" : "L1_TO_L2");
+    } else {
+        printf("Test failed for %s 3D transfer with %d errors\n", 
+               type == L2_TO_L1 ? "L2_TO_L1" : "L1_TO_L2", error);
+    }
+#endif
+
+    return error;
+}
\ No newline at end of file
diff --git a/idma_tests/idma_TCDM_to_TCDM/Makefile b/idma_tests/idma_TCDM_to_TCDM/Makefile
new file mode 100755
index 0000000..4249b6f
--- /dev/null
+++ b/idma_tests/idma_TCDM_to_TCDM/Makefile
@@ -0,0 +1,22 @@
+PULP_APP = test
+
+TEST_SRCS ?= idma_TCDM_to_TCDM.c
+PULP_APP_SRCS = $(TEST_SRCS)
+PULP_APP_FC_SRCS = $(TEST_FC_SRCS)
+ifdef TEST_FC_SRCS
+pulpFc=1
+endif
+
+space :=
+space +=
+
+#BUILD_DIR = $(subst $(space),_,$(CURDIR)/build/$(TEST_SRCS))
+
+ifdef VERBOSE
+PULP_CFLAGS += -DVERBOSE
+endif
+
+PULP_CFLAGS += -O3
+stackSize = 4096
+
+include $(PULP_SDK_HOME)/install/rules/pulp.mk
diff --git a/idma_tests/idma_TCDM_to_TCDM/idma_TCDM_to_TCDM.c b/idma_tests/idma_TCDM_to_TCDM/idma_TCDM_to_TCDM.c
new file mode 100644
index 0000000..ae94f01
--- /dev/null
+++ b/idma_tests/idma_TCDM_to_TCDM/idma_TCDM_to_TCDM.c
@@ -0,0 +1,73 @@
+
+#include <math.h>
+#include <stdint.h>
+#include <stdio.h>
+#include "pulp.h"
+
+#define VERBOSE
+
+#define MAX_BUFFER_SIZE 0x2000
+#define BUFFER_SIZE     0x0200
+
+L1_DATA static uint8_t src_buffer[MAX_BUFFER_SIZE];
+L1_DATA static uint8_t dst_buffer[MAX_BUFFER_SIZE];
+
+int test_idma_l1_to_l1(unsigned int len);
+
+int main(void) {
+    if (rt_cluster_id() != 0) {
+        return bench_cluster_forward(0);
+    }
+
+    int error_count = 0;
+
+    if (get_core_id() == 0) {
+        // Run the test for different buffer sizes
+        for (unsigned int len = 8; len <= BUFFER_SIZE; len *= 2) {
+            error_count += test_idma_l1_to_l1(len);
+        }
+    }
+
+    return error_count;
+}
+
+int test_idma_l1_to_l1(unsigned int len) {
+    volatile unsigned int id;
+    volatile int error = 0;
+    volatile uint32_t test, read;
+
+    printf("STARTING L1 TO L1 DMA TEST FOR LENGTH: %d\n", len);
+
+    // Fill source buffer with known data pattern
+    for (unsigned int i = 0; i < len / 4; i++) {
+        ((uint32_t *)src_buffer)[i] = 0xABCD0000 + i;
+    }
+
+    // Clear destination buffer
+    memset(dst_buffer, 0, len);
+
+    // Perform DMA transfer from src_buffer to dst_buffer within L1 memory
+    id = pulp_cl_idma_L1ToL1((unsigned int)src_buffer, (unsigned int)dst_buffer, len);
+
+    // Wait for DMA transfer to complete
+    plp_cl_dma_barrier();
+
+    // Verify data in destination buffer
+    for (unsigned int i = 0; i < len / 4; i++) {
+        test = 0xABCD0000 + i;
+        read = ((uint32_t *)dst_buffer)[i];
+
+        if (test != read) {
+            printf("Error!!! Read: %x, Expected: %x, Index: %d\n", read, test, i);
+            error++;
+        }
+    }
+
+    if (error == 0) {
+        printf("L1 to L1 DMA transfer test passed for length %d.\n", len);
+    } else {
+        printf("L1 to L1 DMA transfer test failed for length %d with %d errors.\n", len, error);
+    }
+
+    return error;
+}
diff --git a/idma_tests/idma_allcores/Makefile b/idma_tests/idma_allcores/Makefile
new file mode 100755
index 0000000..ea1e235
--- /dev/null
+++ b/idma_tests/idma_allcores/Makefile
@@ -0,0 +1,22 @@
+PULP_APP = test
+
+TEST_SRCS ?= allcores.c
+PULP_APP_SRCS = $(TEST_SRCS)
+PULP_APP_FC_SRCS = $(TEST_FC_SRCS)
+ifdef TEST_FC_SRCS
+pulpFc=1
+endif
+
+space :=
+space +=
+
+#BUILD_DIR = $(subst $(space),_,$(CURDIR)/build/$(TEST_SRCS))
+
+ifdef VERBOSE
+PULP_CFLAGS += -DVERBOSE
+endif
+
+PULP_CFLAGS += -O3
+stackSize = 4096
+
+include $(PULP_SDK_HOME)/install/rules/pulp.mk
diff --git a/idma_tests/idma_allcores/allcores.c b/idma_tests/idma_allcores/allcores.c
new file mode 100644
index 0000000..f833dd5
--- /dev/null
+++ b/idma_tests/idma_allcores/allcores.c
@@ -0,0 +1,112 @@
+#include <math.h>
+#include "pulp.h"
+//#include "mchan_tests.h"
+
+#define VERBOSE
+
+#define MAX_BUFFER_SIZE 0x2200
+
+//static unsigned char *ext;
+//static unsigned char *loc;
+
+//#define EXT_DATA_ADDR  ((unsigned int)ext)
+//#define TCDM_DATA_ADDR ((unsigned int)loc)
+
+L2_DATA static unsigned char ext[MAX_BUFFER_SIZE];
+L1_DATA static unsigned char loc[MAX_BUFFER_SIZE];
+L1_DATA static int error_count = 0;
+#define EXT_DATA_ADDR  ((unsigned int) ext)
+#define TCDM_DATA_ADDR ((unsigned int) loc)
+typedef enum {RX, TX, TCDM} test_type_t;
+
+int test_idma_1d(unsigned int len, test_type_t type, unsigned int ext_addr, unsigned int tcdm_addr);
+int main()
+{
+
+  if (rt_cluster_id() != 0)
+    return bench_cluster_forward(0);
+
+  int n_bytes [] = {5, 19, 400};
+
+  for (int core=0; core<ARCHI_CLUSTER_NB_PE; core++) {
+  if (get_core_id() == core){
+    printf("Testing iDMA with cluster core %d\n", core);
+    for (int i = 0; i<3; i++) {
+      error_count += test_idma_1d(n_bytes[i], TX, loc, ext);
+    }
+    for (int i = 0; i<3; i++) {
+      error_count += test_idma_1d(n_bytes[i], RX, ext, loc);
+    }
+    for (int i = 0; i<3; i++) {
+      error_count += test_idma_1d(n_bytes[i], TCDM, loc, loc+MAX_BUFFER_SIZE/2);
+    }
+
+  }
+  synch_barrier();
+  }
+  return error_count;
+}
+
+int test_idma_1d(unsigned int len, test_type_t type, unsigned int src, unsigned int dst){
+
+  volatile unsigned int i,j,id;
+  volatile unsigned int test,read,error=0;
+
+  for (i=0; i<len; i++){
+    *(unsigned char*)(src + i) = i & 0xFF;
+  }
+
+  for (i=0; i<len+16; i++){
+    *(unsigned char*)(dst + i) = 0;
+  }
+  if (type == RX){
+
+#ifdef VERBOSE
+    printf ("STARTING TEST FOR RX %d OPERATION: \n", len);
+#endif
+
+    id = pulp_cl_idma_memcpy(src, dst, len, IDMA_PROT_AXI, IDMA_PROT_OBI);
+  } else if (type == TX) {
+
+#ifdef VERBOSE
+    printf ("STARTING TEST FOR TX %d OPERATION: \n", len);
+#endif
+    id = pulp_cl_idma_memcpy(src, dst, len, IDMA_PROT_OBI, IDMA_PROT_AXI);
+  } else {
+#ifdef VERBOSE
+    printf ("STARTING TEST FOR TCDM %d OPERATION: \n", len);
+#endif
+    id = pulp_cl_idma_memcpy(src, dst, len, IDMA_PROT_OBI, IDMA_PROT_OBI);
+  }
+
+  plp_dma_barrier();
+
+    for (i=0; i<len; i++){
+
+      test = i & 0xFF;
+      read = *(unsigned char*)(dst + i);
+
+      if ( test != read ){
+  printf("Error!!! Read: %x, Test:%x, Index: %d \n ",read,test,i);
+  error++;
+      }
+    }
+      for (i=len; i<len+16; i++){
+
+        test = 0;
+        read = *(unsigned char*)(dst + i);
+
+        if ( test != read ){
+          printf("Error!!! Read: %x, Test:%x, Index: %d \n ",read,test,i);
+          error++;
+        }
+    }
+
+
+  if (error == 0)
+    printf("OOOOOOK!!!!!!\n");
+  else
+    printf("NOT OK!!!!!\n");
+
+  return error;
+}
diff --git a/idma_tests/idma_allcores/dma_wave.do b/idma_tests/idma_allcores/dma_wave.do
new file mode 100644
index 0000000..ab54467
--- /dev/null
+++ b/idma_tests/idma_allcores/dma_wave.do
@@ -0,0 +1,179 @@
+onerror {resume}
+quietly WaveActivateNextPane {} 0
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/clk_i
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/rst_ni
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/test_mode_i
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/ext_master_req_o
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/ext_master_resp_i
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/term_event_o
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/term_irq_o
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/term_event_pe_o
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/term_irq_pe_o
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/busy_o
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/config_wdata
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/config_add
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/config_req
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/config_wen
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/config_be
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/config_id
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/config_gnt
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/config_r_rdata
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/config_r_valid
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/config_r_opc
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/config_r_id
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/obi_read_req_from_dma
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/obi_read_req_from_rrc
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/obi_reorg_req_from_dma
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/obi_reorg_req_from_rrc
+add wave -noupdate -expand -group {iDMA wrap} -expand -subitemconfig {{/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/obi_write_req_from_dma[0]} -expand} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/obi_write_req_from_dma
+add wave -noupdate -expand -group {iDMA wrap} -expand -subitemconfig {{/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/obi_write_req_from_rrc[0]} -expand {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/obi_write_req_from_rrc[0].a} -expand} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/obi_write_req_from_rrc
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/obi_read_req_muxed
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/obi_read_rsp_to_dma
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/obi_read_rsp_to_rrc
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/obi_reorg_rsp_to_dma
+add wave -noupdate -expand -group {iDMA wrap} -subitemconfig {{/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/obi_reorg_rsp_to_rrc[0]} -expand {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/obi_reorg_rsp_to_rrc[0].r} -expand} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/obi_reorg_rsp_to_rrc
+add wave -noupdate -expand -group {iDMA wrap} -expand -subitemconfig {{/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/obi_write_rsp_to_dma[0]} -expand {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/obi_write_rsp_to_dma[0].r} -expand} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/obi_write_rsp_to_dma
+add wave -noupdate -expand -group {iDMA wrap} -expand -subitemconfig {{/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/obi_write_rsp_to_rrc[0]} -expand {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/obi_write_rsp_to_rrc[0].r} -expand} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/obi_write_rsp_to_rrc
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/obi_read_rsp_to_mux
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/soc_req
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/soc_rsp
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/dma_req
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/dma_rsp
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/dma_regs_req
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/dma_regs_rsp
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/stream_idx
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/twod_req
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/twod_req_queue
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/idma_req
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/idma_rsp
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/one_fe_valid
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/fe_valid
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/twod_queue_valid
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/be_valid
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/be_rsp_valid
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/fe_ready
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/twod_queue_ready
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/be_ready
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/be_rsp_ready
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/trans_complete
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/midend_busy
+add wave -noupdate -expand -group {iDMA wrap} -expand /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/idma_busy
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/done_id
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/next_id
+add wave -noupdate -expand -group {init strm1} -subitemconfig {{/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/init_write_rsp.rsp_chan} -expand} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/init_write_rsp}
+add wave -noupdate -expand -group {init strm1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/init_write_req}
+add wave -noupdate -expand -group {init strm1} -expand {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/init_read_rsp}
+add wave -noupdate -expand -group {init strm1} -expand {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/init_read_req}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/clk_i}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/rst_ni}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/testmode_i}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/axi_read_req_o}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/axi_read_rsp_i}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/init_read_req_o}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/init_read_rsp_i}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/obi_read_req_o}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/obi_read_rsp_i}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/init_write_req_o}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/init_write_rsp_i}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/obi_write_req_o}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/obi_write_rsp_i}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/r_dp_req_i}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/r_dp_valid_i}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/r_dp_ready_o}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/r_dp_rsp_o}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/r_dp_valid_o}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/r_dp_ready_i}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_dp_req_i}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_dp_valid_i}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_dp_ready_o}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_dp_rsp_o}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_dp_valid_o}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_dp_ready_i}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/ar_req_i}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/ar_valid_i}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/ar_ready_o}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/aw_req_i}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/aw_valid_i}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/aw_ready_o}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/dp_poison_i}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/r_chan_ready_o}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/r_chan_valid_o}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/r_dp_busy_o}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_dp_busy_o}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/buffer_busy_o}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/axi_buffer_in_valid}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/init_buffer_in_valid}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/obi_buffer_in_valid}
+add wave -noupdate -expand -group {to OBI transport layer} -expand {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/buffer_in_valid}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/buffer_in_ready}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/buffer_out_valid}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/buffer_out_valid_shifted}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/init_buffer_out_ready}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/obi_buffer_out_ready}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/buffer_out_ready}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/buffer_out_ready_shifted}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/axi_buffer_in}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/init_buffer_in}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/obi_buffer_in}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/buffer_in}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/buffer_in_shifted}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/buffer_out}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/buffer_out_shifted}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/axi_r_chan_valid}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/init_r_chan_valid}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/obi_r_chan_valid}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/axi_r_chan_ready}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/init_r_chan_ready}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/obi_r_chan_ready}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/axi_r_dp_valid}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/init_r_dp_valid}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/obi_r_dp_valid}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/axi_r_dp_ready}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/init_r_dp_ready}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/obi_r_dp_ready}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/axi_r_dp_rsp}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/init_r_dp_rsp}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/obi_r_dp_rsp}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/axi_ar_ready}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/init_ar_ready}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/obi_ar_ready}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/init_w_dp_rsp_valid}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/obi_w_dp_rsp_valid}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/init_w_dp_rsp_ready}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/obi_w_dp_rsp_ready}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/init_w_dp_ready}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/obi_w_dp_ready}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/init_w_dp_rsp}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/obi_w_dp_rsp}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/init_aw_ready}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/obi_aw_ready}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_dp_req_valid}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_dp_req_ready}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_dp_rsp_mux_valid}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_dp_rsp_mux_ready}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_dp_rsp_valid}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_dp_rsp_ready}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_dp_rsp_mux}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_resp_fifo_in_valid}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_resp_fifo_in_ready}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_resp_fifo_out_protocol}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_resp_fifo_out_valid}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_resp_fifo_out_ready}
+TreeUpdate [SetDefaultTree]
+WaveRestoreCursors {{Cursor 1} {226792000000 ps} 0}
+quietly wave cursor active 1
+configure wave -namecolwidth 252
+configure wave -valuecolwidth 120
+configure wave -justifyvalue left
+configure wave -signalnamewidth 1
+configure wave -snapdistance 10
+configure wave -datasetprefix 0
+configure wave -rowmargin 4
+configure wave -childrowmargin 2
+configure wave -gridoffset 0
+configure wave -gridperiod 1
+configure wave -griddelta 40
+configure wave -timeline 0
+configure wave -timelineunits ns
+update
+WaveRestoreZoom {226663267749 ps} {227016368236 ps}
diff --git a/idma_tests/idma_basic_FC_TCDM/Makefile b/idma_tests/idma_basic_FC_TCDM/Makefile
new file mode 100755
index 0000000..563856b
--- /dev/null
+++ b/idma_tests/idma_basic_FC_TCDM/Makefile
@@ -0,0 +1,22 @@
+PULP_APP = test
+
+TEST_SRCS ?= idma_basic_FC_TCDM.c
+PULP_APP_SRCS = $(TEST_SRCS)
+PULP_APP_FC_SRCS = $(TEST_FC_SRCS)
+ifdef TEST_FC_SRCS
+pulpFc=1
+endif
+
+space :=
+space +=
+
+#BUILD_DIR = $(subst $(space),_,$(CURDIR)/build/$(TEST_SRCS))
+
+ifdef VERBOSE
+PULP_CFLAGS += -DVERBOSE
+endif
+
+PULP_CFLAGS += -O3
+stackSize = 4096
+
+include $(PULP_SDK_HOME)/install/rules/pulp.mk
diff --git a/idma_tests/idma_basic_FC_TCDM/idma_basic_FC_TCDM.c b/idma_tests/idma_basic_FC_TCDM/idma_basic_FC_TCDM.c
new file mode 100644
index 0000000..20b9063
--- /dev/null
+++ b/idma_tests/idma_basic_FC_TCDM/idma_basic_FC_TCDM.c
@@ -0,0 +1,124 @@
+#include <math.h>
+#include <stdint.h>
+#include <inttypes.h>
+#include <stdio.h>
+#include "pulp.h"
+
+#define VERBOSE
+
+#define MAX_BUFFER_SIZE 0x2000 
+
+L2_DATA static uint8_t ext[MAX_BUFFER_SIZE]; // External memory buffer (L2)
+L1_DATA static uint8_t loc[MAX_BUFFER_SIZE]; // Local memory buffer (TCDM / L1)
+
+typedef enum {L2_TO_L1, L1_TO_L2} test_type_t;
+
+int test_idma(uint32_t size, test_type_t type, uint32_t ext_addr, uint32_t tcdm_addr);
+
+int main(void) {
+    if (rt_cluster_id() != 0) {
+        printf("Test only runs on cluster 0\n");
+        return bench_cluster_forward(0);
+    }
+    
+    int error_count = 0;
+
+    if (get_core_id() == 0) {
+
+        uint32_t size = 5;
+        // Test for L2_TO_L1 operation
+        for (int i = 0; i < size; i++) {
+            error_count += test_idma(size, L2_TO_L1, (uint32_t)ext, (uint32_t)loc);
+        }
+
+        // Test for L1_TO_L2 operation
+        for (int i = 0; i < size; i++) {
+            error_count += test_idma(size, L1_TO_L2, (uint32_t)ext, (uint32_t)loc);
+        }
+    }
+
+    return error_count;
+}
+
+int test_idma(uint32_t size, test_type_t type, uint32_t ext_addr, uint32_t tcdm_addr) {
+    volatile uint8_t expected, actual;
+    volatile int error = 0;
+    volatile unsigned int id;
+
+    if (type == L2_TO_L1) {
+
+        for (uint32_t i = 0; i < size; i++) {
+            *(uint8_t *)(ext_addr + i) = (uint8_t)(i & 0xFF);
+        }
+
+        for (uint32_t i = 0; i < size; i++) {
+            *(uint8_t *)(tcdm_addr + i) = 0;
+        }
+
+        // memset((void *)tcdm_addr, 0, size + 16);
+
+        id = pulp_idma_memcpy(ext_addr, tcdm_addr, size, IDMA_PROT_AXI, IDMA_PROT_OBI);
+    
+        /* L2_TO_L1 transaction:
+        +----------------+                        +----------------+
+        |                |       DMA Transfer     |                |
+        |  L2 Memory     | ---------------------> |  L1 Memory     |
+        |  (ext buffer)  |   (pulp_idma_memcpy)   |  (loc buffer)  |
+        |                |                        |                |
+        +----------------+                        +----------------+
+        */
+       plp_cl_dma_wait_toL1(id);
+    } else if (type == L1_TO_L2) {
+        // Fill L1 buffer with a pattern
+        for (uint32_t i = 0; i < size; i++) {
+            *(uint8_t *)(tcdm_addr + i) = (uint8_t)(i & 0xFF);
+        }
+
+        // Clear L2
+        memset((void *)ext_addr, 0, size + 16);
+
+        // Perform DMA transfer from local memory to external memory
+        id = pulp_idma_memcpy(tcdm_addr, ext_addr, size, IDMA_PROT_OBI, IDMA_PROT_AXI);
+
+        /* L1_TO_L2 transaction:
+        +----------------+                        +----------------+
+        |                |       DMA Transfer     |                |
+        |  L2 Memory     | <--------------------- |  L1 Memory     |
+        |  (ext buffer)  |   (pulp_idma_memcpy)   |  (loc buffer)  |
+        |                |                        |                |
+        +----------------+                        +----------------+
+        */
+        plp_cl_dma_wait_toL2(id);
+    } else {
+        printf("Invalid test type. It must be either L2_TO_L1 or L1_TO_L2.\n");
+        return 1;
+    }
+
+    // plp_cl_dma_wait();
+    // plp_dma_barrier();
+
+    // Verify data
+    for (uint32_t i = 0; i < size; i++) {
+        expected = (uint8_t)(i & 0xFF); // Same pattern as L2_TO_L1
+        if (type == L2_TO_L1) {
+            // Reading from the local memory buffer "loc" allocated in L1 memory
+            actual = *(uint8_t *)(tcdm_addr + i);
+        } else { // L1_TO_L2
+            // Reading from the external memory buffer "ext" allocated in L2 memory
+            actual = *(uint8_t *)(ext_addr + i);
+        } 
+
+        if (expected != actual) {
+            printf("Error at index %u: Expected 0x%02X, Got 0x%02X\n", i, expected, actual);
+            error++;
+        }
+    }
+
+    if (error == 0) {
+        printf("Test passed for %s of length %u.\n", type == L2_TO_L1 ? "L2_TO_L1" : "L1_TO_L2", size);
+    } else {
+        printf("Test failed for %s of length %u with %u errors.\n", type == L2_TO_L1 ? "L2_TO_L1" : "L1_TO_L2", size, error);
+    }
+    
+    return error;
+}
\ No newline at end of file
diff --git a/idma_tests/idma_multi_core/Makefile b/idma_tests/idma_multi_core/Makefile
new file mode 100644
index 0000000..a1092e3
--- /dev/null
+++ b/idma_tests/idma_multi_core/Makefile
@@ -0,0 +1,31 @@
+PULP_APP = test
+TEST_SRCS ?= idma_multi_core.c
+PULP_APP_SRCS = $(TEST_SRCS)
+
+ifdef VERBOSE
+PULP_CFLAGS += -DVERBOSE
+endif
+
+ifdef MULTI_CORE_P
+PULP_CFLAGS += -DMULTI_CORE_P
+endif
+
+ifdef MULTI_CORE_S
+PULP_CFLAGS += -DMULTI_CORE_S
+endif
+
+ifdef SINGLE_CORE
+PULP_CFLAGS += -DSINGLE_CORE
+endif
+
+ifdef QUICK_MODE
+PULP_CFLAGS += -DQUICK_MODE
+endif
+
+PULP_CFLAGS += -O3
+stackSize = 4096
+
+include $(PULP_SDK_HOME)/install/rules/pulp.mk
+
+stimuli:
+	python gen_stimuli_idma.py
diff --git a/idma_tests/idma_multi_core/gen_stimuli_idma.py b/idma_tests/idma_multi_core/gen_stimuli_idma.py
new file mode 100644
index 0000000..53ce234
--- /dev/null
+++ b/idma_tests/idma_multi_core/gen_stimuli_idma.py
@@ -0,0 +1,38 @@
+#!/usr/bin/env python
+
+import sys
+import random
+import argparse
+import math
+import re
+
+
+parser = argparse.ArgumentParser(description='Generate stimuli for iDMA simple transfer')
+
+args = parser.parse_args()
+
+def write_transfer_sizes_array(f, name, arr):
+    f.write ('unsigned int %s[] = {\n' % name)
+    for v in arr:
+        random_int = random.randint(1, TRANSFER_SIZE)
+        f.write('%d, \n' % random_int)
+    f.write('};\n\n')
+    return
+
+def write_define(f, name,val):
+    f.write('#define %s %d\n\n' % (name,val))
+    return
+
+# Randomize the number of transfers to execute
+NB_TRANSFERS = random.randint(1, 10)
+
+# Randomize between 1 and 1024 the size (in bytes) of each transfer (Must be less than CORE_SPACE defined in the C header)
+TRANSFER_SIZE = 1024
+
+sizes = [None] * NB_TRANSFERS
+
+f_sizes = open('idma_parameters.h', 'w')
+f_defines = open ('idma_defines.h', 'w')
+
+write_transfer_sizes_array(f_sizes, 'sizes', sizes)
+write_define(f_defines, 'NB_TRANSFERS', NB_TRANSFERS)
diff --git a/idma_tests/idma_multi_core/idma_defines.h b/idma_tests/idma_multi_core/idma_defines.h
new file mode 100644
index 0000000..b544610
--- /dev/null
+++ b/idma_tests/idma_multi_core/idma_defines.h
@@ -0,0 +1,2 @@
+#define NB_TRANSFERS 7
+
diff --git a/idma_tests/idma_multi_core/idma_multi_core.c b/idma_tests/idma_multi_core/idma_multi_core.c
new file mode 100644
index 0000000..079c250
--- /dev/null
+++ b/idma_tests/idma_multi_core/idma_multi_core.c
@@ -0,0 +1,187 @@
+#include "idma_multi_core.h"
+
+#define TOT_SIZE 8 * CORE_SPACE
+
+int errors[8] = {0};
+int test_status = 8;
+
+uint32_t l1_addr[8] = {0};
+uint32_t l1_dst_addr[8] = {0};
+uint32_t l2_addr[8] = {0};
+
+int test_idma_1D (int core_id, uint32_t size, int ext2loc, int loc2loc) {
+    volatile uint8_t *src_ptr, *dst_ptr;
+
+    int error = 0;
+
+    if (loc2loc == 1) {
+        // L1 to L1 transfer
+        src_ptr = (uint8_t*) l1_addr[core_id];
+        dst_ptr = (uint8_t*) l1_dst_addr[core_id];
+    } else if (ext2loc == 1) {
+        // L2 to L1 transfer
+        src_ptr = (uint8_t*) l2_addr[core_id];
+        dst_ptr = (uint8_t*) l1_addr[core_id];
+    } else {
+        // L1 to L2 transfer
+        src_ptr = (uint8_t*) l1_addr[core_id];
+        dst_ptr = (uint8_t*) l2_addr[core_id];
+    }
+
+    // Fill source region with test data
+    for (int i = 0; i < size; i++) {
+        src_ptr[i] = (uint8_t)(i & 0xFF);
+    }
+
+    if (loc2loc == 1) {
+        plp_cl_dma_wait_toL1(pulp_cl_idma_L1ToL1((unsigned int) src_ptr, (unsigned int) dst_ptr, size));
+    } else if (ext2loc == 1) {
+        plp_cl_dma_wait_toL1(pulp_cl_idma_L2ToL1((unsigned int) src_ptr, (unsigned int) dst_ptr, size));
+    } else {
+        plp_cl_dma_wait_toL2(pulp_cl_idma_L1ToL2((unsigned int) src_ptr, (unsigned int) dst_ptr, size));
+    }
+
+    // Check the results
+
+    for (int i=0; i < size; i++) {
+        uint8_t expected = src_ptr[i]; 
+        uint8_t actual   = dst_ptr[i];
+
+        if (expected != actual) {
+            error++;
+            if (core_id == 0) {
+                PRINTF ("Error: expected @%8x = %8x vs actual @%8x = %8x \n", expected, &src_ptr[i], actual, &dst_ptr[i]);
+            }
+        }
+    }
+
+    return error;
+}
+
+void allocate_mem_to_cores () {
+    int core_id = rt_core_id();
+
+    // Pre-allocate TOT_SIZE = 8 * CORE_SPACE: then we split this window to assign
+    // each core its available space for iDMA transfers
+    // pi_l1_malloc starts allocating from 0x10004008 in L1
+    // pi_l2_malloc starts allocating from 0x1c000a60 in L2
+
+    if (core_id == 0) {
+        l1_addr[0]     = (uint32_t) pi_l1_malloc(0, TOT_SIZE);
+        l1_dst_addr[0] = (uint32_t) pi_l1_malloc(0, TOT_SIZE);
+        l2_addr[0]     = (uint32_t) pi_l2_malloc(TOT_SIZE);
+    }
+
+    // The following synch_barrier is needed so that
+    // no core can assign its address range until the mallocs are executed
+    synch_barrier();
+
+    l1_addr[core_id]     = l1_addr[0] + core_id * CORE_SPACE;
+    l1_dst_addr[core_id] = l1_dst_addr[0] + core_id * CORE_SPACE;
+    l2_addr[core_id]     = l2_addr[0] + core_id * CORE_SPACE;
+
+    // The following synch_barrier is needed so that
+    // no core can start executing until all address ranges have been assigned
+    synch_barrier();
+
+}
+
+void free_allocated_memory () {
+
+    synch_barrier();
+    // Only Core 0 takes care of freeing the allocated memory, since it's the one
+    // allocated it at the beginning of the test
+    if (core_id == 0) {
+        pi_l1_free(0, l1_addr, TOT_SIZE);
+        pi_l1_free(0, l1_dst_addr, TOT_SIZE);
+        pi_l2_free(l2_addr, TOT_SIZE);
+    }
+}
+
+int main () {
+    int core_id = rt_core_id();
+    unsigned int size = 0;
+
+    allocate_mem_to_cores();
+
+    #ifdef MULTI_CORE_P
+        // MULTI CORE PARALLEL MODE: each core uses the iDMA in a parallel manner
+        if (core_id == 0) {
+            PRINTF ("MULTI CORE PARALLEL MODE \n");
+        }
+        for (int k = 0; k < TRANSFERS; k++) {
+            #ifdef QUICK_MODE
+            size = idma_presets[k];
+            #else
+            size = sizes[k];
+            #endif
+            if (core_id == 0){
+                PRINTF ("Size: %d \n", size);
+            }
+            // L1 -> L2
+            errors[core_id] += test_idma_1D(core_id, size, 0, 0);
+            // L2 -> L1
+            errors[core_id] += test_idma_1D(core_id, size, 1, 0);
+            // L1 -> L1
+            errors[core_id] += test_idma_1D(core_id, size, 0, 1);
+        }
+        synch_barrier();
+    #elif MULTI_CORE_S
+        // MULTI CORE SERIAL MODE: each core uses the iDMA in a serial manner
+        if (core_id == 0) {
+            PRINTF ("MULTI CORE SERIAL MODE \n");
+        }
+        for (int i = 0; i < 8; i++) {
+            if (core_id == i) {
+                for (int k = 0; k < TRANSFERS; k++) {
+                    #ifdef QUICK_MODE
+                    size = idma_presets[k];
+                    #else
+                    size = sizes[k];
+                    #endif
+                    if (core_id == 0){
+                        PRINTF ("Size: %d \n", size);
+                    }
+                    // L1 -> L2
+                    errors[core_id] += test_idma_1D(core_id, size, 0, 0);
+                    // L2 -> L1
+                    errors[core_id] += test_idma_1D(core_id, size, 1, 0);
+                    // L1 -> L1
+                    errors[core_id] += test_idma_1D(core_id, size, 0, 1);
+                }
+            }
+        }
+    #else
+        if (core_id == 0) {
+            // SINGLE CORE MODE: just core 0 uses the iDMA
+            PRINTF ("SINGLE CORE MODE: CORE 0 \n");
+            for (int k = 0; k < TRANSFERS; k++) {
+                #ifdef QUICK_MODE
+                size = idma_presets[k];
+                #else
+                size = sizes[k];
+                #endif
+                PRINTF ("Size: %d \n", size);
+                // L1 -> L2
+                errors[core_id] += test_idma_1D(core_id, size, 0, 0);
+                // L2 -> L1
+                errors[core_id] += test_idma_1D(core_id, size, 1, 0);
+                // L1 -> L1
+                errors[core_id] += test_idma_1D(core_id, size, 0, 1);
+            }
+        }
+    #endif
+
+    if (core_id == 0) {
+        for (int i = 0; i<8; i++) {
+            if (errors[i] == 0) {
+                PRINTF ("Core %d returned %d errors \n", i, errors[i]);
+                test_status--;
+            }
+        }
+    }
+
+    free_allocated_memory();
+
+    return test_status;
+}
diff --git a/idma_tests/idma_multi_core/idma_multi_core.h b/idma_tests/idma_multi_core/idma_multi_core.h
new file mode 100644
index 0000000..af672e0
--- /dev/null
+++ b/idma_tests/idma_multi_core/idma_multi_core.h
@@ -0,0 +1,48 @@
+// Copyright 2022 ETH Zurich and University of Bologna.
+// Solderpad Hardware License, Version 0.51, see LICENSE for details.
+// SPDX-License-Identifier: SHL-0.51
+//
+//Author: Andreas Kuster <kustera@ethz.ch>
+//
+//Description: Generated register defines for dma_frontend
+
+#include "pulp.h"
+#include <stdio.h>
+#include <stdlib.h>
+#include <stdint.h>
+#include <string.h>
+#include <assert.h>
+
+#include "idma_defines.h"
+#include "idma_parameters.h"
+#include "idma_presets.h"
+
+#ifndef _DMA_FRONTEND_REG_DEFS_
+#define _DMA_FRONTEND_REG_DEFS_
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#define DMA_CONF_DECOUPLE 0
+#define DMA_CONF_DEBURST 0
+#define DMA_CONF_SERIALIZE 0
+
+#define CORE_SPACE 2048
+
+#if VERBOSE
+    #define PRINTF(...) printf(__VA_ARGS__)
+#else
+    #define PRINTF(...)
+#endif
+
+#ifdef QUICK_MODE
+    #define TRANSFERS 12
+#else
+    #define TRANSFERS NB_TRANSFERS
+#endif
+
+#ifdef __cplusplus
+}  // extern "C"
+#endif
+#endif  // _DMA_FRONTEND_REG_DEFS_
diff --git a/idma_tests/idma_multi_core/idma_parameters.h b/idma_tests/idma_multi_core/idma_parameters.h
new file mode 100644
index 0000000..0388301
--- /dev/null
+++ b/idma_tests/idma_multi_core/idma_parameters.h
@@ -0,0 +1,10 @@
+unsigned int sizes[] = {
+89, 
+947, 
+156, 
+560, 
+278, 
+869, 
+238, 
+};
+
diff --git a/idma_tests/idma_multi_core/idma_presets.h b/idma_tests/idma_multi_core/idma_presets.h
new file mode 100644
index 0000000..3779763
--- /dev/null
+++ b/idma_tests/idma_multi_core/idma_presets.h
@@ -0,0 +1,3 @@
+unsigned int idma_presets [] = {
+    1, 2, 3, 4, 8, 16, 32, 64, 128, 256, 512, 1024
+};
\ No newline at end of file
diff --git a/idma_tests/idma_multi_core_2d/Makefile b/idma_tests/idma_multi_core_2d/Makefile
new file mode 100644
index 0000000..1d6678e
--- /dev/null
+++ b/idma_tests/idma_multi_core_2d/Makefile
@@ -0,0 +1,27 @@
+PULP_APP = test
+TEST_SRCS ?= idma_multi_core_2d.c
+PULP_APP_SRCS = $(TEST_SRCS)
+
+ifdef VERBOSE
+PULP_CFLAGS += -DVERBOSE
+endif
+
+ifdef MULTI_CORE_P
+PULP_CFLAGS += -DMULTI_CORE_P
+endif
+
+ifdef MULTI_CORE_S
+PULP_CFLAGS += -DMULTI_CORE_S
+endif
+
+ifdef QUICK_MODE
+PULP_CFLAGS += -DQUICK_MODE
+endif
+
+PULP_CFLAGS += -O3
+stackSize = 4096
+
+include $(PULP_SDK_HOME)/install/rules/pulp.mk
+
+stimuli:
+	python gen_stimuli_idma_2d.py
\ No newline at end of file
diff --git a/idma_tests/idma_multi_core_2d/gen_stimuli_idma_2d.py b/idma_tests/idma_multi_core_2d/gen_stimuli_idma_2d.py
new file mode 100644
index 0000000..87b7bcb
--- /dev/null
+++ b/idma_tests/idma_multi_core_2d/gen_stimuli_idma_2d.py
@@ -0,0 +1,56 @@
+#!/usr/bin/env python
+
+import sys
+import random
+import argparse
+import math
+import re
+
+
+parser = argparse.ArgumentParser(description='Generate stimuli for iDMA simple transfer')
+
+args = parser.parse_args()
+
+def write_transfer_parameters_struct(f, name):
+    f.write ('typedef struct {\n')
+    f.write ('  unsigned int size;\n')
+    f.write ('  unsigned int length;\n')
+    f.write ('  unsigned int src_stride;\n')
+    f.write ('  unsigned int dst_stride;\n')
+    f.write ('} %s;\n\n' % name)
+    return
+
+def write_transfer_parameters_array(f, name, arr):
+    f.write ('TransferParameters %s[] = {\n' % name)
+    for v in arr:
+        length      = random.randint(1, MAX_LENGTH)
+        size        = random.randint(1, TRANSFER_SIZE) + length
+        src_stride  = random.randint(1, MAX_STRIDE) + length
+        dst_stride  = random.randint(1, MAX_STRIDE) + length
+        f.write('{%d, %d, %d, %d},\n' % (size, length, src_stride, dst_stride))
+    f.write('};\n\n')
+    return
+
+def write_define(f, name,val):
+    f.write('#define %s %d\n\n' % (name,val))
+    return
+
+# Randomize between 1 and 10 the number of 2D iDMA transfers to be performed
+NB_TRANSFERS = random.randint(1, 10)
+
+# Randomize between 1 and 128 the size in bytes of each transfer
+TRANSFER_SIZE = 128
+MAX_STRIDE    = 10
+MAX_LENGTH    = 10
+
+transfer_params = [None] * NB_TRANSFERS
+
+f_params    = open('idma_parameters.h', 'w')
+f_defines   = open('idma_defines.h', 'w')
+
+write_define(f_defines, 'NB_TRANSFERS', NB_TRANSFERS)
+write_transfer_parameters_struct(f_params, 'TransferParameters')
+write_transfer_parameters_array(f_params, 'transfer_params', transfer_params)
+
+f_params.close()
+f_defines.close()
\ No newline at end of file
diff --git a/idma_tests/idma_multi_core_2d/idma_defines.h b/idma_tests/idma_multi_core_2d/idma_defines.h
new file mode 100644
index 0000000..618d62d
--- /dev/null
+++ b/idma_tests/idma_multi_core_2d/idma_defines.h
@@ -0,0 +1,2 @@
+#define NB_TRANSFERS 8
+
diff --git a/idma_tests/idma_multi_core_2d/idma_multi_core_2d.c b/idma_tests/idma_multi_core_2d/idma_multi_core_2d.c
new file mode 100644
index 0000000..f5dc0eb
--- /dev/null
+++ b/idma_tests/idma_multi_core_2d/idma_multi_core_2d.c
@@ -0,0 +1,209 @@
+#include "idma_multi_core_2d.h"
+
+#define TOT_SIZE 8 * CORE_SPACE
+
+int errors[8] = {0};
+int test_status = 8;
+
+uint32_t l1_addr[8] = {0};
+uint32_t l1_dst_addr[8] = {0};
+uint32_t l2_addr[8] = {0};
+
+void print_transfer (TransferParameters transfer) {
+    if (rt_core_id() == 0) {
+        PRINTF ("Transfer Parameters: \n");
+        PRINTF ("Size: %d | Length: %d \n", transfer.size, transfer.length);
+        PRINTF ("Src_stride_2d: %d | Dst_stride_2d: %d \n", transfer.src_stride, transfer.dst_stride);
+    }
+}
+
+int test_idma_2D (int core_id, TransferParameters transfer, int ext2loc, int loc2loc) {
+    volatile uint8_t *src_ptr, *dst_ptr;
+
+    int error = 0;
+    int offset_2d;
+
+    uint32_t src_stride = transfer.src_stride;
+    uint32_t dst_stride = transfer.dst_stride;
+    uint32_t size = transfer.size;
+    uint32_t length = transfer.length;
+    uint32_t num_reps = size/length;
+
+    if (loc2loc == 1) {
+        // L1 to L1 transfer
+        src_ptr = (uint8_t*) l1_addr[core_id];
+        dst_ptr = (uint8_t*) l1_dst_addr[core_id];
+    } else if (ext2loc == 1) {
+        // L2 to L1 transfer
+        src_ptr = (uint8_t*) l2_addr[core_id];
+        dst_ptr = (uint8_t*) l1_addr[core_id];
+    } else {
+        // L1 to L2 transfer
+        src_ptr = (uint8_t*) l1_addr[core_id];
+        dst_ptr = (uint8_t*) l2_addr[core_id];
+    }
+
+    // Fill source region with test data
+    for (int q = 0; q < num_reps; q++) {
+        offset_2d = q * src_stride;
+        for (int i = 0; i < length; i++) {
+            src_ptr[i+offset_2d] = (uint8_t)(i & 0xFF);
+        }
+    }
+
+    if (loc2loc == 1) {
+        plp_cl_dma_wait_toL1(pulp_cl_idma_L1ToL1_2d((unsigned int)src_ptr, (unsigned int)dst_ptr, length, src_stride, dst_stride, num_reps));
+    } if (ext2loc == 1) {
+        plp_cl_dma_wait_toL1(pulp_cl_idma_L2ToL1_2d((unsigned int)src_ptr, (unsigned int)dst_ptr, length, src_stride, dst_stride, num_reps));
+    } else {
+        plp_cl_dma_wait_toL2(pulp_cl_idma_L1ToL2_2d((unsigned int)src_ptr, (unsigned int)dst_ptr, length, src_stride, dst_stride, num_reps));
+    }
+
+    // Check the results
+    
+    for (unsigned int rep = 0; rep < num_reps; rep++) {
+        unsigned int src_offset = rep * src_stride;
+        unsigned int dst_offset = rep * dst_stride;
+        for (unsigned int i = 0; i < length; i++) {
+            uint8_t expected = src_ptr[src_offset + i];
+            uint8_t actual = dst_ptr[dst_offset + i];
+
+            if (expected != actual) {
+                error++;
+                if (core_id == 0) {
+                    PRINTF ("ERROR: expected[%d] @%8x = %8x vs actual[%d] @%8x = %8x \n", src_offset + i, &src_ptr[src_offset + i], 
+                            expected, dst_offset+i, &dst_ptr[dst_offset + i], actual);
+                }
+            }
+
+        }
+    }
+    
+    return error;
+}
+
+void allocate_mem_to_cores () {
+    int core_id = rt_core_id();
+
+    // Pre-allocate TOT_SIZE = 8 * CORE_SPACE: then we split this window to assign
+    // each core its available space for iDMA transfers
+    // pi_l1_malloc starts allocating from 0x10004008 in L1
+    // pi_l2_malloc starts allocating from 0x1c000a60 in L2
+
+    if (core_id == 0) {
+        l1_addr[0]     = (uint32_t) pi_l1_malloc(0, TOT_SIZE);
+        l1_dst_addr[0] = (uint32_t) pi_l1_malloc(0, TOT_SIZE);
+        l2_addr[0]     = (uint32_t) pi_l2_malloc(TOT_SIZE);
+    }
+
+    // The following synch_barrier is needed so that
+    // no core can assign its address range until the mallocs are executed
+    synch_barrier();
+
+    l1_addr[core_id]     = l1_addr[0] + core_id * CORE_SPACE;
+    l1_dst_addr[core_id] = l1_dst_addr[0] + core_id * CORE_SPACE;
+    l2_addr[core_id]     = l2_addr[0] + core_id * CORE_SPACE;
+
+    // The following synch_barrier is needed so that
+    // no core can start executing until all address ranges have been assigned
+    synch_barrier();
+
+}
+
+void free_allocated_memory () {
+
+    synch_barrier();
+    // Only Core 0 takes care of freeing the allocated memory, since it's the one
+    // allocated it at the beginning of the test
+    if (core_id == 0) {
+        pi_l1_free(0, l1_addr, TOT_SIZE);
+        pi_l1_free(0, l1_dst_addr, TOT_SIZE);
+        pi_l2_free(l2_addr, TOT_SIZE);
+    }
+}
+
+int main () {
+
+    int core_id = rt_core_id();
+
+    allocate_mem_to_cores();
+
+    TransferParameters transfer;
+
+    #ifdef MULTI_CORE_P
+        // MULTI CORE PARALLEL MODE: each core uses the iDMA in a parallel manner
+        if (core_id == 0) {
+            PRINTF ("MULTI CORE PARALLEL MODE \n");
+        }
+        for (int k = 0; k < TRANSFERS; k++) {
+            #ifdef QUICK_MODE
+            transfer = idma_presets[k];
+            #else
+            transfer = transfer_params[k];
+            #endif
+            print_transfer(transfer);
+            // L1 -> L2
+            errors[core_id] += test_idma_2D(core_id, transfer, 0, 0);
+            // L2 -> L1
+            errors[core_id] += test_idma_2D(core_id, transfer, 1, 0);
+            // L1 -> L1
+            errors[core_id] += test_idma_2D(core_id, transfer, 0, 1);
+        }
+        synch_barrier();
+    #elif MULTI_CORE_S
+        // MULTI CORE SERIAL MODE: each core uses the iDMA in a serial manner
+        if (core_id == 0) {
+            PRINTF ("MULTI CORE SERIAL MODE \n");
+        }
+        for (int i = 0; i < 8; i++) {
+            if (core_id == i) {
+                for (int k = 0; k < TRANSFERS; k++) {
+                    #ifdef QUICK_MODE
+                    transfer = idma_presets[k];
+                    #else
+                    transfer = transfer_params[k];
+                    #endif
+                    print_transfer(transfer);
+                    // L1 -> L2
+                    errors[core_id] += test_idma_2D(core_id, transfer, 0, 0);
+                    // L2 -> L1
+                    errors[core_id] += test_idma_2D(core_id, transfer, 1, 0);
+                    // L1 -> L1
+                    errors[core_id] += test_idma_2D(core_id, transfer, 0, 1);
+                }
+            }
+        }
+    #else
+        // SINGLE CORE MODE: just core 0 uses the iDMA
+        if (core_id == 0) {
+            PRINTF ("SINGLE CORE MODE \n");
+            for (int k = 0; k < TRANSFERS; k++) {
+                #ifdef QUICK_MODE
+                transfer = idma_presets[k];
+                #else
+                transfer = transfer_params[k];
+                #endif
+                print_transfer(transfer);
+                // L1 -> L2
+                errors[core_id] += test_idma_2D(core_id, transfer, 0, 0);
+                // L2 -> L1
+                errors[core_id] += test_idma_2D(core_id, transfer, 1, 0);
+                // L1 -> L1
+                errors[core_id] += test_idma_2D(core_id, transfer, 0, 1);
+            }
+        }
+    #endif
+
+    if (core_id == 0) {
+        for (int i = 0; i<8; i++) {
+            if (errors[i] == 0) {
+                PRINTF ("Core %d returned %d errors \n", i, errors[i]);
+                test_status --;
+            }
+        }
+    }
+
+    free_allocated_memory();
+
+    return test_status;
+}
\ No newline at end of file
diff --git a/idma_tests/idma_multi_core_2d/idma_multi_core_2d.h b/idma_tests/idma_multi_core_2d/idma_multi_core_2d.h
new file mode 100644
index 0000000..87d9047
--- /dev/null
+++ b/idma_tests/idma_multi_core_2d/idma_multi_core_2d.h
@@ -0,0 +1,48 @@
+// Copyright 2022 ETH Zurich and University of Bologna.
+// Solderpad Hardware License, Version 0.51, see LICENSE for details.
+// SPDX-License-Identifier: SHL-0.51
+//
+//Author: Andreas Kuster <kustera@ethz.ch>
+//
+//Description: Generated register defines for dma_frontend
+
+#include "pulp.h"
+#include <stdio.h>
+#include <stdlib.h>
+#include <stdint.h>
+#include <string.h>
+#include <assert.h>
+
+#include "idma_defines.h"
+#include "idma_parameters.h"
+#include "idma_presets.h"
+
+#ifndef _DMA_FRONTEND_REG_DEFS_
+#define _DMA_FRONTEND_REG_DEFS_
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#define DMA_CONF_DECOUPLE 0
+#define DMA_CONF_DEBURST 0
+#define DMA_CONF_SERIALIZE 0
+
+#define CORE_SPACE 2048
+
+#ifdef QUICK_MODE
+#define TRANSFERS 7
+#else
+#define TRANSFERS NB_TRANSFERS
+#endif
+
+#ifdef VERBOSE
+    #define PRINTF(...) printf(__VA_ARGS__)
+#else
+    #define PRINTF(...)
+#endif
+
+#ifdef __cplusplus
+}  // extern "C"
+#endif
+#endif  // _DMA_FRONTEND_REG_DEFS_
diff --git a/idma_tests/idma_multi_core_2d/idma_parameters.h b/idma_tests/idma_multi_core_2d/idma_parameters.h
new file mode 100644
index 0000000..bc02dbc
--- /dev/null
+++ b/idma_tests/idma_multi_core_2d/idma_parameters.h
@@ -0,0 +1,18 @@
+typedef struct {
+  unsigned int size;
+  unsigned int length;
+  unsigned int src_stride;
+  unsigned int dst_stride;
+} TransferParameters;
+
+TransferParameters transfer_params[] = {
+{5, 1, 10, 8},
+{64, 6, 12, 8},
+{46, 5, 11, 7},
+{66, 3, 5, 9},
+{54, 7, 13, 10},
+{84, 8, 18, 15},
+{106, 2, 12, 11},
+{65, 9, 12, 12},
+};
+
diff --git a/idma_tests/idma_multi_core_2d/idma_presets.h b/idma_tests/idma_multi_core_2d/idma_presets.h
new file mode 100644
index 0000000..edc740a
--- /dev/null
+++ b/idma_tests/idma_multi_core_2d/idma_presets.h
@@ -0,0 +1,11 @@
+
+TransferParameters idma_presets[] = {
+{1, 1, 1, 1},
+{2, 2, 512, 512},
+{4, 2, 256, 256},
+{16, 8, 64, 64},
+{64, 8, 16, 16},
+{128, 8, 8, 8},
+{512, 2, 4, 4},
+};
+
diff --git a/idma_tests/idma_multi_core_3d/Makefile b/idma_tests/idma_multi_core_3d/Makefile
new file mode 100644
index 0000000..e206393
--- /dev/null
+++ b/idma_tests/idma_multi_core_3d/Makefile
@@ -0,0 +1,27 @@
+PULP_APP = test
+TEST_SRCS ?= idma_multi_core_3d.c
+PULP_APP_SRCS = $(TEST_SRCS)
+
+ifdef VERBOSE
+PULP_CFLAGS += -DVERBOSE
+endif
+
+ifdef MULTI_CORE_P
+PULP_CFLAGS += -DMULTI_CORE_P
+endif
+
+ifdef MULTI_CORE_S
+PULP_CFLAGS += -DMULTI_CORE_S
+endif
+
+ifdef QUICK_MODE
+PULP_CFLAGS += -DQUICK_MODE
+endif
+
+PULP_CFLAGS += -O3
+stackSize = 4096
+
+include $(PULP_SDK_HOME)/install/rules/pulp.mk
+
+stimuli:
+	python gen_stimuli_idma_3d.py
\ No newline at end of file
diff --git a/idma_tests/idma_multi_core_3d/gen_stimuli_idma_3d.py b/idma_tests/idma_multi_core_3d/gen_stimuli_idma_3d.py
new file mode 100644
index 0000000..75469b9
--- /dev/null
+++ b/idma_tests/idma_multi_core_3d/gen_stimuli_idma_3d.py
@@ -0,0 +1,66 @@
+#!/usr/bin/env python
+
+import sys
+import random
+import argparse
+import math
+import re
+
+
+parser = argparse.ArgumentParser(description='Generate stimuli for iDMA simple transfer')
+
+args = parser.parse_args()
+
+def write_transfer_parameters_struct(f, name):
+    f.write ('typedef struct {\n')
+    f.write ('  unsigned int size;\n')
+    f.write ('  unsigned int length;\n')
+    f.write ('  unsigned int src_stride_2d;\n')
+    f.write ('  unsigned int dst_stride_2d;\n')
+    f.write ('  unsigned int src_stride_3d;\n')
+    f.write ('  unsigned int dst_stride_3d;\n')
+    f.write ('  unsigned int num_reps_3d;\n')
+    f.write ('} %s;\n\n' % name)
+    return
+
+def write_transfer_parameters_array(f, name, arr):
+    f.write ('TransferParameters %s[] = {\n' % name)
+    for v in arr:
+        length          = random.randint(1, MAX_LENGTH)
+        size            = random.randint(1, TRANSFER_SIZE)  + length
+        src_stride_2d   = random.randint(1, MAX_STRIDE)     + length
+        dst_stride_2d   = random.randint(1, MAX_STRIDE)     + length
+        src_stride_3d   = random.randint(1, MAX_STRIDE)     + length
+        dst_stride_3d   = random.randint(1, MAX_STRIDE)     + length
+        num_reps_3d     = random.randint(1, MAX_REPS)
+
+        f.write('{%d, %d, %d, %d, %d, %d, %d},\n' % (size, length, src_stride_2d, dst_stride_2d, src_stride_3d, dst_stride_3d, num_reps_3d))
+    f.write('};\n\n')
+    return
+
+def write_define(f, name,val):
+    f.write('#define %s %d\n\n' % (name,val))
+    return
+
+# Randomize between 1 and 10 the number of 3D iDMA transfers to be performed
+# For each transfer we set a size in bytes.
+
+NB_TRANSFERS = random.randint(1, 10)
+
+# Randomize between 1 and 128 the size of each transfer
+TRANSFER_SIZE = 128
+MAX_STRIDE   = 10
+MAX_LENGTH   = 10
+MAX_REPS     = 5
+
+transfer_params = [None] * NB_TRANSFERS
+
+f_params    = open('idma_parameters.h', 'w')
+f_defines   = open('idma_defines.h', 'w')
+
+write_define(f_defines, 'NB_TRANSFERS', NB_TRANSFERS)
+write_transfer_parameters_struct(f_params, 'TransferParameters')
+write_transfer_parameters_array(f_params, 'transfer_params', transfer_params)
+
+f_params.close()
+f_defines.close()
\ No newline at end of file
diff --git a/idma_tests/idma_multi_core_3d/idma_defines.h b/idma_tests/idma_multi_core_3d/idma_defines.h
new file mode 100644
index 0000000..05d5bda
--- /dev/null
+++ b/idma_tests/idma_multi_core_3d/idma_defines.h
@@ -0,0 +1,2 @@
+#define NB_TRANSFERS 10
+
diff --git a/idma_tests/idma_multi_core_3d/idma_multi_core_3d.c b/idma_tests/idma_multi_core_3d/idma_multi_core_3d.c
new file mode 100644
index 0000000..b8479ff
--- /dev/null
+++ b/idma_tests/idma_multi_core_3d/idma_multi_core_3d.c
@@ -0,0 +1,230 @@
+#include "idma_multi_core_3d.h"
+
+#define TOT_SIZE 8 * CORE_SPACE
+
+int errors[8] = {0};
+int test_status = 8;
+
+uint32_t l1_addr[8] = {0};
+uint32_t l1_dst_addr[8] = {0};
+uint32_t l2_addr[8] = {0};
+
+void print_transfer (TransferParameters transfer) {
+    if (rt_core_id() == 0) {
+        PRINTF ("Transfer Parameters: \n");
+        PRINTF ("Size: %d | Length: %d \n", transfer.size, transfer.length);
+        PRINTF ("Src_stride_2d: %d | Dst_stride_2d: %d \n", transfer.src_stride_2d, transfer.dst_stride_2d);
+        PRINTF ("Src_stride_3d: %d | Dst_stride_3d: %d \n", transfer.src_stride_3d, transfer.dst_stride_3d);
+        PRINTF ("Num_reps_3d: %d \n", transfer.num_reps_3d);
+    }
+}
+
+int test_idma_3D (int core_id, TransferParameters transfer, int ext2loc, int loc2loc) {
+    volatile uint8_t *src_ptr, *dst_ptr;
+    unsigned int offset_3d = 0;
+    unsigned int offset_2d = 0;
+
+    int error = 0;
+
+    uint32_t src_stride_2d = transfer.src_stride_2d;
+    uint32_t dst_stride_2d = transfer.dst_stride_2d;
+    uint32_t src_stride_3d = transfer.src_stride_3d;
+    uint32_t dst_stride_3d = transfer.dst_stride_3d;
+    uint32_t size = transfer.size;
+    uint32_t length = transfer.length;
+    uint32_t num_reps = size/length;
+    uint32_t num_reps_3d = transfer.num_reps_3d;
+
+    if (loc2loc == 1) {
+        // L1 to L1 transfer
+        src_ptr = (uint8_t*) l1_addr[core_id];
+        dst_ptr = (uint8_t*) l1_dst_addr[core_id];
+    } else if (ext2loc == 1) {
+        // L2 to L1 transfer
+        src_ptr = (uint8_t*) l2_addr[core_id];
+        dst_ptr = (uint8_t*) l1_addr[core_id];
+    } else {
+        // L1 to L2 transfer
+        src_ptr = (uint8_t*) l1_addr[core_id];
+        dst_ptr = (uint8_t*) l2_addr[core_id];
+    }
+
+    // Fill source region with test data
+    for (int j = 0; j < num_reps_3d; j++) {
+        for (int q = 0; q < num_reps; q++) {
+            for (int i = 0; i < length; i++) {
+                src_ptr[i+offset_2d+offset_3d] = (uint8_t)(i & 0xFF);
+            }
+            offset_2d += src_stride_2d;
+        }
+        offset_2d = 0;
+        offset_3d += (num_reps-1) * src_stride_2d + src_stride_3d;
+    }
+
+
+    if (loc2loc == 1) {
+        plp_cl_dma_wait_toL1(pulp_cl_idma_L1ToL1_3d((unsigned int)src_ptr, (unsigned int)dst_ptr, length, src_stride_2d, dst_stride_2d, num_reps, 
+        src_stride_3d, dst_stride_3d, num_reps_3d));
+    } else if (ext2loc == 1) {
+        plp_cl_dma_wait_toL1(pulp_cl_idma_L2ToL1_3d((unsigned int)src_ptr, (unsigned int)dst_ptr, length, src_stride_2d, dst_stride_2d, num_reps, 
+        src_stride_3d, dst_stride_3d, num_reps_3d));
+    } else {
+        plp_cl_dma_wait_toL2(pulp_cl_idma_L1ToL2_3d((unsigned int)src_ptr, (unsigned int)dst_ptr, length, src_stride_2d, dst_stride_2d, num_reps, 
+        src_stride_3d, dst_stride_3d, num_reps_3d));
+    }
+
+    // Check the results
+    int src_offset_2d = 0;
+    int dst_offset_2d = 0;
+    int src_offset_3d = 0;
+    int dst_offset_3d = 0;
+
+    for (int rep_3d = 0; rep_3d < num_reps_3d; rep_3d ++) {
+        for (unsigned int rep = 0; rep < num_reps; rep++) {
+            for (unsigned int i = 0; i < length; i++) {
+                uint8_t expected = src_ptr[src_offset_2d + src_offset_3d + i];
+                uint8_t actual   = dst_ptr[dst_offset_2d + dst_offset_3d + i];
+
+                if (expected != actual) {
+                    if (core_id == 0) {
+                        PRINTF ("ERROR: expected @%8x[%d] = %8x vs actual @%8x[%d] = %8x \n", &src_ptr[src_offset_2d + src_offset_3d + i], src_offset_2d + src_offset_3d + i, 
+                                expected, &dst_ptr[dst_offset_2d + dst_offset_3d + i], dst_offset_2d + dst_offset_3d + i, actual);
+                    }
+                    error++;
+                }
+            }
+            src_offset_2d += src_stride_2d;
+            dst_offset_2d += dst_stride_2d;
+        }
+        src_offset_2d = 0;
+        dst_offset_2d = 0;
+        src_offset_3d += (num_reps-1) * src_stride_2d + src_stride_3d;
+        dst_offset_3d += (num_reps-1) * dst_stride_2d + dst_stride_3d;
+    }
+
+    return error;
+}
+
+void allocate_mem_to_cores () {
+    int core_id = rt_core_id();
+
+    // Pre-allocate TOT_SIZE = 8 * CORE_SPACE: then we split this window to assign
+    // each core its available space for iDMA transfers
+    // pi_l1_malloc starts allocating from 0x10004008 in L1
+    // pi_l2_malloc starts allocating from 0x1c000a60 in L2
+
+    if (core_id == 0) {
+        l1_addr[0]     = (uint32_t) pi_l1_malloc(0, TOT_SIZE);
+        l1_dst_addr[0] = (uint32_t) pi_l1_malloc(0, TOT_SIZE);
+        l2_addr[0]     = (uint32_t) pi_l2_malloc(TOT_SIZE);
+    }
+
+    // The following synch_barrier is needed so that
+    // no core can assign its address range until the mallocs are executed
+    synch_barrier();
+
+    l1_addr[core_id]     = l1_addr[0] + core_id * CORE_SPACE;
+    l1_dst_addr[core_id] = l1_dst_addr[0] + core_id * CORE_SPACE;
+    l2_addr[core_id]     = l2_addr[0] + core_id * CORE_SPACE;
+
+    // The following synch_barrier is needed so that
+    // no core can start executing until all address ranges have been assigned
+    synch_barrier();
+
+}
+
+void free_allocated_memory () {
+
+    synch_barrier();
+    // Only Core 0 takes care of freeing the allocated memory, since it's the one
+    // allocated it at the beginning of the test
+    if (core_id == 0) {
+        pi_l1_free(0, l1_addr, TOT_SIZE);
+        pi_l1_free(0, l1_dst_addr, TOT_SIZE);
+        pi_l2_free(l2_addr, TOT_SIZE);
+    }
+}
+
+int main () {
+
+    int core_id = rt_core_id();
+
+    allocate_mem_to_cores();
+
+    TransferParameters transfer;
+
+    #ifdef MULTI_CORE_P
+        // MULTI CORE PARALLEL MODE: each core uses the iDMA in a parallel manner
+        if (core_id == 0) {
+            PRINTF ("MULTI CORE PARALLEL MODE \n");
+        }
+        for (int k = 0; k < TRANSFERS; k++) {
+            #ifdef QUICK_MODE
+            transfer = idma_presets[k];
+            #else
+            transfer = transfer_params[k];
+            #endif
+            print_transfer(transfer);
+            // L1 -> L2
+            errors[core_id] += test_idma_3D(core_id, transfer, 0, 0);
+            // L2 -> L1
+            errors[core_id] += test_idma_3D(core_id, transfer, 1, 0);
+            // L1 -> L1 transfer
+            errors[core_id] += test_idma_3D(core_id, transfer, 0, 1);
+        }
+        synch_barrier();
+    #elif MULTI_CORE_S
+        // MULTI CORE SERIAL MODE: each core uses the iDMA in a serial manner
+        if (core_id == 0) {
+            PRINTF ("MULTI CORE SERIAL MODE \n");
+        }
+        for (int i = 0; i < 8; i++) {
+            if (core_id == i) {
+                for (int k = 0; k < TRANSFERS; k++) {
+                    #ifdef QUICK_MODE
+                    transfer = idma_presets[k];
+                    #else
+                    transfer = transfer_params[k];
+                    #endif
+                    print_transfer(transfer);
+                    // L1 -> L2
+                    errors[core_id] += test_idma_3D(core_id, transfer, 0, 0);
+                    // L2 -> L1
+                    errors[core_id] += test_idma_3D(core_id, transfer, 1, 0);
+                    // L1 -> L1 transfer
+                    errors[core_id] += test_idma_3D(core_id, transfer, 0, 1);
+                }
+            }
+        }
+    #else
+        if (core_id == 0) {
+            for (int k = 0; k < TRANSFERS; k++) {
+                #ifdef QUICK_MODE
+                transfer = idma_presets[k];
+                #else
+                transfer = transfer_params[k];
+                #endif
+                print_transfer(transfer);
+                PRINTF ("L1 to L2 \n");
+                errors[core_id] += test_idma_3D(core_id, transfer, 0, 0);
+                PRINTF ("L2 to L1 \n");
+                errors[core_id] += test_idma_3D(core_id, transfer, 1, 0);
+                PRINTF ("L1 to L1 \n");
+                errors[core_id] += test_idma_3D(core_id, transfer, 0, 1);
+            }
+        }
+    #endif
+
+    if (core_id == 0) {
+        for (int i = 0; i<8; i++) {
+            if (errors[i] == 0) {
+                PRINTF ("Core %d returned %d errors \n", i, errors[i]);
+                test_status--;
+            }
+        }
+    }
+
+    free_allocated_memory();
+
+    return test_status;
+}
\ No newline at end of file
diff --git a/idma_tests/idma_multi_core_3d/idma_multi_core_3d.h b/idma_tests/idma_multi_core_3d/idma_multi_core_3d.h
new file mode 100644
index 0000000..a6547ba
--- /dev/null
+++ b/idma_tests/idma_multi_core_3d/idma_multi_core_3d.h
@@ -0,0 +1,48 @@
+// Copyright 2022 ETH Zurich and University of Bologna.
+// Solderpad Hardware License, Version 0.51, see LICENSE for details.
+// SPDX-License-Identifier: SHL-0.51
+//
+//Author: Andreas Kuster <kustera@ethz.ch>
+//
+//Description: Generated register defines for dma_frontend
+
+#include "pulp.h"
+#include <stdio.h>
+#include <stdlib.h>
+#include <stdint.h>
+#include <string.h>
+#include <assert.h>
+
+#include "idma_defines.h"
+#include "idma_parameters.h"
+#include "idma_presets.h"
+
+#ifndef _DMA_FRONTEND_REG_DEFS_
+#define _DMA_FRONTEND_REG_DEFS_
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#define DMA_CONF_DECOUPLE 0
+#define DMA_CONF_DEBURST 0
+#define DMA_CONF_SERIALIZE 0
+
+#define CORE_SPACE 2048
+
+#ifdef QUICK_MODE
+#define TRANSFERS 9
+#else
+#define TRANSFERS NB_TRANSFERS
+#endif
+
+#ifdef VERBOSE
+    #define PRINTF(...) printf(__VA_ARGS__)
+#else
+    #define PRINTF(...)
+#endif
+
+#ifdef __cplusplus
+}  // extern "C"
+#endif
+#endif  // _DMA_FRONTEND_REG_DEFS_
diff --git a/idma_tests/idma_multi_core_3d/idma_parameters.h b/idma_tests/idma_multi_core_3d/idma_parameters.h
new file mode 100644
index 0000000..77c06db
--- /dev/null
+++ b/idma_tests/idma_multi_core_3d/idma_parameters.h
@@ -0,0 +1,23 @@
+typedef struct {
+  unsigned int size;
+  unsigned int length;
+  unsigned int src_stride_2d;
+  unsigned int dst_stride_2d;
+  unsigned int src_stride_3d;
+  unsigned int dst_stride_3d;
+  unsigned int num_reps_3d;
+} TransferParameters;
+
+TransferParameters transfer_params[] = {
+{107, 5, 12, 13, 7, 12, 1},
+{86, 5, 7, 10, 6, 12, 4},
+{95, 10, 12, 20, 15, 17, 4},
+{40, 2, 4, 12, 5, 4, 3},
+{105, 4, 14, 6, 13, 9, 3},
+{72, 8, 11, 17, 13, 15, 2},
+{54, 10, 14, 20, 13, 11, 3},
+{74, 6, 15, 10, 10, 11, 1},
+{114, 4, 9, 8, 12, 9, 3},
+{92, 3, 13, 5, 8, 5, 1},
+};
+
diff --git a/idma_tests/idma_multi_core_3d/idma_presets.h b/idma_tests/idma_multi_core_3d/idma_presets.h
new file mode 100644
index 0000000..82ce6d6
--- /dev/null
+++ b/idma_tests/idma_multi_core_3d/idma_presets.h
@@ -0,0 +1,16 @@
+
+// Parameters are declared in this order:
+// size, length, src_stride_2d, dst_stride_2d, src_stride_3d, dst_stride_3d, num_reps_3d
+
+TransferParameters idma_presets[] = {
+{1,   1,  1,  1,  1,  1, 1},
+{2,   8,  8,  8,  8,  8, 2},
+{3,   8,  8,  8,  8,  8, 4},
+{4,   8, 16, 16, 16, 16, 8},
+{8,   8, 32, 32, 32, 32, 8},
+{16,  8, 16, 16, 16, 16, 8},
+{32,  2,  2,  2,  2,  2, 2},
+{64,  4,  4,  4,  4,  4, 2},
+{128, 4,  4,  4,  4,  4, 4},
+};
+
diff --git a/idma_tests/idma_simple/Makefile b/idma_tests/idma_simple/Makefile
new file mode 100755
index 0000000..15ad295
--- /dev/null
+++ b/idma_tests/idma_simple/Makefile
@@ -0,0 +1,22 @@
+PULP_APP = test
+
+TEST_SRCS ?= simple_tx.c
+PULP_APP_SRCS = $(TEST_SRCS)
+PULP_APP_FC_SRCS = $(TEST_FC_SRCS)
+ifdef TEST_FC_SRCS
+pulpFc=1
+endif
+
+space :=
+space +=
+
+#BUILD_DIR = $(subst $(space),_,$(CURDIR)/build/$(TEST_SRCS))
+
+ifdef VERBOSE
+PULP_CFLAGS += -DVERBOSE
+endif
+
+PULP_CFLAGS += -O3
+stackSize = 4096
+
+include $(PULP_SDK_HOME)/install/rules/pulp.mk
diff --git a/idma_tests/idma_simple/dma_wave.do b/idma_tests/idma_simple/dma_wave.do
new file mode 100644
index 0000000..df25845
--- /dev/null
+++ b/idma_tests/idma_simple/dma_wave.do
@@ -0,0 +1,419 @@
+onerror {resume}
+quietly WaveActivateNextPane {} 0
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/rst_ni
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/test_mode_i
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/ext_master_req_o
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/ext_master_resp_i
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/term_event_o
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/term_irq_o
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/term_event_pe_o
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/term_irq_pe_o
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/busy_o
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/config_wdata
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/config_add
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/config_req
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/config_wen
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/config_be
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/config_id
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/config_gnt
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/config_r_rdata
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/config_r_valid
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/config_r_opc
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/config_r_id
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/obi_read_req_from_dma
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/obi_read_req_from_rrc
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/obi_reorg_req_from_dma
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/obi_reorg_req_from_rrc
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/obi_write_req_from_dma
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/obi_write_req_from_rrc
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/obi_read_req_muxed
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/obi_read_rsp_to_dma
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/obi_read_rsp_to_rrc
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/obi_reorg_rsp_to_dma
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/obi_reorg_rsp_to_rrc
+add wave -noupdate -expand -group {iDMA wrap} -subitemconfig {{/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/obi_write_rsp_to_dma[0]} -expand} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/obi_write_rsp_to_dma
+add wave -noupdate -expand -group {iDMA wrap} -subitemconfig {{/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/obi_write_rsp_to_rrc[0]} -expand} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/obi_write_rsp_to_rrc
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/obi_read_rsp_to_mux
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/soc_req
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/soc_rsp
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/dma_req
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/dma_rsp
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/dma_regs_req
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/dma_regs_rsp
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/stream_idx
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/twod_req
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/twod_req_queue
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/idma_req
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/idma_rsp
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/one_fe_valid
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/fe_valid
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/twod_queue_valid
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/be_valid
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/be_rsp_valid
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/fe_ready
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/twod_queue_ready
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/be_ready
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/be_rsp_ready
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/trans_complete
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/midend_busy
+add wave -noupdate -expand -group {iDMA wrap} -expand /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/idma_busy
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/done_id
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/next_id
+add wave -noupdate -group {init strm1} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/clk_i
+add wave -noupdate -group {init strm1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/init_write_rsp}
+add wave -noupdate -group {init strm1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/init_write_req}
+add wave -noupdate -group {init strm1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/init_read_req.req_valid}
+add wave -noupdate -group {init strm1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/init_read_rsp.req_ready}
+add wave -noupdate -group {init strm1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/init_read_rsp.rsp_valid}
+add wave -noupdate -group {init strm1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/init_read_req.rsp_ready}
+add wave -noupdate -group {init strm1} -expand {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/init_read_rsp}
+add wave -noupdate -group {init strm1} -expand {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/init_read_req}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/clk_i}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/rst_ni}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/testmode_i}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/axi_read_req_o}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/axi_read_rsp_i}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/init_read_req_o}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/init_read_rsp_i}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/obi_read_req_o}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/obi_read_rsp_i}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/init_write_req_o}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/init_write_rsp_i}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/obi_write_req_o}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/obi_write_rsp_i}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/r_dp_req_i}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/r_dp_valid_i}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/r_dp_ready_o}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/r_dp_rsp_o}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/r_dp_valid_o}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/r_dp_ready_i}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_dp_req_i}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_dp_valid_i}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_dp_ready_o}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_dp_rsp_o}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_dp_valid_o}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_dp_ready_i}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/ar_req_i}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/ar_valid_i}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/ar_ready_o}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/aw_req_i}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/aw_valid_i}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/aw_ready_o}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/dp_poison_i}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/r_chan_ready_o}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/r_chan_valid_o}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/r_dp_busy_o}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_dp_busy_o}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/buffer_busy_o}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/axi_buffer_in_valid}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/init_buffer_in_valid}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/obi_buffer_in_valid}
+add wave -noupdate -group {to OBI transport layer} -expand {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/buffer_in_valid}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/buffer_in_ready}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/buffer_out_valid}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/buffer_out_valid_shifted}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/init_buffer_out_ready}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/obi_buffer_out_ready}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/buffer_out_ready}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/buffer_out_ready_shifted}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/axi_buffer_in}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/init_buffer_in}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/obi_buffer_in}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/buffer_in}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/buffer_in_shifted}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/buffer_out}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/buffer_out_shifted}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/axi_r_chan_valid}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/init_r_chan_valid}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/obi_r_chan_valid}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/axi_r_chan_ready}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/init_r_chan_ready}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/obi_r_chan_ready}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/axi_r_dp_valid}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/init_r_dp_valid}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/obi_r_dp_valid}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/axi_r_dp_ready}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/init_r_dp_ready}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/obi_r_dp_ready}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/axi_r_dp_rsp}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/init_r_dp_rsp}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/obi_r_dp_rsp}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/axi_ar_ready}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/init_ar_ready}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/obi_ar_ready}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/init_w_dp_rsp_valid}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/obi_w_dp_rsp_valid}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/init_w_dp_rsp_ready}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/obi_w_dp_rsp_ready}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/init_w_dp_ready}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/obi_w_dp_ready}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/init_w_dp_rsp}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/obi_w_dp_rsp}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/init_aw_ready}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/obi_aw_ready}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_dp_req_valid}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_dp_req_ready}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_dp_rsp_mux_valid}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_dp_rsp_mux_ready}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_dp_rsp_valid}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_dp_rsp_ready}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_dp_rsp_mux}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_resp_fifo_in_valid}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_resp_fifo_in_ready}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_resp_fifo_out_protocol}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_resp_fifo_out_valid}
+add wave -noupdate -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_resp_fifo_out_ready}
+add wave -noupdate -group {backend 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/rst_ni}
+add wave -noupdate -group {backend 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/testmode_i}
+add wave -noupdate -group {backend 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/idma_req_i}
+add wave -noupdate -group {backend 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/req_valid_i}
+add wave -noupdate -group {backend 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/req_ready_o}
+add wave -noupdate -group {backend 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/idma_rsp_o}
+add wave -noupdate -group {backend 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/rsp_valid_o}
+add wave -noupdate -group {backend 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/rsp_ready_i}
+add wave -noupdate -group {backend 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/idma_eh_req_i}
+add wave -noupdate -group {backend 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/eh_req_valid_i}
+add wave -noupdate -group {backend 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/eh_req_ready_o}
+add wave -noupdate -group {backend 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/axi_read_req_o}
+add wave -noupdate -group {backend 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/axi_read_rsp_i}
+add wave -noupdate -group {backend 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/init_read_req_o}
+add wave -noupdate -group {backend 1} -expand {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/init_read_rsp_i}
+add wave -noupdate -group {backend 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/obi_read_req_o}
+add wave -noupdate -group {backend 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/obi_read_rsp_i}
+add wave -noupdate -group {backend 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/init_write_req_o}
+add wave -noupdate -group {backend 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/init_write_rsp_i}
+add wave -noupdate -group {backend 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/obi_write_req_o}
+add wave -noupdate -group {backend 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/obi_write_rsp_i}
+add wave -noupdate -group {backend 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/busy_o}
+add wave -noupdate -group {backend 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/dp_busy}
+add wave -noupdate -group {backend 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/dp_poison}
+add wave -noupdate -group {backend 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/r_req}
+add wave -noupdate -group {backend 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/w_req}
+add wave -noupdate -group {backend 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/r_valid}
+add wave -noupdate -group {backend 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/w_valid}
+add wave -noupdate -group {backend 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/r_ready}
+add wave -noupdate -group {backend 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/w_ready}
+add wave -noupdate -group {backend 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/r_meta_req_tagged}
+add wave -noupdate -group {backend 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/w_meta_req_tagged}
+add wave -noupdate -group {backend 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/w_last_burst}
+add wave -noupdate -group {backend 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/w_last_ready}
+add wave -noupdate -group {backend 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/w_super_last}
+add wave -noupdate -group {backend 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/r_dp_req_in_ready}
+add wave -noupdate -group {backend 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/w_dp_req_in_ready}
+add wave -noupdate -group {backend 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/w_dp_req_out_valid}
+add wave -noupdate -group {backend 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/w_dp_req_out_ready}
+add wave -noupdate -group {backend 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/r_dp_req_out}
+add wave -noupdate -group {backend 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/r_dp_req_out_valid}
+add wave -noupdate -group {backend 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/r_dp_req_out_ready}
+add wave -noupdate -group {backend 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/w_dp_req_out}
+add wave -noupdate -group {backend 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/r_dp_rsp}
+add wave -noupdate -group {backend 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/w_dp_rsp}
+add wave -noupdate -group {backend 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/r_dp_rsp_valid}
+add wave -noupdate -group {backend 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/w_dp_rsp_valid}
+add wave -noupdate -group {backend 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/r_dp_rsp_ready}
+add wave -noupdate -group {backend 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/w_dp_rsp_ready}
+add wave -noupdate -group {backend 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/ar_ready}
+add wave -noupdate -group {backend 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/ar_ready_dp}
+add wave -noupdate -group {backend 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/aw_ready}
+add wave -noupdate -group {backend 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/aw_ready_dp}
+add wave -noupdate -group {backend 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/aw_valid_dp}
+add wave -noupdate -group {backend 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/ar_valid_dp}
+add wave -noupdate -group {backend 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/aw_req_dp}
+add wave -noupdate -group {backend 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/ar_req_dp}
+add wave -noupdate -group {backend 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/legalizer_flush}
+add wave -noupdate -group {backend 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/legalizer_kill}
+add wave -noupdate -group {backend 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/is_length_zero}
+add wave -noupdate -group {backend 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/req_valid}
+add wave -noupdate -group {backend 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/idma_rsp}
+add wave -noupdate -group {backend 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/rsp_valid}
+add wave -noupdate -group {backend 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/rsp_ready}
+add wave -noupdate -group {backend 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/r_chan_valid}
+add wave -noupdate -group {backend 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/r_chan_ready}
+add wave -noupdate -group {legalizer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/gen_hw_legalizer/i_idma_legalizer/clk_i}
+add wave -noupdate -group {legalizer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/gen_hw_legalizer/i_idma_legalizer/rst_ni}
+add wave -noupdate -group {legalizer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/gen_hw_legalizer/i_idma_legalizer/req_i}
+add wave -noupdate -group {legalizer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/gen_hw_legalizer/i_idma_legalizer/valid_i}
+add wave -noupdate -group {legalizer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/gen_hw_legalizer/i_idma_legalizer/ready_o}
+add wave -noupdate -group {legalizer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/gen_hw_legalizer/i_idma_legalizer/r_req_o}
+add wave -noupdate -group {legalizer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/gen_hw_legalizer/i_idma_legalizer/r_valid_o}
+add wave -noupdate -group {legalizer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/gen_hw_legalizer/i_idma_legalizer/r_ready_i}
+add wave -noupdate -group {legalizer 1} -expand -subitemconfig {{/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/gen_hw_legalizer/i_idma_legalizer/w_req_o.w_dp_req} -expand {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/gen_hw_legalizer/i_idma_legalizer/w_req_o.aw_req} -expand} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/gen_hw_legalizer/i_idma_legalizer/w_req_o}
+add wave -noupdate -group {legalizer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/gen_hw_legalizer/i_idma_legalizer/w_valid_o}
+add wave -noupdate -group {legalizer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/gen_hw_legalizer/i_idma_legalizer/w_ready_i}
+add wave -noupdate -group {legalizer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/gen_hw_legalizer/i_idma_legalizer/flush_i}
+add wave -noupdate -group {legalizer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/gen_hw_legalizer/i_idma_legalizer/kill_i}
+add wave -noupdate -group {legalizer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/gen_hw_legalizer/i_idma_legalizer/r_busy_o}
+add wave -noupdate -group {legalizer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/gen_hw_legalizer/i_idma_legalizer/w_busy_o}
+add wave -noupdate -group {legalizer 1} -expand {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/gen_hw_legalizer/i_idma_legalizer/r_tf_d}
+add wave -noupdate -group {legalizer 1} -childformat {{{/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/gen_hw_legalizer/i_idma_legalizer/r_tf_q.length} -radix unsigned}} -subitemconfig {{/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/gen_hw_legalizer/i_idma_legalizer/r_tf_q.length} {-height 16 -radix unsigned}} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/gen_hw_legalizer/i_idma_legalizer/r_tf_q}
+add wave -noupdate -group {legalizer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/gen_hw_legalizer/i_idma_legalizer/w_tf_d}
+add wave -noupdate -group {legalizer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/gen_hw_legalizer/i_idma_legalizer/w_tf_q}
+add wave -noupdate -group {legalizer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/gen_hw_legalizer/i_idma_legalizer/opt_tf_d}
+add wave -noupdate -group {legalizer 1} -expand {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/gen_hw_legalizer/i_idma_legalizer/opt_tf_q}
+add wave -noupdate -group {legalizer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/gen_hw_legalizer/i_idma_legalizer/r_tf_ena}
+add wave -noupdate -group {legalizer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/gen_hw_legalizer/i_idma_legalizer/w_tf_ena}
+add wave -noupdate -group {legalizer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/gen_hw_legalizer/i_idma_legalizer/r_page_num_bytes_to_pb}
+add wave -noupdate -group {legalizer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/gen_hw_legalizer/i_idma_legalizer/r_num_bytes_to_pb}
+add wave -noupdate -group {legalizer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/gen_hw_legalizer/i_idma_legalizer/w_page_num_bytes_to_pb}
+add wave -noupdate -group {legalizer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/gen_hw_legalizer/i_idma_legalizer/w_num_bytes_to_pb}
+add wave -noupdate -group {legalizer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/gen_hw_legalizer/i_idma_legalizer/c_num_bytes_to_pb}
+add wave -noupdate -group {legalizer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/gen_hw_legalizer/i_idma_legalizer/r_num_bytes_possible}
+add wave -noupdate -group {legalizer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/gen_hw_legalizer/i_idma_legalizer/r_num_bytes}
+add wave -noupdate -group {legalizer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/gen_hw_legalizer/i_idma_legalizer/r_addr_offset}
+add wave -noupdate -group {legalizer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/gen_hw_legalizer/i_idma_legalizer/r_done}
+add wave -noupdate -group {legalizer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/gen_hw_legalizer/i_idma_legalizer/w_num_bytes_possible}
+add wave -noupdate -group {legalizer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/gen_hw_legalizer/i_idma_legalizer/w_num_bytes}
+add wave -noupdate -group {legalizer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/gen_hw_legalizer/i_idma_legalizer/w_addr_offset}
+add wave -noupdate -group {legalizer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/gen_hw_legalizer/i_idma_legalizer/w_done}
+add wave -noupdate -group {rd page splitter 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/gen_hw_legalizer/i_idma_legalizer/i_read_page_splitter/addr_i}
+add wave -noupdate -group {rd page splitter 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/gen_hw_legalizer/i_idma_legalizer/i_read_page_splitter/not_bursting_i}
+add wave -noupdate -group {rd page splitter 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/gen_hw_legalizer/i_idma_legalizer/i_read_page_splitter/reduce_len_i}
+add wave -noupdate -group {rd page splitter 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/gen_hw_legalizer/i_idma_legalizer/i_read_page_splitter/max_llen_i}
+add wave -noupdate -group {rd page splitter 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/gen_hw_legalizer/i_idma_legalizer/i_read_page_splitter/num_bytes_to_pb_o}
+add wave -noupdate -group {rd page splitter 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/gen_hw_legalizer/i_idma_legalizer/i_read_page_splitter/page_addr_width}
+add wave -noupdate -group {rd page splitter 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/gen_hw_legalizer/i_idma_legalizer/i_read_page_splitter/page_size}
+add wave -noupdate -group {rd page splitter 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/gen_hw_legalizer/i_idma_legalizer/i_read_page_splitter/page_offset}
+add wave -noupdate -group {wr page splitter 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/gen_hw_legalizer/i_idma_legalizer/i_write_page_splitter/addr_i}
+add wave -noupdate -group {wr page splitter 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/gen_hw_legalizer/i_idma_legalizer/i_write_page_splitter/not_bursting_i}
+add wave -noupdate -group {wr page splitter 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/gen_hw_legalizer/i_idma_legalizer/i_write_page_splitter/reduce_len_i}
+add wave -noupdate -group {wr page splitter 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/gen_hw_legalizer/i_idma_legalizer/i_write_page_splitter/max_llen_i}
+add wave -noupdate -group {wr page splitter 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/gen_hw_legalizer/i_idma_legalizer/i_write_page_splitter/num_bytes_to_pb_o}
+add wave -noupdate -group {wr page splitter 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/gen_hw_legalizer/i_idma_legalizer/i_write_page_splitter/page_addr_width}
+add wave -noupdate -group {wr page splitter 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/gen_hw_legalizer/i_idma_legalizer/i_write_page_splitter/page_size}
+add wave -noupdate -group {wr page splitter 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/gen_hw_legalizer/i_idma_legalizer/i_write_page_splitter/page_offset}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/clk_i}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/rst_ni}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/testmode_i}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/axi_read_req_o}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/axi_read_rsp_i}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/init_read_req_o}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/init_read_rsp_i}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/obi_read_req_o}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/obi_read_rsp_i}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/init_write_req_o}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/init_write_rsp_i}
+add wave -noupdate -group {transport layer 1} -expand {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/obi_write_req_o}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/obi_write_rsp_i}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/r_dp_req_i}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/r_dp_valid_i}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/r_dp_ready_o}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/r_dp_rsp_o}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/r_dp_valid_o}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/r_dp_ready_i}
+add wave -noupdate -group {transport layer 1} -expand {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_dp_req_i}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_dp_valid_i}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_dp_ready_o}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_dp_rsp_o}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_dp_valid_o}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_dp_ready_i}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/ar_req_i}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/ar_valid_i}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/ar_ready_o}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/aw_req_i}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/aw_valid_i}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/aw_ready_o}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/dp_poison_i}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/r_chan_ready_o}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/r_chan_valid_o}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/r_dp_busy_o}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_dp_busy_o}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/buffer_busy_o}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/axi_buffer_in_valid}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/init_buffer_in_valid}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/obi_buffer_in_valid}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/buffer_in_valid}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/buffer_in_ready}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/buffer_out_valid}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/buffer_out_valid_shifted}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/init_buffer_out_ready}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/obi_buffer_out_ready}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/buffer_out_ready}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/buffer_out_ready_shifted}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/axi_buffer_in}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/init_buffer_in}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/obi_buffer_in}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/buffer_in}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/buffer_in_shifted}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/buffer_out}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/buffer_out_shifted}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/axi_r_chan_valid}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/init_r_chan_valid}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/obi_r_chan_valid}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/axi_r_chan_ready}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/init_r_chan_ready}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/obi_r_chan_ready}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/axi_r_dp_valid}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/init_r_dp_valid}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/obi_r_dp_valid}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/axi_r_dp_ready}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/init_r_dp_ready}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/obi_r_dp_ready}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/axi_r_dp_rsp}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/init_r_dp_rsp}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/obi_r_dp_rsp}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/axi_ar_ready}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/init_ar_ready}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/obi_ar_ready}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/init_w_dp_rsp_valid}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/obi_w_dp_rsp_valid}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/init_w_dp_rsp_ready}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/obi_w_dp_rsp_ready}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/init_w_dp_ready}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/obi_w_dp_ready}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/init_w_dp_rsp}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/obi_w_dp_rsp}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/init_aw_ready}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/obi_aw_ready}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_dp_req_valid}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_dp_req_ready}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_dp_rsp_mux_valid}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_dp_rsp_mux_ready}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_dp_rsp_valid}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_dp_rsp_ready}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_dp_rsp_mux}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_resp_fifo_in_valid}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_resp_fifo_in_ready}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_resp_fifo_out_protocol}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_resp_fifo_out_valid}
+add wave -noupdate -group {transport layer 1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_resp_fifo_out_ready}
+add wave -noupdate -group {strm1 init read} -expand {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/i_idma_init_read/r_dp_req_i}
+add wave -noupdate -group {strm1 init read} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/i_idma_init_read/r_dp_valid_i}
+add wave -noupdate -group {strm1 init read} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/i_idma_init_read/r_dp_ready_o}
+add wave -noupdate -group {strm1 init read} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/i_idma_init_read/r_dp_rsp_o}
+add wave -noupdate -group {strm1 init read} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/i_idma_init_read/r_dp_valid_o}
+add wave -noupdate -group {strm1 init read} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/i_idma_init_read/r_dp_ready_i}
+add wave -noupdate -group {strm1 init read} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/i_idma_init_read/read_meta_req_i}
+add wave -noupdate -group {strm1 init read} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/i_idma_init_read/read_meta_valid_i}
+add wave -noupdate -group {strm1 init read} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/i_idma_init_read/read_meta_ready_o}
+add wave -noupdate -group {strm1 init read} -expand {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/i_idma_init_read/read_req_o}
+add wave -noupdate -group {strm1 init read} -expand {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/i_idma_init_read/read_rsp_i}
+add wave -noupdate -group {strm1 init read} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/i_idma_init_read/r_chan_ready_o}
+add wave -noupdate -group {strm1 init read} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/i_idma_init_read/r_chan_valid_o}
+add wave -noupdate -group {strm1 init read} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/i_idma_init_read/buffer_in_o}
+add wave -noupdate -group {strm1 init read} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/i_idma_init_read/buffer_in_valid_o}
+add wave -noupdate -group {strm1 init read} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/i_idma_init_read/buffer_in_ready_i}
+add wave -noupdate -group {strm1 init read} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/i_idma_init_read/read_aligned_in_mask}
+add wave -noupdate -group {strm1 init read} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/i_idma_init_read/mask_in}
+add wave -noupdate -group {strm1 init read} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/i_idma_init_read/in_valid}
+add wave -noupdate -group {strm1 init read} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/i_idma_init_read/in_ready}
+TreeUpdate [SetDefaultTree]
+quietly WaveActivateNextPane
+add wave -noupdate -color Gold {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/clk_i}
+TreeUpdate [SetDefaultTree]
+WaveRestoreCursors {{Cursor 1} {1574893420000 ps} 0}
+quietly wave cursor active 1
+configure wave -namecolwidth 252
+configure wave -valuecolwidth 224
+configure wave -justifyvalue left
+configure wave -signalnamewidth 1
+configure wave -snapdistance 10
+configure wave -datasetprefix 0
+configure wave -rowmargin 4
+configure wave -childrowmargin 2
+configure wave -gridoffset 0
+configure wave -gridperiod 1
+configure wave -griddelta 40
+configure wave -timeline 0
+configure wave -timelineunits ns
+update
+WaveRestoreZoom {248010100 ns} {3541692100 ns}
diff --git a/idma_tests/idma_simple/simple_tx.c b/idma_tests/idma_simple/simple_tx.c
new file mode 100644
index 0000000..5156078
--- /dev/null
+++ b/idma_tests/idma_simple/simple_tx.c
@@ -0,0 +1,133 @@
+#include <math.h>
+#include "pulp.h"
+//#include "mchan_tests.h"
+
+#define VERBOSE
+
+#define MAX_BUFFER_SIZE 0x2000
+
+//static unsigned char *ext;
+//static unsigned char *loc;
+
+//#define EXT_DATA_ADDR  ((unsigned int)ext)
+//#define TCDM_DATA_ADDR ((unsigned int)loc)
+
+L2_DATA static unsigned char ext[MAX_BUFFER_SIZE];
+L1_DATA static unsigned char loc[MAX_BUFFER_SIZE];
+
+#define EXT_DATA_ADDR  ((unsigned int) ext)
+#define TCDM_DATA_ADDR ((unsigned int) loc)
+typedef enum {RX, TX} test_type_t;
+
+int testMCHAN(unsigned int len, test_type_t type, unsigned int ext_addr, unsigned int tcdm_addr);
+int main()
+{
+
+  if (rt_cluster_id() != 0)
+    return bench_cluster_forward(0);
+
+
+  int   error_count = 0;
+
+
+ /*  if (rt_core_id() == 0) */
+/*   { */
+/*     if ((ext = rt_alloc(RT_ALLOC_L2_CL_DATA, MAX_BUFFER_SIZE)) == 0) return -1; */
+/*     if ((loc = rt_alloc(RT_ALLOC_CL_DATA, MAX_BUFFER_SIZE)) == 0) return -1; */
+/*   } */
+
+  if (get_core_id() == 0){
+
+
+    for ( int i = 5; i < 8045; i=5*i) {
+      error_count += testMCHAN(i, TX, ext, loc);
+    }
+    for ( int i = 5; i < 8045; i=5*i ) {
+      error_count += testMCHAN(i, RX, ext, loc);
+    }
+
+  }
+
+  return error_count;
+
+}
+
+int testMCHAN(unsigned int len, test_type_t type, unsigned int ext_addr, unsigned int tcdm_addr){
+
+  volatile unsigned int i,j,id;
+  volatile unsigned int test,read,error=0;
+
+  if (type == RX){
+
+#ifdef VERBOSE
+    printf ("STARTING TEST FOR RX %d OPERATION: \n", len);
+#endif
+
+    for (i=0; i<len; i++){
+      *(unsigned char*)(ext_addr + i) = i & 0xFF;
+    }
+
+    for (i=0; i<len+16; i++){
+      *(unsigned char*)(tcdm_addr + i) = 0;
+    }
+    id = pulp_idma_memcpy(ext_addr, tcdm_addr, len, IDMA_PROT_AXI, IDMA_PROT_OBI);
+  } else {
+
+#ifdef VERBOSE
+    printf ("STARTING TEST FOR TX %d OPERATION: \n", len);
+#endif
+
+    for (i=0; i<len; i++){
+      *(unsigned char*)(tcdm_addr + i) = i & 0xFF;
+    }
+
+    for (i=0; i<len+16; i++){
+      *(unsigned char*)(ext_addr+ i) = 0;
+    }
+    id = pulp_idma_memcpy(tcdm_addr, ext_addr, len, IDMA_PROT_OBI, IDMA_PROT_AXI);
+  }
+
+  plp_dma_barrier();
+
+
+  if (type == RX){
+
+    for (i=0; i<len; i++){
+
+      test = i & 0xFF;
+      read = *(unsigned char*)(tcdm_addr + i);
+
+      if ( test != read ){
+  printf("Error!!! Read: %x, Test:%x, Index: %d \n ",read,test,i);
+  error++;
+      }
+
+    }
+
+
+
+
+  }else{
+
+    for (i=0; i<len; i++){
+
+      test = i & 0xFF;
+      read = *(unsigned char*)(ext_addr + i);
+
+      if ( test != read ){
+  printf("Error!!! Read: %x, Test:%x, Index: %d \n ",read,test,i);
+  error++;
+      }
+
+    }
+
+  }
+
+  if (error == 0)
+    printf("OOOOOOK!!!!!!\n");
+  else
+    printf("NOT OK!!!!!\n");
+
+  return error;
+
+}
diff --git a/idma_tests/idma_simple/wave.do b/idma_tests/idma_simple/wave.do
new file mode 100644
index 0000000..e15c840
--- /dev/null
+++ b/idma_tests/idma_simple/wave.do
@@ -0,0 +1,66 @@
+onerror {resume}
+quietly WaveActivateNextPane {} 0
+add wave -noupdate -subitemconfig {/pulp_cluster_tb/cluster_i/s_dma_master_req.ar {-childformat {{size -radix unsigned}} -expand} /pulp_cluster_tb/cluster_i/s_dma_master_req.ar.size {-radix unsigned}} /pulp_cluster_tb/cluster_i/s_dma_master_req
+add wave -noupdate -subitemconfig {/pulp_cluster_tb/cluster_i/s_dma_master_resp.r -expand} /pulp_cluster_tb/cluster_i/s_dma_master_resp
+add wave -noupdate -group {s_hci_dma[0]} {/pulp_cluster_tb/cluster_i/s_hci_dma[0]/clk}
+add wave -noupdate -group {s_hci_dma[0]} {/pulp_cluster_tb/cluster_i/s_hci_dma[0]/req}
+add wave -noupdate -group {s_hci_dma[0]} {/pulp_cluster_tb/cluster_i/s_hci_dma[0]/gnt}
+add wave -noupdate -group {s_hci_dma[0]} {/pulp_cluster_tb/cluster_i/s_hci_dma[0]/r_valid}
+add wave -noupdate -group {s_hci_dma[0]} {/pulp_cluster_tb/cluster_i/s_hci_dma[0]/r_ready}
+add wave -noupdate -group {s_hci_dma[0]} {/pulp_cluster_tb/cluster_i/s_hci_dma[0]/add}
+add wave -noupdate -group {s_hci_dma[0]} {/pulp_cluster_tb/cluster_i/s_hci_dma[0]/wen}
+add wave -noupdate -group {s_hci_dma[0]} {/pulp_cluster_tb/cluster_i/s_hci_dma[0]/data}
+add wave -noupdate -group {s_hci_dma[0]} {/pulp_cluster_tb/cluster_i/s_hci_dma[0]/be}
+add wave -noupdate -group {s_hci_dma[0]} {/pulp_cluster_tb/cluster_i/s_hci_dma[0]/user}
+add wave -noupdate -group {s_hci_dma[0]} {/pulp_cluster_tb/cluster_i/s_hci_dma[0]/id}
+add wave -noupdate -group {s_hci_dma[0]} {/pulp_cluster_tb/cluster_i/s_hci_dma[0]/r_data}
+add wave -noupdate -group {s_hci_dma[0]} {/pulp_cluster_tb/cluster_i/s_hci_dma[0]/r_user}
+add wave -noupdate -group {s_hci_dma[0]} {/pulp_cluster_tb/cluster_i/s_hci_dma[0]/r_id}
+add wave -noupdate -group {s_hci_dma[0]} {/pulp_cluster_tb/cluster_i/s_hci_dma[0]/r_opc}
+add wave -noupdate -group {s_hci_dma[0]} {/pulp_cluster_tb/cluster_i/s_hci_dma[0]/ecc}
+add wave -noupdate -group {s_hci_dma[0]} {/pulp_cluster_tb/cluster_i/s_hci_dma[0]/r_ecc}
+add wave -noupdate -group {s_hci_dma[0]} {/pulp_cluster_tb/cluster_i/s_hci_dma[0]/ereq}
+add wave -noupdate -group {s_hci_dma[0]} {/pulp_cluster_tb/cluster_i/s_hci_dma[0]/egnt}
+add wave -noupdate -group {s_hci_dma[0]} {/pulp_cluster_tb/cluster_i/s_hci_dma[0]/r_evalid}
+add wave -noupdate -group {s_hci_dma[0]} {/pulp_cluster_tb/cluster_i/s_hci_dma[0]/r_eready}
+add wave -noupdate -group {s_hci_dma[0]} {/pulp_cluster_tb/cluster_i/s_hci_dma[0]/clk_assert}
+add wave -noupdate -expand -group {s_hci_dma[1]} {/pulp_cluster_tb/cluster_i/s_hci_dma[1]/clk}
+add wave -noupdate -expand -group {s_hci_dma[1]} {/pulp_cluster_tb/cluster_i/s_hci_dma[1]/req}
+add wave -noupdate -expand -group {s_hci_dma[1]} {/pulp_cluster_tb/cluster_i/s_hci_dma[1]/gnt}
+add wave -noupdate -expand -group {s_hci_dma[1]} {/pulp_cluster_tb/cluster_i/s_hci_dma[1]/r_valid}
+add wave -noupdate -expand -group {s_hci_dma[1]} {/pulp_cluster_tb/cluster_i/s_hci_dma[1]/r_ready}
+add wave -noupdate -expand -group {s_hci_dma[1]} {/pulp_cluster_tb/cluster_i/s_hci_dma[1]/add}
+add wave -noupdate -expand -group {s_hci_dma[1]} {/pulp_cluster_tb/cluster_i/s_hci_dma[1]/wen}
+add wave -noupdate -expand -group {s_hci_dma[1]} {/pulp_cluster_tb/cluster_i/s_hci_dma[1]/data}
+add wave -noupdate -expand -group {s_hci_dma[1]} {/pulp_cluster_tb/cluster_i/s_hci_dma[1]/be}
+add wave -noupdate -expand -group {s_hci_dma[1]} {/pulp_cluster_tb/cluster_i/s_hci_dma[1]/user}
+add wave -noupdate -expand -group {s_hci_dma[1]} {/pulp_cluster_tb/cluster_i/s_hci_dma[1]/id}
+add wave -noupdate -expand -group {s_hci_dma[1]} {/pulp_cluster_tb/cluster_i/s_hci_dma[1]/r_data}
+add wave -noupdate -expand -group {s_hci_dma[1]} {/pulp_cluster_tb/cluster_i/s_hci_dma[1]/r_user}
+add wave -noupdate -expand -group {s_hci_dma[1]} {/pulp_cluster_tb/cluster_i/s_hci_dma[1]/r_id}
+add wave -noupdate -expand -group {s_hci_dma[1]} {/pulp_cluster_tb/cluster_i/s_hci_dma[1]/r_opc}
+add wave -noupdate -expand -group {s_hci_dma[1]} {/pulp_cluster_tb/cluster_i/s_hci_dma[1]/ecc}
+add wave -noupdate -expand -group {s_hci_dma[1]} {/pulp_cluster_tb/cluster_i/s_hci_dma[1]/r_ecc}
+add wave -noupdate -expand -group {s_hci_dma[1]} {/pulp_cluster_tb/cluster_i/s_hci_dma[1]/ereq}
+add wave -noupdate -expand -group {s_hci_dma[1]} {/pulp_cluster_tb/cluster_i/s_hci_dma[1]/egnt}
+add wave -noupdate -expand -group {s_hci_dma[1]} {/pulp_cluster_tb/cluster_i/s_hci_dma[1]/r_evalid}
+add wave -noupdate -expand -group {s_hci_dma[1]} {/pulp_cluster_tb/cluster_i/s_hci_dma[1]/r_eready}
+add wave -noupdate -expand -group {s_hci_dma[1]} {/pulp_cluster_tb/cluster_i/s_hci_dma[1]/clk_assert}
+TreeUpdate [SetDefaultTree]
+WaveRestoreCursors {{Cursor 1} {262653051593 ps} 1} {{Cursor 2} {262655244483 ps} 0}
+quietly wave cursor active 2
+configure wave -namecolwidth 150
+configure wave -valuecolwidth 567
+configure wave -justifyvalue left
+configure wave -signalnamewidth 1
+configure wave -snapdistance 10
+configure wave -datasetprefix 0
+configure wave -rowmargin 4
+configure wave -childrowmargin 2
+configure wave -gridoffset 0
+configure wave -gridperiod 1
+configure wave -griddelta 40
+configure wave -timeline 0
+configure wave -timelineunits ps
+update
+WaveRestoreZoom {262620992344 ps} {262771194315 ps}
diff --git a/idma_tests/idma_wide/Makefile b/idma_tests/idma_wide/Makefile
new file mode 100755
index 0000000..d75dc52
--- /dev/null
+++ b/idma_tests/idma_wide/Makefile
@@ -0,0 +1,22 @@
+PULP_APP = test
+
+TEST_SRCS ?= idma_wide.c
+PULP_APP_SRCS = $(TEST_SRCS)
+PULP_APP_FC_SRCS = $(TEST_FC_SRCS)
+ifdef TEST_FC_SRCS
+pulpFc=1
+endif
+
+space :=
+space +=
+
+#BUILD_DIR = $(subst $(space),_,$(CURDIR)/build/$(TEST_SRCS))
+
+ifdef VERBOSE
+PULP_CFLAGS += -DVERBOSE
+endif
+
+PULP_CFLAGS += -O3
+stackSize = 4096
+
+include $(PULP_SDK_HOME)/install/rules/pulp.mk
diff --git a/idma_tests/idma_wide/idma_wide.c b/idma_tests/idma_wide/idma_wide.c
new file mode 100644
index 0000000..75e68cb
--- /dev/null
+++ b/idma_tests/idma_wide/idma_wide.c
@@ -0,0 +1,120 @@
+#include <math.h>
+#include <stdint.h>
+#include <inttypes.h>
+#include <stdio.h>
+#include "pulp.h"
+
+#define VERBOSE
+
+#define MAX_BUFFER_SIZE 0x2000 
+
+#define min_size 32 // 256 for wide AXI port
+#define max_size 45
+
+L2_DATA static uint8_t ext[MAX_BUFFER_SIZE]; // External memory buffer (L2)
+L1_DATA static uint8_t loc[MAX_BUFFER_SIZE]; // Local memory buffer (TCDM / L1)
+
+typedef enum {L2_TO_L1, L1_TO_L2} test_type_t;
+
+int test_idma(uint32_t size, test_type_t type, uint32_t ext_addr, uint32_t tcdm_addr);
+
+int main(void) {
+    if (rt_cluster_id() != 0) {
+        printf("Test only runs on cluster 0\n");
+        return bench_cluster_forward(0);
+    }
+    
+    int error_count = 0;
+
+    if (get_core_id() == 0) {
+
+        // Test for L2_TO_L1 operation
+        for (int size = min_size; size < max_size; size++) {
+            error_count += test_idma(size, L2_TO_L1, (uint32_t)ext, (uint32_t)loc);
+        }
+
+        // Test for L1_TO_L2 operation
+        for (int size = min_size; size < max_size; size++) {
+            error_count += test_idma(size, L1_TO_L2, (uint32_t)ext, (uint32_t)loc);
+        }
+    }
+
+    return error_count;
+}
+
+int test_idma(uint32_t size, test_type_t type, uint32_t ext_addr, uint32_t tcdm_addr) {
+    volatile uint8_t expected, actual;
+    volatile int error = 0;
+    volatile unsigned int id;
+
+    if (type == L2_TO_L1) {
+
+        for (uint32_t i = 0; i < size; i++) {
+            *(uint8_t *)(ext_addr + i) = (uint8_t)(i & 0xFF);
+        }
+
+        memset((void *)tcdm_addr, 0, size + 16);
+
+        id = pulp_idma_memcpy(ext_addr, tcdm_addr, size, IDMA_PROT_AXI, IDMA_PROT_OBI);
+    
+        /* L2_TO_L1 transaction:
+        +----------------+                        +----------------+
+        |                |       DMA Transfer     |                |
+        |  L2 Memory     | ---------------------> |  L1 Memory     |
+        |  (ext buffer)  |   (pulp_idma_memcpy)   |  (loc buffer)  |
+        |                |                        |                |
+        +----------------+                        +----------------+
+        */
+    
+    } else if (type == L1_TO_L2) {
+        // Fill L1 buffer with a pattern
+        for (uint32_t i = 0; i < size; i++) {
+            *(uint8_t *)(tcdm_addr + i) = (uint8_t)(i & 0xFF);
+        }
+
+        // Clear L2
+        memset((void *)ext_addr, 0, size + 16);
+
+        // Perform DMA transfer from local memory to external memory
+        id = pulp_idma_memcpy(tcdm_addr, ext_addr, size, IDMA_PROT_OBI, IDMA_PROT_AXI);
+
+        /* L1_TO_L2 transaction:
+        +----------------+                        +----------------+
+        |                |       DMA Transfer     |                |
+        |  L2 Memory     | <--------------------- |  L1 Memory     |
+        |  (ext buffer)  |   (pulp_idma_memcpy)   |  (loc buffer)  |
+        |                |                        |                |
+        +----------------+                        +----------------+
+        */
+    } else {
+        printf("Invalid test type. It must be either L2_TO_L1 or L1_TO_L2.\n");
+        return 1;
+    }
+
+    plp_dma_barrier();
+
+    // Verify data
+    for (uint32_t i = 0; i < size; i++) {
+        expected = (uint8_t)(i & 0xFF); // Same pattern as L2_TO_L1
+        if (type == L2_TO_L1) {
+            // Reading from the local memory buffer "loc" allocated in L1 memory
+            actual = *(uint8_t *)(tcdm_addr + i);
+        } else { // L1_TO_L2
+            // Reading from the external memory buffer "ext" allocated in L2 memory
+            actual = *(uint8_t *)(ext_addr + i);
+        } 
+
+        if (expected != actual) {
+            printf("Error at index %u: Expected 0x%02X, Got 0x%02X\n", i, expected, actual);
+            error++;
+        }
+    }
+
+    if (error == 0) {
+        printf("Test passed for %s of length %u.\n", type == L2_TO_L1 ? "L2_TO_L1" : "L1_TO_L2", size);
+    } else {
+        printf("Test failed for %s of length %u with %u errors.\n", type == L2_TO_L1 ? "L2_TO_L1" : "L1_TO_L2", size, error);
+    }
+    
+    return error;
+}
\ No newline at end of file
diff --git a/idma_tests/idma_wide/wave.do b/idma_tests/idma_wide/wave.do
new file mode 100644
index 0000000..e15c840
--- /dev/null
+++ b/idma_tests/idma_wide/wave.do
@@ -0,0 +1,66 @@
+onerror {resume}
+quietly WaveActivateNextPane {} 0
+add wave -noupdate -subitemconfig {/pulp_cluster_tb/cluster_i/s_dma_master_req.ar {-childformat {{size -radix unsigned}} -expand} /pulp_cluster_tb/cluster_i/s_dma_master_req.ar.size {-radix unsigned}} /pulp_cluster_tb/cluster_i/s_dma_master_req
+add wave -noupdate -subitemconfig {/pulp_cluster_tb/cluster_i/s_dma_master_resp.r -expand} /pulp_cluster_tb/cluster_i/s_dma_master_resp
+add wave -noupdate -group {s_hci_dma[0]} {/pulp_cluster_tb/cluster_i/s_hci_dma[0]/clk}
+add wave -noupdate -group {s_hci_dma[0]} {/pulp_cluster_tb/cluster_i/s_hci_dma[0]/req}
+add wave -noupdate -group {s_hci_dma[0]} {/pulp_cluster_tb/cluster_i/s_hci_dma[0]/gnt}
+add wave -noupdate -group {s_hci_dma[0]} {/pulp_cluster_tb/cluster_i/s_hci_dma[0]/r_valid}
+add wave -noupdate -group {s_hci_dma[0]} {/pulp_cluster_tb/cluster_i/s_hci_dma[0]/r_ready}
+add wave -noupdate -group {s_hci_dma[0]} {/pulp_cluster_tb/cluster_i/s_hci_dma[0]/add}
+add wave -noupdate -group {s_hci_dma[0]} {/pulp_cluster_tb/cluster_i/s_hci_dma[0]/wen}
+add wave -noupdate -group {s_hci_dma[0]} {/pulp_cluster_tb/cluster_i/s_hci_dma[0]/data}
+add wave -noupdate -group {s_hci_dma[0]} {/pulp_cluster_tb/cluster_i/s_hci_dma[0]/be}
+add wave -noupdate -group {s_hci_dma[0]} {/pulp_cluster_tb/cluster_i/s_hci_dma[0]/user}
+add wave -noupdate -group {s_hci_dma[0]} {/pulp_cluster_tb/cluster_i/s_hci_dma[0]/id}
+add wave -noupdate -group {s_hci_dma[0]} {/pulp_cluster_tb/cluster_i/s_hci_dma[0]/r_data}
+add wave -noupdate -group {s_hci_dma[0]} {/pulp_cluster_tb/cluster_i/s_hci_dma[0]/r_user}
+add wave -noupdate -group {s_hci_dma[0]} {/pulp_cluster_tb/cluster_i/s_hci_dma[0]/r_id}
+add wave -noupdate -group {s_hci_dma[0]} {/pulp_cluster_tb/cluster_i/s_hci_dma[0]/r_opc}
+add wave -noupdate -group {s_hci_dma[0]} {/pulp_cluster_tb/cluster_i/s_hci_dma[0]/ecc}
+add wave -noupdate -group {s_hci_dma[0]} {/pulp_cluster_tb/cluster_i/s_hci_dma[0]/r_ecc}
+add wave -noupdate -group {s_hci_dma[0]} {/pulp_cluster_tb/cluster_i/s_hci_dma[0]/ereq}
+add wave -noupdate -group {s_hci_dma[0]} {/pulp_cluster_tb/cluster_i/s_hci_dma[0]/egnt}
+add wave -noupdate -group {s_hci_dma[0]} {/pulp_cluster_tb/cluster_i/s_hci_dma[0]/r_evalid}
+add wave -noupdate -group {s_hci_dma[0]} {/pulp_cluster_tb/cluster_i/s_hci_dma[0]/r_eready}
+add wave -noupdate -group {s_hci_dma[0]} {/pulp_cluster_tb/cluster_i/s_hci_dma[0]/clk_assert}
+add wave -noupdate -expand -group {s_hci_dma[1]} {/pulp_cluster_tb/cluster_i/s_hci_dma[1]/clk}
+add wave -noupdate -expand -group {s_hci_dma[1]} {/pulp_cluster_tb/cluster_i/s_hci_dma[1]/req}
+add wave -noupdate -expand -group {s_hci_dma[1]} {/pulp_cluster_tb/cluster_i/s_hci_dma[1]/gnt}
+add wave -noupdate -expand -group {s_hci_dma[1]} {/pulp_cluster_tb/cluster_i/s_hci_dma[1]/r_valid}
+add wave -noupdate -expand -group {s_hci_dma[1]} {/pulp_cluster_tb/cluster_i/s_hci_dma[1]/r_ready}
+add wave -noupdate -expand -group {s_hci_dma[1]} {/pulp_cluster_tb/cluster_i/s_hci_dma[1]/add}
+add wave -noupdate -expand -group {s_hci_dma[1]} {/pulp_cluster_tb/cluster_i/s_hci_dma[1]/wen}
+add wave -noupdate -expand -group {s_hci_dma[1]} {/pulp_cluster_tb/cluster_i/s_hci_dma[1]/data}
+add wave -noupdate -expand -group {s_hci_dma[1]} {/pulp_cluster_tb/cluster_i/s_hci_dma[1]/be}
+add wave -noupdate -expand -group {s_hci_dma[1]} {/pulp_cluster_tb/cluster_i/s_hci_dma[1]/user}
+add wave -noupdate -expand -group {s_hci_dma[1]} {/pulp_cluster_tb/cluster_i/s_hci_dma[1]/id}
+add wave -noupdate -expand -group {s_hci_dma[1]} {/pulp_cluster_tb/cluster_i/s_hci_dma[1]/r_data}
+add wave -noupdate -expand -group {s_hci_dma[1]} {/pulp_cluster_tb/cluster_i/s_hci_dma[1]/r_user}
+add wave -noupdate -expand -group {s_hci_dma[1]} {/pulp_cluster_tb/cluster_i/s_hci_dma[1]/r_id}
+add wave -noupdate -expand -group {s_hci_dma[1]} {/pulp_cluster_tb/cluster_i/s_hci_dma[1]/r_opc}
+add wave -noupdate -expand -group {s_hci_dma[1]} {/pulp_cluster_tb/cluster_i/s_hci_dma[1]/ecc}
+add wave -noupdate -expand -group {s_hci_dma[1]} {/pulp_cluster_tb/cluster_i/s_hci_dma[1]/r_ecc}
+add wave -noupdate -expand -group {s_hci_dma[1]} {/pulp_cluster_tb/cluster_i/s_hci_dma[1]/ereq}
+add wave -noupdate -expand -group {s_hci_dma[1]} {/pulp_cluster_tb/cluster_i/s_hci_dma[1]/egnt}
+add wave -noupdate -expand -group {s_hci_dma[1]} {/pulp_cluster_tb/cluster_i/s_hci_dma[1]/r_evalid}
+add wave -noupdate -expand -group {s_hci_dma[1]} {/pulp_cluster_tb/cluster_i/s_hci_dma[1]/r_eready}
+add wave -noupdate -expand -group {s_hci_dma[1]} {/pulp_cluster_tb/cluster_i/s_hci_dma[1]/clk_assert}
+TreeUpdate [SetDefaultTree]
+WaveRestoreCursors {{Cursor 1} {262653051593 ps} 1} {{Cursor 2} {262655244483 ps} 0}
+quietly wave cursor active 2
+configure wave -namecolwidth 150
+configure wave -valuecolwidth 567
+configure wave -justifyvalue left
+configure wave -signalnamewidth 1
+configure wave -snapdistance 10
+configure wave -datasetprefix 0
+configure wave -rowmargin 4
+configure wave -childrowmargin 2
+configure wave -gridoffset 0
+configure wave -gridperiod 1
+configure wave -griddelta 40
+configure wave -timeline 0
+configure wave -timelineunits ps
+update
+WaveRestoreZoom {262620992344 ps} {262771194315 ps}
diff --git a/idma_tests/idma_zeromem/Makefile b/idma_tests/idma_zeromem/Makefile
new file mode 100755
index 0000000..d896e12
--- /dev/null
+++ b/idma_tests/idma_zeromem/Makefile
@@ -0,0 +1,22 @@
+PULP_APP = test
+
+TEST_SRCS ?= zeromem.c
+PULP_APP_SRCS = $(TEST_SRCS)
+PULP_APP_FC_SRCS = $(TEST_FC_SRCS)
+ifdef TEST_FC_SRCS
+pulpFc=1
+endif
+
+space :=
+space +=
+
+#BUILD_DIR = $(subst $(space),_,$(CURDIR)/build/$(TEST_SRCS))
+
+ifdef VERBOSE
+PULP_CFLAGS += -DVERBOSE
+endif
+
+PULP_CFLAGS += -O3
+stackSize = 4096
+
+include $(PULP_SDK_HOME)/install/rules/pulp.mk
diff --git a/idma_tests/idma_zeromem/dma_wave.do b/idma_tests/idma_zeromem/dma_wave.do
new file mode 100644
index 0000000..ab54467
--- /dev/null
+++ b/idma_tests/idma_zeromem/dma_wave.do
@@ -0,0 +1,179 @@
+onerror {resume}
+quietly WaveActivateNextPane {} 0
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/clk_i
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/rst_ni
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/test_mode_i
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/ext_master_req_o
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/ext_master_resp_i
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/term_event_o
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/term_irq_o
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/term_event_pe_o
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/term_irq_pe_o
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/busy_o
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/config_wdata
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/config_add
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/config_req
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/config_wen
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/config_be
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/config_id
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/config_gnt
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/config_r_rdata
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/config_r_valid
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/config_r_opc
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/config_r_id
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/obi_read_req_from_dma
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/obi_read_req_from_rrc
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/obi_reorg_req_from_dma
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/obi_reorg_req_from_rrc
+add wave -noupdate -expand -group {iDMA wrap} -expand -subitemconfig {{/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/obi_write_req_from_dma[0]} -expand} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/obi_write_req_from_dma
+add wave -noupdate -expand -group {iDMA wrap} -expand -subitemconfig {{/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/obi_write_req_from_rrc[0]} -expand {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/obi_write_req_from_rrc[0].a} -expand} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/obi_write_req_from_rrc
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/obi_read_req_muxed
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/obi_read_rsp_to_dma
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/obi_read_rsp_to_rrc
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/obi_reorg_rsp_to_dma
+add wave -noupdate -expand -group {iDMA wrap} -subitemconfig {{/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/obi_reorg_rsp_to_rrc[0]} -expand {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/obi_reorg_rsp_to_rrc[0].r} -expand} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/obi_reorg_rsp_to_rrc
+add wave -noupdate -expand -group {iDMA wrap} -expand -subitemconfig {{/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/obi_write_rsp_to_dma[0]} -expand {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/obi_write_rsp_to_dma[0].r} -expand} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/obi_write_rsp_to_dma
+add wave -noupdate -expand -group {iDMA wrap} -expand -subitemconfig {{/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/obi_write_rsp_to_rrc[0]} -expand {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/obi_write_rsp_to_rrc[0].r} -expand} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/obi_write_rsp_to_rrc
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/obi_read_rsp_to_mux
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/soc_req
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/soc_rsp
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/dma_req
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/dma_rsp
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/dma_regs_req
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/dma_regs_rsp
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/stream_idx
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/twod_req
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/twod_req_queue
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/idma_req
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/idma_rsp
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/one_fe_valid
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/fe_valid
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/twod_queue_valid
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/be_valid
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/be_rsp_valid
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/fe_ready
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/twod_queue_ready
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/be_ready
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/be_rsp_ready
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/trans_complete
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/midend_busy
+add wave -noupdate -expand -group {iDMA wrap} -expand /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/idma_busy
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/done_id
+add wave -noupdate -expand -group {iDMA wrap} /pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/next_id
+add wave -noupdate -expand -group {init strm1} -subitemconfig {{/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/init_write_rsp.rsp_chan} -expand} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/init_write_rsp}
+add wave -noupdate -expand -group {init strm1} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/init_write_req}
+add wave -noupdate -expand -group {init strm1} -expand {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/init_read_rsp}
+add wave -noupdate -expand -group {init strm1} -expand {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/init_read_req}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/clk_i}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/rst_ni}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/testmode_i}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/axi_read_req_o}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/axi_read_rsp_i}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/init_read_req_o}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/init_read_rsp_i}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/obi_read_req_o}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/obi_read_rsp_i}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/init_write_req_o}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/init_write_rsp_i}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/obi_write_req_o}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/obi_write_rsp_i}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/r_dp_req_i}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/r_dp_valid_i}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/r_dp_ready_o}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/r_dp_rsp_o}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/r_dp_valid_o}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/r_dp_ready_i}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_dp_req_i}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_dp_valid_i}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_dp_ready_o}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_dp_rsp_o}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_dp_valid_o}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_dp_ready_i}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/ar_req_i}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/ar_valid_i}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/ar_ready_o}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/aw_req_i}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/aw_valid_i}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/aw_ready_o}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/dp_poison_i}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/r_chan_ready_o}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/r_chan_valid_o}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/r_dp_busy_o}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_dp_busy_o}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/buffer_busy_o}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/axi_buffer_in_valid}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/init_buffer_in_valid}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/obi_buffer_in_valid}
+add wave -noupdate -expand -group {to OBI transport layer} -expand {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/buffer_in_valid}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/buffer_in_ready}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/buffer_out_valid}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/buffer_out_valid_shifted}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/init_buffer_out_ready}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/obi_buffer_out_ready}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/buffer_out_ready}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/buffer_out_ready_shifted}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/axi_buffer_in}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/init_buffer_in}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/obi_buffer_in}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/buffer_in}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/buffer_in_shifted}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/buffer_out}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/buffer_out_shifted}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/axi_r_chan_valid}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/init_r_chan_valid}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/obi_r_chan_valid}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/axi_r_chan_ready}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/init_r_chan_ready}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/obi_r_chan_ready}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/axi_r_dp_valid}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/init_r_dp_valid}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/obi_r_dp_valid}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/axi_r_dp_ready}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/init_r_dp_ready}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/obi_r_dp_ready}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/axi_r_dp_rsp}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/init_r_dp_rsp}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/obi_r_dp_rsp}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/axi_ar_ready}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/init_ar_ready}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/obi_ar_ready}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/init_w_dp_rsp_valid}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/obi_w_dp_rsp_valid}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/init_w_dp_rsp_ready}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/obi_w_dp_rsp_ready}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/init_w_dp_ready}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/obi_w_dp_ready}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/init_w_dp_rsp}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/obi_w_dp_rsp}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/init_aw_ready}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/obi_aw_ready}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_dp_req_valid}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_dp_req_ready}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_dp_rsp_mux_valid}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_dp_rsp_mux_ready}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_dp_rsp_valid}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_dp_rsp_ready}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_dp_rsp_mux}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_resp_fifo_in_valid}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_resp_fifo_in_ready}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_resp_fifo_out_protocol}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_resp_fifo_out_valid}
+add wave -noupdate -expand -group {to OBI transport layer} {/pulp_cluster_tb/cluster_i/inst_idma/dmac_wrap_i/gen_streams[1]/gen_cpy_in/i_idma_backend_r_axi_rw_init_rw_obi/i_idma_transport_layer/w_resp_fifo_out_ready}
+TreeUpdate [SetDefaultTree]
+WaveRestoreCursors {{Cursor 1} {226792000000 ps} 0}
+quietly wave cursor active 1
+configure wave -namecolwidth 252
+configure wave -valuecolwidth 120
+configure wave -justifyvalue left
+configure wave -signalnamewidth 1
+configure wave -snapdistance 10
+configure wave -datasetprefix 0
+configure wave -rowmargin 4
+configure wave -childrowmargin 2
+configure wave -gridoffset 0
+configure wave -gridperiod 1
+configure wave -griddelta 40
+configure wave -timeline 0
+configure wave -timelineunits ns
+update
+WaveRestoreZoom {226663267749 ps} {227016368236 ps}
diff --git a/idma_tests/idma_zeromem/zeromem.c b/idma_tests/idma_zeromem/zeromem.c
new file mode 100644
index 0000000..993e457
--- /dev/null
+++ b/idma_tests/idma_zeromem/zeromem.c
@@ -0,0 +1,81 @@
+#include <math.h>
+#include "pulp.h"
+//#include "mchan_tests.h"
+
+#define VERBOSE
+
+#define MAX_BUFFER_SIZE 0x2000
+
+//static unsigned char *ext;
+//static unsigned char *loc;
+
+//#define EXT_DATA_ADDR  ((unsigned int)ext)
+//#define TCDM_DATA_ADDR ((unsigned int)loc)
+
+L2_DATA static unsigned char ext[MAX_BUFFER_SIZE];
+L1_DATA static unsigned char loc[MAX_BUFFER_SIZE];
+
+#define EXT_DATA_ADDR  ((unsigned int) ext)
+#define TCDM_DATA_ADDR ((unsigned int) loc)
+typedef enum {L1_TEST, L2_TEST} test_type_t;
+
+int zeromem_test(unsigned int len, test_type_t type, unsigned int buf);
+int main()
+{
+
+  if (rt_cluster_id() != 0)
+    return bench_cluster_forward(0);
+
+
+  int error_count_l2 = 0;
+  int error_count_l1 = 0;
+
+
+
+  if (get_core_id() == 0){
+
+
+    for ( int i = 5; i < 8045; i=4*i) {
+      error_count_l1 += zeromem_test(i, L1_TEST, loc);
+    }
+
+    if (error_count_l1)
+      printf("OOPS -- got %d errors in L1 zeromem tests!\n", error_count_l1);
+    else
+      printf("L1 zeromem tests passed!\n");
+    for ( int i = 5; i < 8045; i=4*i ) {
+      error_count_l2 += zeromem_test(i, L2_TEST, ext);
+    }
+    if (error_count_l2)
+      printf("OOPS -- got %d errors in L2 zeromem tests!\n", error_count_l2);
+    else
+      printf("L2 zeromem tests passed!\n");
+
+  }
+
+  return error_count_l1 + error_count_l2;
+}
+
+
+int zeromem_test(unsigned int len, test_type_t type, unsigned int buf) {
+  unsigned int tx_id;
+  int error_cnt = 0;
+  uint8_t * buf_ptr = (uint8_t *) buf;
+  // fill the buffer with data to make sure it gets erased
+  for (int i=0; i<len; i++)
+    buf_ptr[i] = i & 0xFF;
+
+  if (type == L1_TEST) {
+    tx_id = pulp_idma_zeromem(buf, len, IDMA_PROT_OBI);
+  } else {
+    tx_id = pulp_idma_zeromem(buf, len, IDMA_PROT_AXI);
+  }
+  plp_dma_barrier();
+  for (int i=0; i<len; i++) {
+    if (buf_ptr[i] != 0) {
+      error_cnt++;
+      printf("Error in %s test at element %d: expected 0, got 0x%x\n", type==L1_TEST ? "L1" : "L2", i, buf_ptr[i]);
+    }
+  }
+  return error_cnt;
+}
diff --git a/mchan_tests/testMCHAN_2D_ext_tcdm/mchan_tests.h b/mchan_tests/testMCHAN_2D_ext_tcdm/mchan_tests.h
index 254df64..3cd8117 100644
--- a/mchan_tests/testMCHAN_2D_ext_tcdm/mchan_tests.h
+++ b/mchan_tests/testMCHAN_2D_ext_tcdm/mchan_tests.h
@@ -10,14 +10,14 @@
 #define LIN 0
 #define TWD 1
 
-#define MCHAN_COMMAND_QUEUE    0x10204400 //0x10201800
-#define MCHAN_STATUS_REGISTER  0x10204404 //0x10201804
+#define MCHAN_COMMAND_QUEUE    0x50204400 //0x10201800
+#define MCHAN_STATUS_REGISTER  0x50204404 //0x10201804
 
 // TEMPORARY FIX DAVIDE
 #define PLP_DMA_2D_TCDM_BIT 22
 
-#define PE_MCHAN_COMMAND_QUEUE   0x10201C00
-#define PE_MCHAN_STATUS_REGISTER 0x10201C04
+#define PE_MCHAN_COMMAND_QUEUE   0x50201C00
+#define PE_MCHAN_STATUS_REGISTER 0x50201C04
 
 #define PLP_DMA_TYPE_BIT  0x00000011
 #define PLP_DMA_INCR_BIT 0x00000012
diff --git a/mchan_tests/testMCHAN_4k_crossing/mchan_tests.h b/mchan_tests/testMCHAN_4k_crossing/mchan_tests.h
index 254df64..3cd8117 100644
--- a/mchan_tests/testMCHAN_4k_crossing/mchan_tests.h
+++ b/mchan_tests/testMCHAN_4k_crossing/mchan_tests.h
@@ -10,14 +10,14 @@
 #define LIN 0
 #define TWD 1
 
-#define MCHAN_COMMAND_QUEUE    0x10204400 //0x10201800
-#define MCHAN_STATUS_REGISTER  0x10204404 //0x10201804
+#define MCHAN_COMMAND_QUEUE    0x50204400 //0x10201800
+#define MCHAN_STATUS_REGISTER  0x50204404 //0x10201804
 
 // TEMPORARY FIX DAVIDE
 #define PLP_DMA_2D_TCDM_BIT 22
 
-#define PE_MCHAN_COMMAND_QUEUE   0x10201C00
-#define PE_MCHAN_STATUS_REGISTER 0x10201C04
+#define PE_MCHAN_COMMAND_QUEUE   0x50201C00
+#define PE_MCHAN_STATUS_REGISTER 0x50201C04
 
 #define PLP_DMA_TYPE_BIT  0x00000011
 #define PLP_DMA_INCR_BIT 0x00000012
diff --git a/mchan_tests/testMCHAN_TCDM2TCDM_tx_rx/mchan_tests.h b/mchan_tests/testMCHAN_TCDM2TCDM_tx_rx/mchan_tests.h
index 254df64..3cd8117 100644
--- a/mchan_tests/testMCHAN_TCDM2TCDM_tx_rx/mchan_tests.h
+++ b/mchan_tests/testMCHAN_TCDM2TCDM_tx_rx/mchan_tests.h
@@ -10,14 +10,14 @@
 #define LIN 0
 #define TWD 1
 
-#define MCHAN_COMMAND_QUEUE    0x10204400 //0x10201800
-#define MCHAN_STATUS_REGISTER  0x10204404 //0x10201804
+#define MCHAN_COMMAND_QUEUE    0x50204400 //0x10201800
+#define MCHAN_STATUS_REGISTER  0x50204404 //0x10201804
 
 // TEMPORARY FIX DAVIDE
 #define PLP_DMA_2D_TCDM_BIT 22
 
-#define PE_MCHAN_COMMAND_QUEUE   0x10201C00
-#define PE_MCHAN_STATUS_REGISTER 0x10201C04
+#define PE_MCHAN_COMMAND_QUEUE   0x50201C00
+#define PE_MCHAN_STATUS_REGISTER 0x50201C04
 
 #define PLP_DMA_TYPE_BIT  0x00000011
 #define PLP_DMA_INCR_BIT 0x00000012
diff --git a/mchan_tests/testMCHAN_basic/mchan_tests.h b/mchan_tests/testMCHAN_basic/mchan_tests.h
index 254df64..3cd8117 100644
--- a/mchan_tests/testMCHAN_basic/mchan_tests.h
+++ b/mchan_tests/testMCHAN_basic/mchan_tests.h
@@ -10,14 +10,14 @@
 #define LIN 0
 #define TWD 1
 
-#define MCHAN_COMMAND_QUEUE    0x10204400 //0x10201800
-#define MCHAN_STATUS_REGISTER  0x10204404 //0x10201804
+#define MCHAN_COMMAND_QUEUE    0x50204400 //0x10201800
+#define MCHAN_STATUS_REGISTER  0x50204404 //0x10201804
 
 // TEMPORARY FIX DAVIDE
 #define PLP_DMA_2D_TCDM_BIT 22
 
-#define PE_MCHAN_COMMAND_QUEUE   0x10201C00
-#define PE_MCHAN_STATUS_REGISTER 0x10201C04
+#define PE_MCHAN_COMMAND_QUEUE   0x50201C00
+#define PE_MCHAN_STATUS_REGISTER 0x50201C04
 
 #define PLP_DMA_TYPE_BIT  0x00000011
 #define PLP_DMA_INCR_BIT 0x00000012
diff --git a/mchan_tests/testMCHAN_basic_8cores/mchan_tests.h b/mchan_tests/testMCHAN_basic_8cores/mchan_tests.h
index 8865d98..28d412a 100755
--- a/mchan_tests/testMCHAN_basic_8cores/mchan_tests.h
+++ b/mchan_tests/testMCHAN_basic_8cores/mchan_tests.h
@@ -10,14 +10,14 @@
 #define LIN 0
 #define TWD 1
 
-#define MCHAN_COMMAND_QUEUE    0x10204400 //0x10201800
-#define MCHAN_STATUS_REGISTER  0x10204404 //0x10201804
+#define MCHAN_COMMAND_QUEUE    0x50204400 //0x10201800
+#define MCHAN_STATUS_REGISTER  0x50204404 //0x10201804
 
 // TEMPORARY FIX DAVIDE
 #define PLP_DMA_2D_TCDM_BIT 22
 
-#define PE_MCHAN_COMMAND_QUEUE   0x10201800 //0x10201C00
-#define PE_MCHAN_STATUS_REGISTER 0x10201804 //0x10201C04
+#define PE_MCHAN_COMMAND_QUEUE   0x50201800 //0x10201C00
+#define PE_MCHAN_STATUS_REGISTER 0x50201804 //0x10201C04
 
 #define PLP_DMA_TYPE_BIT  0x00000011
 #define PLP_DMA_INCR_BIT 0x00000012
diff --git a/mchan_tests/testMCHAN_basic_FC_TCDM/mchan_tests.h b/mchan_tests/testMCHAN_basic_FC_TCDM/mchan_tests.h
index 254df64..3cd8117 100644
--- a/mchan_tests/testMCHAN_basic_FC_TCDM/mchan_tests.h
+++ b/mchan_tests/testMCHAN_basic_FC_TCDM/mchan_tests.h
@@ -10,14 +10,14 @@
 #define LIN 0
 #define TWD 1
 
-#define MCHAN_COMMAND_QUEUE    0x10204400 //0x10201800
-#define MCHAN_STATUS_REGISTER  0x10204404 //0x10201804
+#define MCHAN_COMMAND_QUEUE    0x50204400 //0x10201800
+#define MCHAN_STATUS_REGISTER  0x50204404 //0x10201804
 
 // TEMPORARY FIX DAVIDE
 #define PLP_DMA_2D_TCDM_BIT 22
 
-#define PE_MCHAN_COMMAND_QUEUE   0x10201C00
-#define PE_MCHAN_STATUS_REGISTER 0x10201C04
+#define PE_MCHAN_COMMAND_QUEUE   0x50201C00
+#define PE_MCHAN_STATUS_REGISTER 0x50201C04
 
 #define PLP_DMA_TYPE_BIT  0x00000011
 #define PLP_DMA_INCR_BIT 0x00000012
diff --git a/mchan_tests/testMCHAN_not_incremental/mchan_tests.h b/mchan_tests/testMCHAN_not_incremental/mchan_tests.h
index 254df64..3cd8117 100644
--- a/mchan_tests/testMCHAN_not_incremental/mchan_tests.h
+++ b/mchan_tests/testMCHAN_not_incremental/mchan_tests.h
@@ -10,14 +10,14 @@
 #define LIN 0
 #define TWD 1
 
-#define MCHAN_COMMAND_QUEUE    0x10204400 //0x10201800
-#define MCHAN_STATUS_REGISTER  0x10204404 //0x10201804
+#define MCHAN_COMMAND_QUEUE    0x50204400 //0x10201800
+#define MCHAN_STATUS_REGISTER  0x50204404 //0x10201804
 
 // TEMPORARY FIX DAVIDE
 #define PLP_DMA_2D_TCDM_BIT 22
 
-#define PE_MCHAN_COMMAND_QUEUE   0x10201C00
-#define PE_MCHAN_STATUS_REGISTER 0x10201C04
+#define PE_MCHAN_COMMAND_QUEUE   0x50201C00
+#define PE_MCHAN_STATUS_REGISTER 0x50201C04
 
 #define PLP_DMA_TYPE_BIT  0x00000011
 #define PLP_DMA_INCR_BIT 0x00000012
diff --git a/mchan_tests/testMCHAN_pe_basic/mchan_tests.h b/mchan_tests/testMCHAN_pe_basic/mchan_tests.h
index 254df64..3cd8117 100644
--- a/mchan_tests/testMCHAN_pe_basic/mchan_tests.h
+++ b/mchan_tests/testMCHAN_pe_basic/mchan_tests.h
@@ -10,14 +10,14 @@
 #define LIN 0
 #define TWD 1
 
-#define MCHAN_COMMAND_QUEUE    0x10204400 //0x10201800
-#define MCHAN_STATUS_REGISTER  0x10204404 //0x10201804
+#define MCHAN_COMMAND_QUEUE    0x50204400 //0x10201800
+#define MCHAN_STATUS_REGISTER  0x50204404 //0x10201804
 
 // TEMPORARY FIX DAVIDE
 #define PLP_DMA_2D_TCDM_BIT 22
 
-#define PE_MCHAN_COMMAND_QUEUE   0x10201C00
-#define PE_MCHAN_STATUS_REGISTER 0x10201C04
+#define PE_MCHAN_COMMAND_QUEUE   0x50201C00
+#define PE_MCHAN_STATUS_REGISTER 0x50201C04
 
 #define PLP_DMA_TYPE_BIT  0x00000011
 #define PLP_DMA_INCR_BIT 0x00000012
diff --git a/mchan_tests/testMCHAN_pe_fc_basic/mchan_tests.h b/mchan_tests/testMCHAN_pe_fc_basic/mchan_tests.h
index 254df64..3cd8117 100644
--- a/mchan_tests/testMCHAN_pe_fc_basic/mchan_tests.h
+++ b/mchan_tests/testMCHAN_pe_fc_basic/mchan_tests.h
@@ -10,14 +10,14 @@
 #define LIN 0
 #define TWD 1
 
-#define MCHAN_COMMAND_QUEUE    0x10204400 //0x10201800
-#define MCHAN_STATUS_REGISTER  0x10204404 //0x10201804
+#define MCHAN_COMMAND_QUEUE    0x50204400 //0x10201800
+#define MCHAN_STATUS_REGISTER  0x50204404 //0x10201804
 
 // TEMPORARY FIX DAVIDE
 #define PLP_DMA_2D_TCDM_BIT 22
 
-#define PE_MCHAN_COMMAND_QUEUE   0x10201C00
-#define PE_MCHAN_STATUS_REGISTER 0x10201C04
+#define PE_MCHAN_COMMAND_QUEUE   0x50201C00
+#define PE_MCHAN_STATUS_REGISTER 0x50201C04
 
 #define PLP_DMA_TYPE_BIT  0x00000011
 #define PLP_DMA_INCR_BIT 0x00000012
diff --git a/mchan_tests/testMCHAN_unaligned/mchan_tests.h b/mchan_tests/testMCHAN_unaligned/mchan_tests.h
index 254df64..3cd8117 100644
--- a/mchan_tests/testMCHAN_unaligned/mchan_tests.h
+++ b/mchan_tests/testMCHAN_unaligned/mchan_tests.h
@@ -10,14 +10,14 @@
 #define LIN 0
 #define TWD 1
 
-#define MCHAN_COMMAND_QUEUE    0x10204400 //0x10201800
-#define MCHAN_STATUS_REGISTER  0x10204404 //0x10201804
+#define MCHAN_COMMAND_QUEUE    0x50204400 //0x10201800
+#define MCHAN_STATUS_REGISTER  0x50204404 //0x10201804
 
 // TEMPORARY FIX DAVIDE
 #define PLP_DMA_2D_TCDM_BIT 22
 
-#define PE_MCHAN_COMMAND_QUEUE   0x10201C00
-#define PE_MCHAN_STATUS_REGISTER 0x10201C04
+#define PE_MCHAN_COMMAND_QUEUE   0x50201C00
+#define PE_MCHAN_STATUS_REGISTER 0x50201C04
 
 #define PLP_DMA_TYPE_BIT  0x00000011
 #define PLP_DMA_INCR_BIT 0x00000012
diff --git a/neureka/app/Makefile b/neureka/app/Makefile
new file mode 100644
index 0000000..b28b359
--- /dev/null
+++ b/neureka/app/Makefile
@@ -0,0 +1,60 @@
+# Copyright (C) 2022-2023 ETH Zurich and University of Bologna
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+# SPDX-License-Identifier: Apache-2.0
+#
+# Author: Yvan Tortorella (yvan.tortorella@unibo.it)
+#         Francesco Conti (f.conti@unibo.it)
+#		  Arpan Suravi Prasad (prasadar@iis.ee.ethz.ch)
+#
+
+
+LIBDIR := ../app
+ACCELERATOR := neureka
+STIM_DIR := ./
+
+INC_DIRS += $(LIBDIR)/inc
+
+ifndef OVERRIDE_MAIN  
+APP_SRCS += $(wildcard $(LIBDIR)/src/*.c)
+endif
+
+APP_SRCS += $(wildcard $(LIBDIR)/*.c)
+
+## Library
+INC_DIRS += $(LIBDIR) $(LIBDIR)/util
+APP_SRCS += $(wildcard $(LIBDIR)/util/*.c)
+
+## Accelerator
+INC_DIRS += $(LIBDIR)/hal $(LIBDIR)/gvsoc $(LIBDIR)/bsp $(LIBDIR)/bsp/testbench
+APP_SRCS += $(wildcard $(LIBDIR)/hal/*.c) $(wildcard $(LIBDIR)/gvsoc/*.c) $(wildcard $(LIBDIR)/bsp/testbench/*.c)
+
+
+## Generated 
+INC_DIRS += inc
+APP_SRCS += $(wildcard src/*.c)
+
+INC_FLAGS += $(addprefix -I,$(INC_DIRS))
+
+# Flags
+ACCELERATOR_UPPERCASE := $(shell echo $(ACCELERATOR) | tr [:lower:] [:upper:])
+APP_CFLAGS += -DNNX_ACCELERATOR=\"$(ACCELERATOR)\" -DNNX_$(ACCELERATOR_UPPERCASE) -DNNX_NEUREKA_TESTBENCH -DNNX_NEUREKA_PE_H=4 -DNNX_NEUREKA_PE_W=4
+APP_CFLAGS += $(INC_FLAGS)
+
+PULP_CFLAGS = $(APP_CFLAGS)
+
+PULP_APP = test
+PULP_APP_FC_SRCS = $(APP_SRCS)
+PULP_APP_HOST_SRCS = $(APP_SRCS)
+
+include $(PULP_SDK_HOME)/install/rules/pulp_rt.mk
\ No newline at end of file
diff --git a/neureka/app/bsp/neureka_bsp.h b/neureka/app/bsp/neureka_bsp.h
new file mode 100644
index 0000000..41df5c7
--- /dev/null
+++ b/neureka/app/bsp/neureka_bsp.h
@@ -0,0 +1,13 @@
+#ifdef NNX_NEUREKA_TESTBENCH
+    #include "neureka_testbench_bsp.h"
+    #define neureka_bsp_conf_t neureka_testbench_conf_t
+    #define neureka_bsp_open neureka_testbench_open
+    #define neureka_bsp_close neureka_testbench_close
+    #define neureka_bsp_event_wait_and_clear neureka_testbench_event_wait_and_clear
+#else
+    #include "neureka_siracusa_bsp.h"
+    #define neureka_bsp_conf_t neureka_siracusa_conf_t
+    #define neureka_bsp_open neureka_siracusa_open
+    #define neureka_bsp_close neureka_siracusa_close
+    #define neureka_bsp_event_wait_and_clear neureka_siracusa_event_wait_and_clear
+#endif
\ No newline at end of file
diff --git a/neureka/app/bsp/siracusa/neureka_siracusa_bsp.c b/neureka/app/bsp/siracusa/neureka_siracusa_bsp.c
new file mode 100644
index 0000000..57136fd
--- /dev/null
+++ b/neureka/app/bsp/siracusa/neureka_siracusa_bsp.c
@@ -0,0 +1,78 @@
+/*
+ * Luka Macan <luka.macan@unibo.it>
+ *
+ * Copyright 2023 ETH Zurich and University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ *
+ * SPDX-License-Identifier: Apache-2.0
+ */
+
+#include "neureka_siracusa_bsp.h"
+#include <pmsis.h>
+
+#define NEUREKA_SIRACUSA_CLUSTER_CTRL_BASE_ADDR (0x00200000)
+#define NEUREKA_SIRACUSA_CLUSTER_CTRL_HWPE_OFFS 0x18
+#define NEUREKA_SIRACUSA_CLUSTER_CTRL_HWPE_ADDR                                \
+  (NEUREKA_SIRACUSA_CLUSTER_CTRL_BASE_ADDR +                                   \
+   NEUREKA_SIRACUSA_CLUSTER_CTRL_HWPE_OFFS)
+#define NEUREKA_SIRACUSA_CLUSTER_CTRL_HWPE_MASK_HCI_PRIO 0x100
+#define NEUREKA_SIRACUSA_CLUSTER_CTRL_HWPE_MASK_HCI_MAXSTALL 0xff
+#define NEUREKA_SIRACUSA_MAX_STALL (8)
+#define NEUREKA_SIRACUSA_EVENT (1 << 12)
+#define NEUREKA_SIRACUSA_BASE_ADDR (0x00201000)
+#define NEUREKA_SIRACUSA_WEIGHT_MEM_BASE_ADDR (0x10400000)
+#define NEUREKA_SIRACUSA_WEIGHT_MEM_MRAM_OFFSET (0x00000000)
+#define NEUREKA_SIRACUSA_WEIGHT_MEM_SRAM_OFFSET (0x00400000)
+
+void neureka_siracusa_hci_setpriority_neureka() {
+  *(volatile uint32_t *)NEUREKA_SIRACUSA_CLUSTER_CTRL_HWPE_ADDR |=
+      NEUREKA_SIRACUSA_CLUSTER_CTRL_HWPE_MASK_HCI_PRIO;
+}
+
+void neureka_siracusa_hci_setpriority_core() {
+  *(volatile uint32_t *)NEUREKA_SIRACUSA_CLUSTER_CTRL_HWPE_ADDR &=
+      ~NEUREKA_SIRACUSA_CLUSTER_CTRL_HWPE_MASK_HCI_PRIO;
+}
+
+void neureka_siracusa_hci_reset_max_stall() {
+  *(volatile uint32_t *)NEUREKA_SIRACUSA_CLUSTER_CTRL_HWPE_ADDR &=
+      ~NEUREKA_SIRACUSA_CLUSTER_CTRL_HWPE_MASK_HCI_MAXSTALL;
+}
+
+void neureka_siracusa_hci_set_max_stall(uint32_t max_stall) {
+  *(volatile uint32_t *)NEUREKA_SIRACUSA_CLUSTER_CTRL_HWPE_ADDR |=
+      max_stall & NEUREKA_SIRACUSA_CLUSTER_CTRL_HWPE_MASK_HCI_MAXSTALL;
+}
+
+void neureka_siracusa_open(neureka_siracusa_conf_t *conf) {
+  neureka_siracusa_hci_setpriority_neureka();
+  neureka_siracusa_hci_set_max_stall(conf->max_stall);
+}
+
+void neureka_siracusa_close() {
+  neureka_siracusa_hci_reset_max_stall();
+  neureka_siracusa_hci_setpriority_core();
+}
+
+void neureka_siracusa_event_wait_and_clear() {
+  eu_evt_maskWaitAndClr(NEUREKA_SIRACUSA_EVENT);
+}
+
+static const neureka_dev_t neureka_siracusa_dev = {
+    .hwpe_dev = (struct hwpe_dev_t){
+        .base_addr = (volatile uint32_t *)NEUREKA_SIRACUSA_BASE_ADDR}};
+
+const neureka_dev_t *neureka_siracusa_get_dev() {
+  return &neureka_siracusa_dev;
+}
diff --git a/neureka/app/bsp/siracusa/neureka_siracusa_bsp.h b/neureka/app/bsp/siracusa/neureka_siracusa_bsp.h
new file mode 100644
index 0000000..458fae7
--- /dev/null
+++ b/neureka/app/bsp/siracusa/neureka_siracusa_bsp.h
@@ -0,0 +1,69 @@
+/*
+ * Luka Macan <luka.macan@unibo.it>
+ *
+ * Copyright 2023 ETH Zurich and University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ *
+ * SPDX-License-Identifier: Apache-2.0
+ */
+
+#ifndef __NEUREKA_SIRACUSA_BSP_H__
+#define __NEUREKA_SIRACUSA_BSP_H__
+
+#include <pulp.h>
+
+#include "neureka.h"
+#include <stdint.h>
+
+/**
+ * neureka_siracusa_setpriority_neureka
+ *
+ * Set HCI interconnect bus priority to prioritize neureka.
+ */
+void neureka_siracusa_hci_setpriority_neureka();
+
+/**
+ * neureka_siracusa_setpriority_core
+ *
+ * Set HCI bus priority to prioritize cores.
+ */
+void neureka_siracusa_hci_setpriority_core();
+
+/**
+ * neureka_siracusa_hci_reset_maxstall
+ *
+ * Reset the HCI bus maxstall parameter.
+ * TODO: Check if it disables it also or just resets?
+ */
+void neureka_siracusa_hci_reset_max_stall();
+
+/**
+ * neureka_siracusa_hci_set_maxstall
+ *
+ * Set the HCI bus maxstall. Maxstall defines how many cycles
+ * will the HCI bus stall the lower priority master, i.e. neureka or core,
+ * before letting it do a transaction.
+ */
+void neureka_siracusa_hci_set_max_stall(uint32_t max_stall);
+
+typedef struct neureka_siracusa_conf_t {
+  int max_stall;
+} neureka_siracusa_conf_t;
+
+void neureka_siracusa_open(neureka_siracusa_conf_t *conf);
+void neureka_siracusa_close();
+void neureka_siracusa_event_wait_and_clear();
+const neureka_dev_t *neureka_siracusa_get_dev();
+
+#endif // !__NEUREKA_SIRACUSA_BSP_H__
diff --git a/neureka/app/bsp/testbench/neureka_testbench_bsp.c b/neureka/app/bsp/testbench/neureka_testbench_bsp.c
new file mode 100644
index 0000000..2a5d4a1
--- /dev/null
+++ b/neureka/app/bsp/testbench/neureka_testbench_bsp.c
@@ -0,0 +1,48 @@
+/*
+ * Luka Macan <luka.macan@unibo.it>
+ * Francesco Conti <f.conti@unibo.it>
+ * Arpan Suravi Prasad <prasadar@iis.ee.ethz.ch>
+ *
+ * Copyright 2023 ETH Zurich and University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ *
+ * SPDX-License-Identifier: Apache-2.0
+ */
+
+#include "neureka_testbench_bsp.h"
+#include <pmsis.h>
+
+#define NEUREKA_TESTBENCH_CLUSTER_CTRL_BASE_ADDR (0x00100000)
+#define NEUREKA_TESTBENCH_BASE_ADDR (0x10201000)
+#define NEUREKA_TESTBENCH_EVENT (1 << 12)
+
+void neureka_testbench_open(neureka_testbench_conf_t *conf) {
+  return;
+}
+
+void neureka_testbench_close() {
+  return;
+}
+
+void neureka_testbench_event_wait_and_clear() {
+  eu_evt_maskWaitAndClr(NEUREKA_TESTBENCH_EVENT);
+}
+
+static const neureka_dev_t neureka_testbench_dev = {
+    .hwpe_dev = (struct hwpe_dev_t){
+        .base_addr = (volatile uint32_t *)NEUREKA_TESTBENCH_BASE_ADDR}};
+
+const neureka_dev_t *neureka_testbench_get_dev() {
+  return &neureka_testbench_dev;
+}
diff --git a/neureka/app/bsp/testbench/neureka_testbench_bsp.h b/neureka/app/bsp/testbench/neureka_testbench_bsp.h
new file mode 100644
index 0000000..a6d4bd1
--- /dev/null
+++ b/neureka/app/bsp/testbench/neureka_testbench_bsp.h
@@ -0,0 +1,37 @@
+/*
+ * Luka Macan <luka.macan@unibo.it>
+ *
+ * Copyright 2023 ETH Zurich and University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ *
+ * SPDX-License-Identifier: Apache-2.0
+ */
+
+#ifndef __NEUREKA_TESTBENCH_BSP_H__
+#define __NEUREKA_TESTBENCH_BSP_H__
+
+#include "pulp.h"
+#include "neureka.h"
+#include <stdint.h>
+
+typedef struct neureka_testbench_conf_t {
+  int max_stall;
+} neureka_testbench_conf_t;
+
+void neureka_testbench_open(neureka_testbench_conf_t *conf);
+void neureka_testbench_close();
+void neureka_testbench_event_wait_and_clear();
+const neureka_dev_t *neureka_testbench_get_dev();
+
+#endif // !__NEUREKA_TESTBENCH_BSP_H__
diff --git a/neureka/app/gvsoc/neureka_gvsoc.h b/neureka/app/gvsoc/neureka_gvsoc.h
new file mode 100644
index 0000000..37eeab0
--- /dev/null
+++ b/neureka/app/gvsoc/neureka_gvsoc.h
@@ -0,0 +1,54 @@
+/*
+ * Luka Macan <luka.macan@unibo.it>
+ *
+ * Copyright 2023 ETH Zurich and University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ *
+ * SPDX-License-Identifier: Apache-2.0
+ */
+
+#ifndef __NEUREKA_GVSOC_H__
+#define __NEUREKA_GVSOC_H__
+
+#include "neureka.h"
+#include "neureka_task.h"
+
+#define NEUREKA_REG_GVSOC_LOG_LEVEL 24
+#define NEUREKA_REG_GVSOC_LOG_FORMAT 25
+
+typedef enum neureka_gvsoc_log_format_e {
+  NEUREKA_GVSOC_LOG_FORMAT_DECIMAL = 0,
+  NEUREKA_GVSOC_LOG_FORMAT_HEXADECIMAL = 3
+} neureka_gvsoc_log_format_e;
+
+typedef enum neureka_gvsoc_log_level_e {
+  NEUREKA_GVSOC_LOG_LEVEL_JOB_START_END = 0,
+  NEUREKA_GVSOC_LOG_LEVEL_CONFIG = 1,
+  NEUREKA_GVSOC_LOG_LEVEL_ACTIV_INOUT = 2,
+  NEUREKA_GVSOC_LOG_LEVEL_ALL = 3
+} neureka_gvsoc_log_level_e;
+
+static void neureka_gvsoc_log_activate(neureka_dev_t *dev,
+                                       neureka_gvsoc_log_level_e log_level,
+                                       neureka_gvsoc_log_format_e format) {
+  hwpe_task_reg_write(&dev->hwpe_dev, NEUREKA_REG_GVSOC_LOG_LEVEL, log_level);
+  hwpe_task_reg_write(&dev->hwpe_dev, NEUREKA_REG_GVSOC_LOG_FORMAT, format);
+}
+
+static void neureka_gvsoc_log_deactivate(neureka_dev_t *dev) {
+  hwpe_task_reg_write(&dev->hwpe_dev, NEUREKA_REG_GVSOC_LOG_LEVEL,
+                      NEUREKA_GVSOC_LOG_LEVEL_JOB_START_END);
+}
+
+#endif // __NEUREKA_GVSOC_H__
diff --git a/neureka/app/hal/neureka.c b/neureka/app/hal/neureka.c
new file mode 100644
index 0000000..dc829d9
--- /dev/null
+++ b/neureka/app/hal/neureka.c
@@ -0,0 +1,37 @@
+/*
+ * Luka Macan <luka.macan@unibo.it>
+ *
+ * Copyright 2023 ETH Zurich and University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ *
+ * SPDX-License-Identifier: Apache-2.0
+ */
+
+#include "neureka.h"
+
+#define NEUREKA_STATUS_EMPTY (0x000)
+#define NEUREKA_STATUS_FULL (0x101)
+
+inline int neureka_task_queue_tasks_in_flight(neureka_dev_t *dev) {
+  uint32_t status = hwpe_task_queue_status(&dev->hwpe_dev);
+  return (status & 0x1) + ((status >> 8) & 0x1);
+}
+
+inline int neureka_task_queue_empty(neureka_dev_t *dev) {
+  return hwpe_task_queue_status(&dev->hwpe_dev) == NEUREKA_STATUS_EMPTY;
+}
+
+inline int neureka_task_queue_full(neureka_dev_t *dev) {
+  return hwpe_task_queue_status(&dev->hwpe_dev) == NEUREKA_STATUS_FULL;
+}
diff --git a/neureka/app/hal/neureka.h b/neureka/app/hal/neureka.h
new file mode 100644
index 0000000..eae77a1
--- /dev/null
+++ b/neureka/app/hal/neureka.h
@@ -0,0 +1,37 @@
+/*
+ * Luka Macan <luka.macan@unibo.it>
+ *
+ * Copyright 2023 ETH Zurich and University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ *
+ * SPDX-License-Identifier: Apache-2.0
+ */
+
+#ifndef __NEUREKA_H__
+#define __NEUREKA_H__
+
+#include "hwpe.h"
+#include <stdint.h>
+
+#define NEUREKA_TASK_QUEUE_SIZE (2)
+
+typedef struct neureka_dev_t {
+  hwpe_dev_t hwpe_dev; /* Implements the HWPE device interface */
+} neureka_dev_t;
+
+int neureka_task_queue_tasks_in_flight(neureka_dev_t *dev);
+int neureka_task_queue_empty(neureka_dev_t *dev);
+int neureka_task_queue_full(neureka_dev_t *dev);
+
+#endif // __NEUREKA_H__
diff --git a/neureka/app/hal/neureka_task.c b/neureka/app/hal/neureka_task.c
new file mode 100644
index 0000000..d31c934
--- /dev/null
+++ b/neureka/app/hal/neureka_task.c
@@ -0,0 +1,248 @@
+/*
+ * Luka Macan <luka.macan@unibo.it>
+ *
+ * Copyright 2023 ETH Zurich and University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ *
+ * SPDX-License-Identifier: Apache-2.0
+ */
+
+#include "neureka_task.h"
+#include "neureka_task_defs.h"
+#include "pulp_nnx_util.h"
+
+uint32_t neureka_get_tile_padding(uint32_t padding, uint32_t i_height,
+                                  uint32_t i_width, uint32_t n_height,
+                                  uint32_t n_width) {
+  uint32_t tile_padding = padding;
+  if (i_height > 0) {
+    tile_padding &= ~(0xf << 28);
+  }
+  if (i_width < n_width - 1) {
+    tile_padding &= ~(0xf << 24);
+  }
+  if (i_height < n_height - 1) {
+    tile_padding &= ~(0xf << 20);
+  }
+  if (i_width > 0) {
+    tile_padding &= ~(0xf << 16);
+  }
+  return tile_padding;
+}
+
+void neureka_task_init(neureka_task_t *task) {
+  *task = (neureka_task_t){.data = {0}};
+}
+
+void neureka_task_set_op_to_conv(neureka_task_t *task,
+                                 const uint8_t kernel_shape,
+                                 const uint8_t depthwise) {
+  task->depthwise = depthwise;
+  task->kernel_shape = kernel_shape;
+  task->subtile_output_channel = depthwise ? NEUREKA_SUBTILE_INPUT_CHANNEL_3x3
+                                           : NEUREKA_SUBTILE_OUTPUT_CHANNEL;
+  task->subtile_input_channel = kernel_shape == 3
+                                    ? NEUREKA_SUBTILE_INPUT_CHANNEL_3x3
+                                    : NEUREKA_SUBTILE_INPUT_CHANNEL_1x1;
+
+  const int flag_mode = kernel_shape == 1 ? NEUREKA_FLAG_MODE_1x1
+                        : depthwise == 1  ? NEUREKA_FLAG_MODE_3x3_DW
+                                          : NEUREKA_FLAG_MODE_3x3;
+
+  task->data.cfg.conf0 &= ~(NEUREKA_MASK_FLAG_MODE);
+  task->data.cfg.conf0 |= flag_mode;
+}
+
+void neureka_task_set_bits(neureka_task_t *task, const uint8_t input_bits,
+                           const uint8_t output_bits,
+                           const uint8_t weight_bits) {
+  neureka_quant_mode_e quantMode;
+  if (output_bits == 8) {
+    quantMode = quantMode8Bit;
+  } else {
+    quantMode = quantMode32Bit;
+  }
+
+  task->qw = weight_bits;
+  task->data.cfg.conf0 &=
+      ~(NEUREKA_MASK_QUANT_MODE | NEUREKA_MASK_FLAG_WEIGHT_BITS);
+  task->data.cfg.conf0 |= quantMode | (weight_bits - 1);
+}
+
+void neureka_task_set_norm_quant(neureka_task_t *task, neureka_quant_t quant,
+                                 neureka_norm_t norm) {
+  task->data.cfg.conf0 &=
+      ~(NEUREKA_MASK_QUANT_FUNCTION | NEUREKA_MASK_SHIFT_AMOUNT |
+        NEUREKA_MASK_NORM_MODE | NEUREKA_MASK_FLAG_NORM_BIAS |
+        NEUREKA_MASK_FLAG_NORM_SHIFT);
+  task->data.cfg.conf0 |= NEUREKA_FLAG_NORM_QUANT | quant.function |
+                          (quant.shift_amount << 16) | norm.mode |
+                          norm.flag_bias << NEUREKA_SHIFT_FLAG_NORM_BIAS |
+                          norm.flag_shift << NEUREKA_SHIFT_FLAG_NORM_SHIFT;
+}
+
+void neureka_task_set_weight_offset(
+    neureka_task_t *task, neureka_weight_offset_mode_e weight_offset_mode,
+    const int32_t weight_offset) {
+  task->data.cfg.conf0 &= ~NEUREKA_MASK_WEIGHT_OFFSET_MODE;
+  task->data.cfg.conf0 |= weight_offset_mode;
+  task->data.cfg.weight_offset_factor = weight_offset;
+}
+
+void neureka_task_set_input_signed(neureka_task_t *task) {
+  task->data.cfg.conf0 |= NEUREKA_FLAG_INPUT_SIGNED;
+}
+
+void neureka_task_set_input_unsigned(neureka_task_t *task) {
+  task->data.cfg.conf0 &= ~NEUREKA_FLAG_INPUT_SIGNED;
+}
+
+void neureka_task_set_weight_source(neureka_task_t *task,
+                                    neureka_weight_source_e weight_source) {
+  task->data.cfg.conf0 &= ~NEUREKA_MASK_FLAG_WEIGHT_SOURCE;
+  task->data.cfg.conf0 |= weight_source;
+}
+
+void neureka_task_set_activation_prefetch(neureka_task_t *task,
+                                    neureka_activation_prefetch_e activation_prefetch) {
+  task->data.cfg.conf0 &= ~NEUREKA_MASK_FLAG_ACTIVATION_PREFETCH;
+  task->data.cfg.conf0 |= activation_prefetch;
+}
+
+/** neureka_pad_ptr
+ *
+ * Calculate the pointer to the start of the ptr as if
+ * it was the start to the padded data.
+ * Necessary for input pointer when it's padded.
+ */
+uint32_t neureka_pad_ptr(uint32_t ptr, const uint32_t width,
+                         const uint32_t width_stride, const uint8_t padding_top,
+                         const uint8_t padding_left) {
+  return ptr - (padding_top * width + padding_left) * width_stride;
+}
+
+void neureka_task_set_ptrs_conv(neureka_task_t *task, uint32_t input_ptr,
+                                uint32_t w_in, uint32_t w_in_stride,
+                                uint8_t padding_top, uint8_t padding_left,
+                                uint32_t output_ptr, uint32_t weights_ptr) {
+  task->data.infeat_ptr =
+      neureka_pad_ptr(input_ptr, w_in, w_in_stride, padding_top, padding_left);
+  task->data.outfeat_ptr = output_ptr;
+  task->data.weights_ptr = weights_ptr;
+}
+
+void neureka_task_set_ptrs_norm_quant(neureka_task_t *task, uint32_t scale_ptr,
+                                      uint32_t shift_ptr, uint32_t bias_ptr) {
+  task->data.scale_ptr = scale_ptr;
+  task->data.scale_shift_ptr = shift_ptr;
+  task->data.scale_bias_ptr = bias_ptr;
+}
+
+void neureka_task_set_strides(neureka_task_t *task, const uint32_t k_in,
+                              const uint32_t h_in_stride,
+                              const uint32_t w_in_stride,
+                              const uint32_t h_out_stride,
+                              const uint32_t w_out_stride) {
+  const uint32_t num_k_in =
+      nnx_calculate_number_of_tiles(k_in, task->subtile_input_channel);
+
+  const neureka_stride_t input_stride = {
+      .d0 = w_in_stride, .d1 = h_in_stride, .d2 = 0};
+  task->data.cfg.input_stride = input_stride;
+
+  const neureka_stride_t output_stride = {.d0 = NEUREKA_OUTPUT_BANDWIDTH_BYTES,
+                                          .d1 = w_out_stride,
+                                          .d2 = h_out_stride};
+  task->data.cfg.output_stride = output_stride;
+
+  if (task->kernel_shape == 1) { // 1x1
+    task->data.cfg.weights_stride.d0 = NEUREKA_WEIGHT_BANDWIDTH_BYTES_1x1;
+    task->data.cfg.weights_stride.d1 =
+        NEUREKA_WEIGHT_BANDWIDTH_BYTES_1x1 * num_k_in;
+  } else if (!task->depthwise) { // 3x3
+    task->data.cfg.weights_stride.d0 = NEUREKA_WEIGHT_BANDWIDTH_BYTES_3x3;
+    task->data.cfg.weights_stride.d1 =
+        NEUREKA_WEIGHT_BANDWIDTH_BYTES_3x3 * task->qw * num_k_in;
+  } else { // 3x3 depthwise
+    task->data.cfg.weights_stride.d0 = NEUREKA_WEIGHT_BANDWIDTH_BYTES_3x3;
+    task->data.cfg.weights_stride.d1 = 0;
+  }
+  task->data.cfg.weights_stride.d2 = 0;
+}
+
+void neureka_task_set_counters(neureka_task_t *task, const uint32_t k_in,
+                               const uint32_t h_out, const uint32_t w_out,
+                               const uint32_t k_out,
+                               const uint8_t padding_bottom,
+                               const uint8_t padding_right) {
+  const uint16_t num_Ko =
+      nnx_calculate_number_of_tiles(k_out, task->subtile_output_channel);
+  const uint16_t num_Ki =
+      nnx_calculate_number_of_tiles(k_in, task->subtile_input_channel);
+  const uint16_t num_Ho =
+      nnx_calculate_number_of_tiles(h_out, NEUREKA_SUBTILE_OUTPUT_HEIGHT);
+  const uint16_t num_Wo =
+      nnx_calculate_number_of_tiles(w_out, NEUREKA_SUBTILE_OUTPUT_WIDTH);
+
+  const uint16_t rem_Ko =
+      nnx_calculate_last_tile_size(k_out, task->subtile_output_channel);
+  const uint16_t rem_Ki =
+      nnx_calculate_last_tile_size(k_in, task->subtile_input_channel);
+  const uint16_t rem_Ho =
+      nnx_calculate_last_tile_size(h_out, NEUREKA_SUBTILE_OUTPUT_HEIGHT);
+  const uint16_t rem_Wo =
+      nnx_calculate_last_tile_size(w_out, NEUREKA_SUBTILE_OUTPUT_WIDTH);
+  const uint16_t rem_Hi =
+      (task->kernel_shape == 1 ? rem_Ho : rem_Ho + 2) - padding_bottom;
+  const uint16_t rem_Wi =
+      (task->kernel_shape == 1 ? rem_Wo : rem_Wo + 2) - padding_right;
+
+  const neureka_subtile_t subtile = {
+      .number = {.KoKi = nnx_concat_half(num_Ko, num_Ki),
+                 .HoWo = nnx_concat_half(num_Ho, num_Wo)},
+      .remainder = {.KoKi = nnx_concat_half(rem_Ko, rem_Ki),
+                    .HoWo = nnx_concat_half(rem_Ho, rem_Wo),
+                    .HiWi = nnx_concat_half(rem_Hi, rem_Wi)}};
+  task->data.cfg.subtile = subtile;
+}
+
+void neureka_task_set_padding(neureka_task_t *task, const uint8_t top,
+                              const uint8_t bottom, const uint8_t left,
+                              const uint8_t right, const uint8_t value) {
+  task->data.cfg.padding = ((top & 0xf) << 28) | ((right & 0xf) << 24) |
+                           ((bottom & 0xf) << 20) | ((left & 0xf) << 16) |
+                           (value & 0xff);
+}
+
+void neureka_task_set_mask_filter(neureka_task_t *task, const uint8_t top,
+                                  const uint8_t bottom, const uint8_t left,
+                                  const uint8_t right) {
+  task->data.cfg.filter_mask = ((top & 0xff) << 24) | ((right & 0xff) << 16) |
+                               ((bottom & 0xff) << 8) | ((left & 0xff) << 0);
+}
+
+void neureka_task_set_dims(
+    neureka_task_t *task, const uint32_t w_in, const uint32_t k_in,
+    const uint32_t h_in_stride, const uint32_t w_in_stride,
+    const uint32_t h_out, const uint32_t w_out, const uint32_t k_out,
+    const uint32_t h_out_stride, const uint32_t w_out_stride,
+    const uint8_t padding_top, const uint8_t padding_bottom,
+    const uint8_t padding_left, const uint8_t padding_right) {
+  neureka_task_set_strides(task, k_in, h_in_stride, w_in_stride, h_out_stride,
+                           w_out_stride);
+  neureka_task_set_counters(task, k_in, h_out, w_out, k_out, padding_bottom,
+                            padding_right);
+  neureka_task_set_padding(task, padding_top, padding_bottom, padding_left,
+                           padding_right, 0);
+}
diff --git a/neureka/app/hal/neureka_task.h b/neureka/app/hal/neureka_task.h
new file mode 100644
index 0000000..4022fc0
--- /dev/null
+++ b/neureka/app/hal/neureka_task.h
@@ -0,0 +1,192 @@
+/*
+ * Luka Macan <luka.macan@unibo.it>
+ *
+ * Copyright 2023 ETH Zurich and University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ *
+ * SPDX-License-Identifier: Apache-2.0
+ */
+
+#ifndef __NEUREKA_TASK_H__
+#define __NEUREKA_TASK_H__
+
+#include "neureka_task_defs.h"
+#include <stdint.h>
+
+typedef enum neureka_task_flag_e {
+  neurekaTaskFlagFalse = 0,
+  neurekaTaskFlagTrue = 1
+} neureka_task_flag_e;
+
+typedef enum neureka_weight_source_e {
+  neurekaWeightSourceTcdm = NEUREKA_FLAG_WEIGHT_SOURCE_TCDM,
+  neurekaWeightSourceWmem = NEUREKA_FLAG_WEIGHT_SOURCE_WMEM
+} neureka_weight_source_e;
+
+typedef enum neureka_weight_offset_mode_e {
+  weightOffsetModeSymmetric = NEUREKA_FLAG_WEIGHT_OFFSET_SYMMETRIC,
+  weightOffsetModeLayerWise = NEUREKA_FLAG_WEIGHT_OFFSET_LAYER_WISE
+} neureka_weight_offset_mode_e;
+
+typedef enum neureka_activation_prefetch_e {
+  activationPrefetchOn  = NEUREKA_FLAG_ACTIVATION_PREFETCH_ON,
+  activationPrefetchOff = NEUREKA_FLAG_ACTIVATION_PREFETCH_OFF
+} neureka_activation_prefetch_e;
+
+typedef enum {
+  normMode8Bit = NEUREKA_NORM_MODE_8BIT,
+  normMode32Bit = NEUREKA_NORM_MODE_32BIT
+} neureka_norm_mode_e;
+
+typedef struct neureka_norm_t {
+  neureka_norm_mode_e mode;
+  neureka_task_flag_e flag_bias;
+  neureka_task_flag_e flag_shift;
+} neureka_norm_t;
+
+typedef enum neureka_quant_mode_e {
+  quantMode8Bit = NEUREKA_QUANT_MODE_8BIT,
+  quantMode32Bit = NEUREKA_QUANT_MODE_32BIT
+} neureka_quant_mode_e;
+
+typedef enum neureka_quant_function_e {
+  quantFunctionIdentity = NEUREKA_FLAG_QUANT_FUNCTION_IDENTITY,
+  quantFunctionRelu = NEUREKA_FLAG_QUANT_FUNCTION_RELU
+} neureka_quant_function_e;
+
+typedef struct neureka_quant_t {
+  // Shift amount must be in range 0x00-0x1F
+  uint8_t shift_amount;
+  neureka_quant_function_e function;
+  neureka_task_flag_e flag_rounding;
+} neureka_quant_t;
+
+typedef struct neureka_stride_t {
+  uint32_t d0;
+  uint32_t d1;
+  uint32_t d2;
+} neureka_stride_t;
+
+typedef struct neureka_subtile_remainder_t {
+  uint32_t KoKi;
+  uint32_t HoWo;
+  uint32_t HiWi;
+} neureka_subtile_remainder_t;
+
+typedef struct neureka_subtile_number_t {
+  uint32_t KoKi;
+  uint32_t HoWo;
+} neureka_subtile_number_t;
+
+typedef struct neureka_subtile_t {
+  neureka_subtile_remainder_t remainder;
+  neureka_subtile_number_t number;
+} neureka_subtile_t;
+
+typedef struct neureka_cfg_t {
+  neureka_stride_t input_stride;
+  neureka_stride_t output_stride;
+  neureka_stride_t weights_stride;
+  neureka_subtile_t subtile;
+  uint32_t padding;
+  uint32_t weight_offset_factor;
+  uint32_t filter_mask;
+  uint32_t conf0;
+} neureka_cfg_t;
+
+typedef struct neureka_task_data_t {
+  uint32_t weights_ptr;
+  uint32_t infeat_ptr;
+  uint32_t outfeat_ptr;
+  uint32_t scale_ptr;
+  uint32_t scale_shift_ptr;
+  uint32_t scale_bias_ptr;
+  neureka_cfg_t cfg;
+} neureka_task_data_t;
+
+typedef struct neureka_task_t {
+  neureka_task_data_t data;
+  uint8_t qw;
+  uint8_t subtile_output_channel;
+  uint8_t subtile_input_channel;
+  uint8_t kernel_shape;
+  uint8_t depthwise;
+  uint8_t id;
+} neureka_task_t;
+
+void neureka_task_init(neureka_task_t *task);
+void neureka_task_set_op_to_conv(neureka_task_t *task,
+                                 const uint8_t kernel_shape,
+                                 const uint8_t depthwise);
+void neureka_task_set_bits(neureka_task_t *task, const uint8_t input_bits,
+                           const uint8_t output_bits,
+                           const uint8_t weight_bits);
+void neureka_task_set_norm_quant(neureka_task_t *task, neureka_quant_t quant,
+                                 neureka_norm_t norm);
+void neureka_task_set_weight_offset(
+    neureka_task_t *task, neureka_weight_offset_mode_e weight_offset_mode,
+    const int32_t weight_offset);
+void neureka_task_set_input_signed(neureka_task_t *task);
+void neureka_task_set_input_unsigned(neureka_task_t *task);
+void neureka_task_set_weight_source(neureka_task_t *task,
+                                    neureka_weight_source_e weight_source);
+uint32_t neureka_get_tile_padding(uint32_t padding, uint32_t i_height,
+                                  uint32_t i_width, uint32_t n_height,
+                                  uint32_t n_width);
+uint32_t neureka_pad_ptr(uint32_t ptr, const uint32_t width,
+                         const uint32_t width_stride, const uint8_t padding_top,
+                         const uint8_t padding_left);
+void neureka_task_set_ptrs_conv(neureka_task_t *task, uint32_t input_ptr,
+                                uint32_t w_in, uint32_t w_in_stride,
+                                uint8_t padding_top, uint8_t padding_left,
+                                uint32_t output_ptr, uint32_t weights_ptr);
+void neureka_task_set_ptrs_norm_quant(neureka_task_t *task, uint32_t scale_ptr,
+                                      uint32_t shift_ptr, uint32_t bias_ptr);
+/** neureka_task_set_strides
+ *
+ * All the strides variables are strides between elements alongside that
+ * dimension and expressed in bytes. There is no stride variable for the channel
+ * dimension because the N-EUREKA requires the channels to be contiguous.
+ */
+void neureka_task_set_strides(neureka_task_t *task, const uint32_t k_in,
+                              const uint32_t h_in_stride,
+                              const uint32_t w_in_stride,
+                              const uint32_t h_out_stride,
+                              const uint32_t w_out_stride);
+void neureka_task_set_counters(neureka_task_t *task, const uint32_t k_in,
+                               const uint32_t h_out, const uint32_t w_out,
+                               const uint32_t k_out,
+                               const uint8_t padding_bottom,
+                               const uint8_t padding_right);
+void neureka_task_set_padding(neureka_task_t *task, const uint8_t top,
+                              const uint8_t bottom, const uint8_t left,
+                              const uint8_t right, const uint8_t value);
+void neureka_task_set_mask_filter(neureka_task_t *task, const uint8_t top,
+                                  const uint8_t bottom, const uint8_t left,
+                                  const uint8_t right);
+/** neureka_task_set_dims
+ *
+ * All the strides variables are strides between elements alongside that
+ * dimension and expressed in bytes. There is no stride variable for the channel
+ * dimension because the N-EUREKA requires the channels to be contiguous.
+ */
+void neureka_task_set_dims(
+    neureka_task_t *task, const uint32_t w_in, const uint32_t k_in,
+    const uint32_t h_in_stride, const uint32_t w_in_stride,
+    const uint32_t h_out, const uint32_t w_out, const uint32_t k_out,
+    const uint32_t h_out_stride, const uint32_t w_out_stride,
+    const uint8_t padding_top, const uint8_t padding_bottom,
+    const uint8_t padding_left, const uint8_t padding_right);
+
+#endif // !__NEUREKA_TASK_H__
diff --git a/neureka/app/hal/neureka_task_defs.h b/neureka/app/hal/neureka_task_defs.h
new file mode 100644
index 0000000..7417347
--- /dev/null
+++ b/neureka/app/hal/neureka_task_defs.h
@@ -0,0 +1,131 @@
+/*
+ * Luka Macan <luka.macan@unibo.it>
+ *
+ * Copyright 2023 ETH Zurich and University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ *
+ * SPDX-License-Identifier: Apache-2.0
+ */
+
+#ifndef __NEUREKA_DEFS_H__
+#define __NEUREKA_DEFS_H__
+
+/* ARCHITECTURE */
+
+#define NNX_NEUREKA_PE_H (4)
+#define NNX_NEUREKA_PE_W (4)
+#define NNX_NEUREKA_BANDWIDTH_1x1 (256)
+#define NNX_NEUREKA_BANDWIDTH_3x3 (288)
+
+#define NEUREKA_SUBTILE_INPUT_HEIGHT_1x1 (NNX_NEUREKA_PE_H)
+#define NEUREKA_SUBTILE_INPUT_WIDTH_1x1  (NNX_NEUREKA_PE_W)
+#define NEUREKA_SUBTILE_INPUT_CHANNEL_1x1 (32)
+
+#define NEUREKA_SUBTILE_INPUT_HEIGHT_3x3 (NNX_NEUREKA_PE_H+2)
+#define NEUREKA_SUBTILE_INPUT_WIDTH_3x3  (NNX_NEUREKA_PE_W+2)
+#define NEUREKA_SUBTILE_INPUT_CHANNEL_3x3 (32)
+
+#define NEUREKA_SUBTILE_OUTPUT_HEIGHT (NNX_NEUREKA_PE_H)
+#define NEUREKA_SUBTILE_OUTPUT_WIDTH (NNX_NEUREKA_PE_W)
+#define NEUREKA_SUBTILE_OUTPUT_CHANNEL (32)
+
+#define NEUREKA_OUTPUT_BANDWIDTH_BYTES (32)
+#define NEUREKA_WEIGHT_BANDWIDTH_BYTES_1x1 (NNX_NEUREKA_BANDWIDTH_1x1/8)
+#define NEUREKA_WEIGHT_BANDWIDTH_BYTES_3x3 (NNX_NEUREKA_BANDWIDTH_3x3/8)
+
+/* TASK REGISTERS */
+
+// job configuration
+#define NEUREKA_REG_WEIGHTS_PTR 0
+#define NEUREKA_REG_INFEAT_PTR 1
+#define NEUREKA_REG_OUTFEAT_PTR 2
+#define NEUREKA_REG_SCALE_PTR 3
+#define NEUREKA_REG_SCALE_SHIFT_PTR 4
+#define NEUREKA_REG_SCALE_BIAS_PTR 5
+#define NEUREKA_REG_INFEAT_D0_STRIDE 6
+#define NEUREKA_REG_INFEAT_D1_STRIDE 7
+#define NEUREKA_REG_INFEAT_D2_STRIDE 8
+#define NEUREKA_REG_OUTFEAT_D0_STRIDE 9
+#define NEUREKA_REG_OUTFEAT_D1_STRIDE 10
+#define NEUREKA_REG_OUTFEAT_D2_STRIDE 11
+#define NEUREKA_REG_WEIGHTS_D0_STRIDE 12
+#define NEUREKA_REG_WEIGHTS_D1_STRIDE 13
+#define NEUREKA_REG_WEIGHTS_D2_STRIDE 14
+#define NEUREKA_REG_SUBTILE_REMAINDER_0 15
+#define NEUREKA_REG_SUBTILE_REMAINDER_1 16
+#define NEUREKA_REG_SUBTILE_REMAINDER_2 17
+#define NEUREKA_REG_SUBTILE_NUMBER_0 18
+#define NEUREKA_REG_SUBTILE_NUMBER_1 19
+#define NEUREKA_REG_PADDING 20
+#define NEUREKA_REG_WEIGHT_OFFSET_FACTOR 21
+#define NEUREKA_REG_FILTER_MASKING 22
+#define NEUREKA_REG_CONF0 23
+
+/*  SHIFT  */
+
+#define NEUREKA_SHIFT_FLAG_INPUT_SIGNED (26)
+#define NEUREKA_SHIFT_FLAG_NORM_BIAS (25)
+#define NEUREKA_SHIFT_FLAG_NORM_SHIFT (24)
+#define NEUREKA_SHIFT_QUANT_SHIFT (16)
+
+/*  CONF0 FLAGS */
+
+#define NEUREKA_FLAG_INPUT_SIGNED (1 << 26)
+#define NEUREKA_FLAG_NORM_BIAS (1 << 25)
+#define NEUREKA_FLAG_NORM_SHIFT (1 << 24)
+#define NEUREKA_FLAG_QUANT_FUNCTION_IDENTITY (1 << 23)
+#define NEUREKA_FLAG_QUANT_FUNCTION_RELU (0 << 23)
+#define NEUREKA_QUANT_MODE_8BIT (0 << 21)
+#define NEUREKA_QUANT_MODE_32BIT (2 << 21)
+// conf0[20:16] - quantization shift amount
+#define NEUREKA_FLAG_WEIGHT_OFFSET_SYMMETRIC (0 << 15) // Unimplemented in gvsoc
+#define NEUREKA_FLAG_WEIGHT_OFFSET_LAYER_WISE                                  \
+  (1 << 15) // Unimplemented in gvsoc
+#define NEUREKA_FLAG_STREAMIN (1 << 14)
+#define NEUREKA_NORM_MODE_8BIT (0 << 12)
+#define NEUREKA_NORM_MODE_32BIT (2 << 12)
+#define NEUREKA_FLAG_ACTIVATION_PREFETCH_ON (1 << 10)
+#define NEUREKA_FLAG_ACTIVATION_PREFETCH_OFF (0 << 10)
+#define NEUREKA_FLAG_WEIGHT_SOURCE_WMEM (1 << 9)
+#define NEUREKA_FLAG_WEIGHT_SOURCE_TCDM (0 << 9)
+#define NEUREKA_FLAG_LINEAR_MODE (1 << 7) // not tested
+#define NEUREKA_FLAG_MODE_3x3 (0 << 5)
+#define NEUREKA_FLAG_MODE_3x3_DW (1 << 5)
+#define NEUREKA_FLAG_MODE_1x1 (2 << 5)
+#define NEUREKA_FLAG_NORM_QUANT (1 << 4)
+
+/* Masks */
+
+#define NEUREKA_MASK_FLAG_INPUT_SIGNED (0x1 << 26)
+#define NEUREKA_MASK_FLAG_NORM_BIAS (0x1 << 25)
+#define NEUREKA_MASK_FLAG_NORM_SHIFT (0x1 << 24)
+#define NEUREKA_MASK_QUANT_FUNCTION (0x1 << 23)
+#define NEUREKA_MASK_QUANT_MODE (0x3 << 21)
+#define NEUREKA_MASK_SHIFT_AMOUNT (0x1f << 16)
+#define NEUREKA_MASK_WEIGHT_OFFSET_MODE (0x1 << 15)
+#define NEUREKA_MASK_NORM_MODE (0x3 << 12)
+#define NEUREKA_MASK_FLAG_ACTIVATION_PREFETCH (0x1 << 10)
+#define NEUREKA_MASK_FLAG_WEIGHT_SOURCE (0x1 << 9)
+#define NEUREKA_MASK_FLAG_MODE (0x3 << 5)
+#define NEUREKA_MASK_FLAG_WEIGHT_BITS (0x7 << 0)
+
+/* PADDING */
+
+#define NEUREKA_DONT_PAD (0)
+#define NEUREKA_MAX_PAD (2)
+
+/* NORM */
+#define NEUREKA_NORM_MAX_LEN (32)
+
+#endif // __NEUREKA_DEFS_H__
diff --git a/neureka/app/inc/layer_util.h b/neureka/app/inc/layer_util.h
new file mode 100644
index 0000000..e44ede9
--- /dev/null
+++ b/neureka/app/inc/layer_util.h
@@ -0,0 +1,40 @@
+/*
+ * Luka Macan <luka.macan@unibo.it>
+ *
+ * Copyright 2023 ETH Zurich and University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ *
+ * SPDX-License-Identifier: Apache-2.0
+ */
+
+#ifndef __LAYER_UTIL_H__
+#define __LAYER_UTIL_H__
+
+#include "layer_conf.h"
+#include <pmsis.h>
+
+static void layer_info() {
+  printf("Layer info:\n"
+         " - input: (%dx%dx%d)\n"
+         " - output: (%dx%dx%d)\n"
+         " - weight: (%dx%dx%dx%d)\n"
+         " - stride: (%dx%d)\n"
+         " - padding: (%dx%dx%dx%d)\n",
+         INPUT_HEIGHT, INPUT_WIDTH, INPUT_CHANNEL, OUTPUT_HEIGHT, OUTPUT_WIDTH,
+         OUTPUT_CHANNEL, WEIGHT_CHANNEL_OUT, WEIGHT_HEIGHT, WEIGHT_WIDTH,
+         WEIGHT_CHANNEL_IN, STRIDE_HEIGHT, STRIDE_WIDTH, PADDING_TOP,
+         PADDING_BOTTOM, PADDING_LEFT, PADDING_RIGHT);
+}
+
+#endif // __LAYER_UTIL_H__
diff --git a/neureka/app/inc/nnx_layer.h b/neureka/app/inc/nnx_layer.h
new file mode 100644
index 0000000..cbaf4b9
--- /dev/null
+++ b/neureka/app/inc/nnx_layer.h
@@ -0,0 +1,26 @@
+/*
+ * Luka Macan <luka.macan@unibo.it>
+ *
+ * Copyright 2023 ETH Zurich and University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ *
+ * SPDX-License-Identifier: Apache-2.0
+ */
+
+#ifndef __NNX_LAYER_H__
+#define __NNX_LAYER_H__
+
+void execute_nnx_layer(void *unused_args);
+
+#endif // __NNX_LAYER_H__
diff --git a/neureka/app/inc/pmsis.h b/neureka/app/inc/pmsis.h
new file mode 100644
index 0000000..1f16f1b
--- /dev/null
+++ b/neureka/app/inc/pmsis.h
@@ -0,0 +1,5 @@
+// fake pmsis.h
+#include <stdint.h>
+#define PI_L1 __attribute__((section(".data_l1")))
+#define PI_L2 __attribute__((section(".data_l1")))
+// #include "tinyprintf.h"
\ No newline at end of file
diff --git a/neureka/app/inc/tinyprintf.h b/neureka/app/inc/tinyprintf.h
new file mode 100644
index 0000000..0af0609
--- /dev/null
+++ b/neureka/app/inc/tinyprintf.h
@@ -0,0 +1,248 @@
+/*
+
+Copyright (c) 2004,2012 Kustaa Nyholm / SpareTimeLabs
+
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are met:
+    * Redistributions of source code must retain the above copyright
+      notice, this list of conditions and the following disclaimer.
+    * Redistributions in binary form must reproduce the above copyright
+      notice, this list of conditions and the following disclaimer in the
+      documentation and/or other materials provided with the distribution.
+    * Neither the name of Kustaa Nyholm or SpareTimeLabs nor the
+      names of its contributors may be used to endorse or promote products
+      derived from this software without specific prior written permission.
+
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
+ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
+WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
+DISCLAIMED. IN NO EVENT SHALL COPYRIGHT HOLDER BE LIABLE FOR ANY
+DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
+(INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
+LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
+ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
+SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+This modified tinyprintf is distributed under the conditions of the BSD-new license
+as explicitly allowed:
+ -> https://github.com/cjlano/tinyprintf commit 2ee3012
+
+They provide a simple and small (+400 loc) printf functionality to
+be used in embedded systems.
+
+I've found them so useful in debugging that I do not bother with a
+debugger at all.
+
+They are distributed in source form, so to use them, just compile them
+into your project.
+
+Two printf variants are provided: printf and the 'sprintf' family of
+functions ('snprintf', 'sprintf', 'vsnprintf', 'vsprintf').
+
+The formats supported by this implementation are:
+'c' 'd' 'i' 'o' 'p' 'u' 's' 'x' 'X'.
+
+Zero padding and field width are also supported.
+
+If the library is compiled with 'PRINTF_SUPPORT_LONG' defined, then
+the long specifier is also supported. Note that this will pull in some
+long math routines (pun intended!) and thus make your executable
+noticeably longer. Likewise with 'PRINTF_LONG_LONG_SUPPORT' for the
+long long specifier, and with 'PRINTF_SIZE_T_SUPPORT' for the size_t
+specifier.
+
+The memory footprint of course depends on the target CPU, compiler and
+compiler options, but a rough guesstimate (based on a H8S target) is about
+1.4 kB for code and some twenty 'int's and 'char's, say 60 bytes of stack space.
+Not too bad. Your mileage may vary. By hacking the source code you can
+get rid of some hundred bytes, I'm sure, but personally I feel the balance of
+functionality and flexibility versus  code size is close to optimal for
+many embedded systems.
+
+To use the printf, you need to supply your own character output function,
+something like :
+
+void putc ( void* p, char c)
+{
+    while (!SERIAL_PORT_EMPTY) ;
+    SERIAL_PORT_TX_REGISTER = c;
+}
+
+Before you can call printf, you need to initialize it to use your
+character output function with something like:
+
+init_printf(NULL,putc);
+
+Notice the 'NULL' in 'init_printf' and the parameter 'void* p' in 'putc',
+the NULL (or any pointer) you pass into the 'init_printf' will eventually be
+passed to your 'putc' routine. This allows you to pass some storage space (or
+anything really) to the character output function, if necessary.
+This is not often needed but it was implemented like that because it made
+implementing the sprintf function so neat (look at the source code).
+
+The code is re-entrant, except for the 'init_printf' function, so it is safe
+to call it from interrupts too, although this may result in mixed output.
+If you rely on re-entrancy, take care that your 'putc' function is re-entrant!
+
+The printf and sprintf functions are actually macros that translate to
+'tfp_printf' and 'tfp_sprintf' when 'TINYPRINTF_OVERRIDE_LIBC' is set
+(default). Setting it to 0 makes it possible to use them along with
+'stdio.h' printf's in a single source file. When
+'TINYPRINTF_OVERRIDE_LIBC' is set, please note that printf/sprintf are
+not function-like macros, so if you have variables or struct members
+with these names, things will explode in your face.  Without variadic
+macros this is the best we can do to wrap these function. If it is a
+problem, just give up the macros and use the functions directly, or
+rename them.
+
+It is also possible to avoid defining tfp_printf and/or tfp_sprintf by
+clearing 'TINYPRINTF_DEFINE_TFP_PRINTF' and/or
+'TINYPRINTF_DEFINE_TFP_SPRINTF' to 0. This allows for example to
+export only tfp_format, which is at the core of all the other
+functions.
+
+For further details see source code.
+
+regs Kusti, 23.10.2004
+*/
+
+#ifndef __TFP_PRINTF__
+#define __TFP_PRINTF__
+
+#include <stdarg.h>
+
+void putf(char *null, char c);
+
+/* Global configuration */
+
+/* Set this to 0 if you do not want to provide tfp_printf */
+#ifndef TINYPRINTF_DEFINE_TFP_PRINTF
+# define TINYPRINTF_DEFINE_TFP_PRINTF 1
+#endif
+
+/* Set this to 0 if you do not want to provide
+   tfp_sprintf/snprintf/vsprintf/vsnprintf */
+#ifndef TINYPRINTF_DEFINE_TFP_SPRINTF
+# define TINYPRINTF_DEFINE_TFP_SPRINTF 1
+#endif
+
+/* Set this to 0 if you do not want tfp_printf and
+   tfp_{vsn,sn,vs,s}printf to be also available as
+   printf/{vsn,sn,vs,s}printf */
+#ifndef TINYPRINTF_OVERRIDE_LIBC
+# define TINYPRINTF_OVERRIDE_LIBC 1
+#endif
+
+/* Optional external types dependencies */
+
+#if TINYPRINTF_DEFINE_TFP_SPRINTF
+# include <sys/types.h>  /* size_t */
+#endif
+
+/* Declarations */
+
+#ifdef __GNUC__
+# define _TFP_SPECIFY_PRINTF_FMT(fmt_idx,arg1_idx) \
+    __attribute__((format (printf, fmt_idx, arg1_idx)))
+#else
+# define _TFP_SPECIFY_PRINTF_FMT(fmt_idx,arg1_idx)
+#endif
+
+#ifdef  __cplusplus
+extern "C" {
+#endif
+
+typedef void (*putcf) (void *, char);
+
+/*
+   'tfp_format' really is the central function for all tinyprintf. For
+   each output character after formatting, the 'putf' callback is
+   called with 2 args:
+     - an arbitrary void* 'putp' param defined by the user and
+       passed unmodified from 'tfp_format',
+     - the character.
+   The 'tfp_printf' and 'tfp_sprintf' functions simply define their own
+   callback and pass to it the right 'putp' it is expecting.
+*/
+void tfp_format(void *putp, putcf putf, const char *fmt, va_list va);
+
+#if TINYPRINTF_DEFINE_TFP_SPRINTF
+int tfp_vsnprintf(char *str, size_t size, const char *fmt, va_list ap);
+int tfp_snprintf(char *str, size_t size, const char *fmt, ...) \
+     _TFP_SPECIFY_PRINTF_FMT(3, 4);
+int tfp_vsprintf(char *str, const char *fmt, va_list ap);
+int tfp_sprintf(char *str, const char *fmt, ...) \
+    _TFP_SPECIFY_PRINTF_FMT(2, 3);
+# if TINYPRINTF_OVERRIDE_LIBC
+#  define vsnprintf tfp_vsnprintf
+#  define snprintf tfp_snprintf
+#  define vsprintf tfp_vsprintf
+#  define sprintf tfp_sprintf
+# endif
+#endif
+
+#if TINYPRINTF_DEFINE_TFP_PRINTF
+void init_printf(void *putp, putcf putf);
+void tfp_printf(char *fmt, ...) _TFP_SPECIFY_PRINTF_FMT(1, 2);
+# if TINYPRINTF_OVERRIDE_LIBC
+#  ifndef DISABLE_PRINTF
+#   define printf tfp_printf
+#  else
+#   define printf(...)
+#  endif
+# endif
+#endif
+
+#ifdef  __cplusplus
+}
+#endif
+
+#endif
+
+
+
+/*
+ * Configuration
+ */
+
+/* Enable long int support */
+// #define PRINTF_LONG_SUPPORT
+
+/* Enable long long int support (implies long int support) */
+// #define PRINTF_LONG_LONG_SUPPORT
+
+/* Enable %z (size_t) support */
+// #define PRINTF_SIZE_T_SUPPORT
+
+/*
+ * Configuration adjustments
+ */
+#ifdef PRINTF_SIZE_T_SUPPORT
+#include <sys/types.h>
+#endif
+
+#ifdef PRINTF_LONG_LONG_SUPPORT
+# define PRINTF_LONG_SUPPORT
+#endif
+
+/* __SIZEOF_<type>__ defined at least by gcc */
+#ifdef __SIZEOF_POINTER__
+# define SIZEOF_POINTER __SIZEOF_POINTER__
+#endif
+#ifdef __SIZEOF_LONG_LONG__
+# define SIZEOF_LONG_LONG __SIZEOF_LONG_LONG__
+#endif
+#ifdef __SIZEOF_LONG__
+# define SIZEOF_LONG __SIZEOF_LONG__
+#endif
+#ifdef __SIZEOF_INT__
+# define SIZEOF_INT __SIZEOF_INT__
+#endif
+
+#ifdef __GNUC__
+# define _TFP_GCC_NO_INLINE_  __attribute__ ((noinline))
+#else
+# define _TFP_GCC_NO_INLINE_
+#endif
+
diff --git a/neureka/app/pulp_nnx_neureka.c b/neureka/app/pulp_nnx_neureka.c
new file mode 100644
index 0000000..b814f23
--- /dev/null
+++ b/neureka/app/pulp_nnx_neureka.c
@@ -0,0 +1,76 @@
+/*
+ * Luka Macan <luka.macan@unibo.it>
+ *
+ * Copyright 2023 ETH Zurich and University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ *
+ * SPDX-License-Identifier: Apache-2.0
+ */
+
+#include "pulp_nnx_neureka.h"
+#include "hwpe.h"
+#include "neureka.h"
+#include "pulp_nnx_util.h"
+#include <pmsis.h>
+#include <stdint.h>
+#include <sys/types.h>
+
+void neureka_nnx_init(neureka_dev_t *dev, neureka_bsp_conf_t *conf) {
+  neureka_bsp_open(conf);
+  hwpe_soft_clear(&dev->hwpe_dev);
+}
+
+void neureka_nnx_term(neureka_dev_t *dev) {
+  hwpe_soft_clear(&dev->hwpe_dev);
+  neureka_bsp_close();
+}
+
+int neureka_nnx_dispatch_check(neureka_dev_t *dev) {
+  return !neureka_task_queue_full(dev);
+}
+
+void neureka_nnx_dispatch_wait(neureka_dev_t *dev) {
+  while (!neureka_nnx_dispatch_check(dev)) {
+    neureka_bsp_event_wait_and_clear();
+  }
+}
+
+int neureka_nnx_dispatch(neureka_dev_t *dev, neureka_task_t *task) {
+  if (hwpe_task_queue_acquire_task(&dev->hwpe_dev, &task->id)) {
+    return 1;
+  }
+  hwpe_task_queue_write_task(&dev->hwpe_dev, (uint32_t *)&task->data,
+                             (int)(sizeof(neureka_task_data_t) / 4));
+  hwpe_task_queue_release_and_run(&dev->hwpe_dev);
+  return 0;
+}
+
+int neureka_nnx_resolve_check(neureka_dev_t *dev, neureka_task_t *task) {
+#if __PLATFORM__ == ARCHI_PLATFORM_GVSOC
+  // GVSOC model has a broken running_id so resolve_check
+  // conservativly looks if the task queue is empty.
+  return neureka_task_queue_empty(dev);
+#else
+  uint8_t prev_task_id = task->id - 1;
+  return !(hwpe_last_task_id(&dev->hwpe_dev) == prev_task_id ||
+           (hwpe_last_task_id(&dev->hwpe_dev) == task->id &&
+            !neureka_task_queue_empty(dev)));
+#endif
+}
+
+void neureka_nnx_resolve_wait(neureka_dev_t *dev, neureka_task_t *task) {
+  while (!neureka_nnx_resolve_check(dev, task)) {
+    neureka_bsp_event_wait_and_clear();
+  }
+}
diff --git a/neureka/app/pulp_nnx_neureka.h b/neureka/app/pulp_nnx_neureka.h
new file mode 100644
index 0000000..80096c8
--- /dev/null
+++ b/neureka/app/pulp_nnx_neureka.h
@@ -0,0 +1,61 @@
+/*
+ * Luka Macan <luka.macan@unibo.it>
+ *
+ * Copyright 2023 ETH Zurich and University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ *
+ * SPDX-License-Identifier: Apache-2.0
+ */
+
+#include "neureka.h"
+#include "neureka_bsp.h"
+#include "neureka_task.h"
+#include <stdint.h>
+
+/* PULP-NNX interface */
+
+void neureka_nnx_init(neureka_dev_t *dev, neureka_bsp_conf_t *conf);
+void neureka_nnx_term(neureka_dev_t *dev);
+
+/** neureka_nnx_dispatch_check
+ *
+ * Check whether you can dispatch to the accelerator.
+ */
+int neureka_nnx_dispatch_check(neureka_dev_t *dev);
+
+/** neureka_nnx_dispatch_wait
+ *
+ * Block until you can dispatch to the accelerator.
+ */
+void neureka_nnx_dispatch_wait(neureka_dev_t *dev);
+
+/** neureka_nnx_dispatch
+ *
+ * Dispatch a task to the accelerator.
+ * Fails with return code 1 if the task cannot be dispatched. Otherwise returns
+ * 0.
+ */
+int neureka_nnx_dispatch(neureka_dev_t *dev, neureka_task_t *task);
+
+/** neureka_nnx_resolve_check
+ *
+ * Check whether the task has been resolved.
+ */
+int neureka_nnx_resolve_check(neureka_dev_t *dev, neureka_task_t *task);
+
+/** neureka_nnx_resolve_wait
+ *
+ * Block until you can resolve the task.
+ */
+void neureka_nnx_resolve_wait(neureka_dev_t *dev, neureka_task_t *task);
diff --git a/neureka/app/src/main.c b/neureka/app/src/main.c
new file mode 100644
index 0000000..818ddf7
--- /dev/null
+++ b/neureka/app/src/main.c
@@ -0,0 +1,56 @@
+/*
+ * Copyright (C) 2020-2024 ETH Zurich and University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+/*
+ * Authors:  Francesco Conti <fconti@iis.ee.ethz.ch>
+ *           Gianna Paulin <pauling@iis.ee.ethz.ch>
+ *           Renzo Andri <andrire@iis.ee.ethz.ch>
+ *           Arpan Suravi Prasad <prasadar@iis.ee.ethz.ch>
+ *           Luka Macan <luka.macan@unibo.it>
+ * Main Test Program for N-EUREKA
+ */
+
+#include <stdint.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <pulp.h>
+
+#include "layer_util.h"
+#include "nnx_layer.h"
+#include "output.h"
+#include "input.h"
+#include "weight.h"
+
+int main() {
+  if (rt_cluster_id() != 0)
+    return bench_cluster_forward(0);
+
+  int err = 0;
+  if(rt_core_id()==0)
+  {
+    // execute NNX layer
+    execute_nnx_layer(NULL);
+
+    printf("Checking outputs");
+
+    // output checking
+    err = check_output();
+    
+  }
+  synch_barrier();
+  return err;
+
+}
diff --git a/neureka/app/src/nnx_layer.c b/neureka/app/src/nnx_layer.c
new file mode 100644
index 0000000..f38e830
--- /dev/null
+++ b/neureka/app/src/nnx_layer.c
@@ -0,0 +1,168 @@
+/*
+ * Luka Macan <luka.macan@unibo.it>
+ *
+ * Copyright 2023 ETH Zurich and University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ *
+ * SPDX-License-Identifier: Apache-2.0
+ */
+
+#include "nnx_layer.h"
+#include <pmsis.h>
+
+#include "neureka.h"
+#include "neureka_gvsoc.h"
+#include "neureka_testbench_bsp.h"
+#include "neureka_task.h"
+#include "pulp_nnx_neureka.h"
+
+#define NULL 0
+
+typedef neureka_norm_mode_e nnx_norm_mode_e;
+typedef neureka_quant_t nnx_quant_t;
+typedef neureka_quant_function_e nnx_quant_function_e;
+typedef neureka_norm_t nnx_norm_t;
+typedef neureka_task_t nnx_task_t;
+typedef neureka_dev_t nnx_dev_t;
+typedef neureka_testbench_conf_t nnx_bsp_conf_t;
+typedef neureka_task_flag_e nnx_task_flag_e;
+
+#define nnxTaskFlagTrue neurekaTaskFlagTrue
+#define nnxTaskFlagFalse neurekaTaskFlagFalse
+
+#define nnx_task_init neureka_task_init
+#define nnx_task_set_op_to_conv neureka_task_set_op_to_conv
+#define nnx_task_set_bits neureka_task_set_bits
+#define nnx_task_set_norm_quant neureka_task_set_norm_quant
+#define nnx_task_set_weight_offset neureka_task_set_weight_offset
+#define nnx_task_set_weight_source neureka_task_set_weight_source
+#define nnx_task_set_activation_prefetch neureka_task_set_activation_prefetch
+#define nnx_task_set_dims neureka_task_set_dims
+#define nnx_task_set_ptrs_conv neureka_task_set_ptrs_conv
+#define nnx_task_set_ptrs_norm_quant neureka_task_set_ptrs_norm_quant
+
+#define nnx_bsp_get_dev neureka_testbench_get_dev
+
+#define nnx_init neureka_nnx_init
+#define nnx_dispatch_wait neureka_nnx_dispatch_wait
+#define nnx_dispatch neureka_nnx_dispatch
+#define nnx_resolve_wait neureka_nnx_resolve_wait
+#define nnx_term neureka_nnx_term
+
+// Generated headers
+#include "bias.h"
+#include "input.h"
+#include "layer_conf.h"
+#include "output.h"
+#include "scale.h"
+#include "weight.h"
+
+static void task_prepare(nnx_task_t *task) {
+  nnx_task_init(task);
+  nnx_task_set_op_to_conv(task, WEIGHT_HEIGHT, GROUPS > 1);
+  nnx_task_set_bits(task, INPUT_BITS, OUTPUT_BITS, WEIGHT_BITS);
+
+  nnx_task_set_weight_offset(task, weightOffsetModeLayerWise, WEIGHT_OFFSET);
+
+#ifdef NEUREKA_WEIGHT_SOURCE_WMEM
+  nnx_task_set_weight_source(task, neurekaWeightSourceWmem);
+  nnx_task_set_activation_prefetch(task, activationPrefetchOn);
+#else
+  neureka_task_set_weight_source(task, neurekaWeightSourceTcdm);
+  nnx_task_set_activation_prefetch(task, activationPrefetchOff);
+#endif
+#if INPUT_SIGNED == 1
+  neureka_task_set_input_signed(task);
+#else
+  neureka_task_set_input_unsigned(task);
+#endif
+
+  const uint32_t w_in_stride = INPUT_CHANNEL * INPUT_BITS / 8;
+  const uint32_t h_in_stride = INPUT_WIDTH * w_in_stride;
+  const uint32_t w_out_stride = OUTPUT_CHANNEL * OUTPUT_BITS / 8;
+  const uint32_t h_out_stride = OUTPUT_WIDTH * w_out_stride;
+
+#if STRIDE_HEIGHT == 2 && STRIDE_WIDTH == 2
+  nnx_task_set_dims_stride2x2(
+      task, INPUT_HEIGHT, INPUT_WIDTH, INPUT_CHANNEL, h_in_stride, w_in_stride,
+      OUTPUT_HEIGHT, OUTPUT_WIDTH, OUTPUT_CHANNEL, h_out_stride, w_out_stride,
+      WEIGHT_HEIGHT, WEIGHT_WIDTH, PADDING_TOP, PADDING_BOTTOM, PADDING_LEFT,
+      PADDING_RIGHT);
+#else
+  nnx_task_set_dims(task, INPUT_WIDTH, INPUT_CHANNEL, h_in_stride, w_in_stride,
+                    OUTPUT_HEIGHT, OUTPUT_WIDTH, OUTPUT_CHANNEL, h_out_stride,
+                    w_out_stride, PADDING_TOP, PADDING_BOTTOM, PADDING_LEFT,
+                    PADDING_RIGHT);
+#endif
+
+  nnx_task_set_ptrs_conv(task, (uint32_t)input, INPUT_WIDTH, w_in_stride,
+                         PADDING_TOP, PADDING_LEFT, (uint32_t)output,
+                         (uint32_t)weight);
+#if HAS_NORM_QUANT == 1
+#if SCALE_BITS == 8
+  const nnx_norm_mode_e normMode = normMode8Bit;
+#elif SCALE_BITS == 32
+  const nnx_norm_mode_e normMode = normMode32Bit;
+#endif
+
+  const nnx_task_flag_e flag_bias =
+      HAS_BIAS ? nnxTaskFlagTrue : nnxTaskFlagFalse;
+  const uint32_t bias_ptr = (uint32_t)(HAS_BIAS ? bias : NULL);
+
+  nnx_quant_function_e quant_function =
+      HAS_RELU ? quantFunctionRelu : quantFunctionIdentity;
+
+  nnx_task_set_norm_quant(task,
+                          (nnx_quant_t){.shift_amount = OUTSHIFT,
+                                        .function = quant_function,
+                                        .flag_rounding = nnxTaskFlagFalse},
+                          (nnx_norm_t){.mode = normMode,
+                                       .flag_bias = flag_bias,
+                                       .flag_shift = nnxTaskFlagFalse});
+
+  nnx_task_set_ptrs_norm_quant(task, (uint32_t)scale, NULL, bias_ptr);
+#endif // HAS_NORM_QUANT
+}
+
+static void task_execute(nnx_task_t *task) {
+  nnx_dev_t *dev = nnx_bsp_get_dev();
+
+  nnx_bsp_conf_t conf = {.max_stall = 8};
+  nnx_init(dev, &conf);
+
+  nnx_dispatch_wait(dev);
+
+  // printf("CFG:\n");
+  // for (int i=0; i<sizeof(neureka_task_data_t)/4; i++) {
+  //   printf("%08x\n", ((uint32_t *) &task->data)[i]);
+  // }
+#if STRIDE_HEIGHT == 2 && STRIDE_WIDTH == 2
+  nnx_dispatch_stride2x2(dev, task, INPUT_WIDTH, INPUT_CHANNEL, OUTPUT_HEIGHT,
+                         OUTPUT_WIDTH, OUTPUT_CHANNEL, WEIGHT_HEIGHT,
+                         WEIGHT_WIDTH);
+#else
+  nnx_dispatch(dev, task);
+#endif
+
+  nnx_resolve_wait(dev, task);
+
+  nnx_term(dev);
+
+}
+
+void execute_nnx_layer(void *args) {
+  nnx_task_t task;
+  task_prepare(&task);
+  task_execute(&task);
+}
diff --git a/neureka/app/util/hwpe.c b/neureka/app/util/hwpe.c
new file mode 100644
index 0000000..0430081
--- /dev/null
+++ b/neureka/app/util/hwpe.c
@@ -0,0 +1,85 @@
+/*
+ * Luka Macan <luka.macan@unibo.it>
+ *
+ * Copyright 2023 ETH Zurich and University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ *
+ * SPDX-License-Identifier: Apache-2.0
+ */
+
+#include "hwpe.h"
+#include <stdint.h>
+
+#define HWPE_TRIGGER 0
+#define HWPE_ACQUIRE 1
+#define HWPE_FINISHED 2
+#define HWPE_STATUS 3
+#define HWPE_RUNNING_JOB 4
+#define HWPE_SOFT_CLEAR 5
+#define HWPE_SWSYNC 6
+#define HWPE_TASK_REG_OFFSET 8
+
+inline void hwpe_reg_write(hwpe_dev_t *dev, int reg, uint32_t value) {
+  dev->base_addr[reg] = value;
+}
+
+inline uint32_t hwpe_reg_read(hwpe_dev_t *dev, int reg) {
+  return dev->base_addr[reg];
+}
+
+inline void hwpe_task_reg_write(hwpe_dev_t *dev, int reg, uint32_t value) {
+  hwpe_reg_write(dev, HWPE_TASK_REG_OFFSET + reg, value);
+}
+
+inline uint32_t hwpe_task_reg_read(hwpe_dev_t *dev, int reg) {
+  return hwpe_reg_read(dev, HWPE_TASK_REG_OFFSET + reg);
+}
+
+void hwpe_soft_clear(hwpe_dev_t *dev) {
+  hwpe_reg_write(dev, HWPE_SOFT_CLEAR, 0);
+  for (volatile int i = 0; i < 10; i++)
+    ;
+}
+
+uint32_t hwpe_task_queue_status(hwpe_dev_t *dev) {
+  return hwpe_reg_read(dev, HWPE_STATUS);
+}
+
+int hwpe_task_queue_acquire_task(hwpe_dev_t *dev, uint8_t *id) {
+  uint32_t read_value = (int32_t)hwpe_reg_read(dev, HWPE_ACQUIRE);
+  if (read_value >= 256) {
+    return 1;
+  } else {
+    *id = (uint8_t)read_value;
+    return 0;
+  }
+}
+
+void hwpe_task_queue_write_task(hwpe_dev_t *dev, uint32_t *data, int len) {
+  for (int i = 0; i < len; i++) {
+    hwpe_task_reg_write(dev, i, data[i]);
+  }
+}
+
+void hwpe_task_queue_release_and_run(hwpe_dev_t *dev) {
+  hwpe_reg_write(dev, HWPE_TRIGGER, 0);
+}
+
+void hwpe_task_queue_release(hwpe_dev_t *dev) {
+  hwpe_reg_write(dev, HWPE_TRIGGER, 1);
+}
+
+uint8_t hwpe_last_task_id(hwpe_dev_t *dev) {
+  return (uint8_t)hwpe_reg_read(dev, HWPE_RUNNING_JOB);
+}
diff --git a/neureka/app/util/hwpe.h b/neureka/app/util/hwpe.h
new file mode 100644
index 0000000..52bf912
--- /dev/null
+++ b/neureka/app/util/hwpe.h
@@ -0,0 +1,43 @@
+/*
+ * Luka Macan <luka.macan@unibo.it>
+ *
+ * Copyright 2023 ETH Zurich and University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ *
+ * SPDX-License-Identifier: Apache-2.0
+ */
+
+#ifndef __HWPE_H__
+#define __HWPE_H__
+
+#include <stdint.h>
+
+/* HWPE device */
+typedef struct hwpe_dev_t {
+  volatile uint32_t *base_addr;
+} hwpe_dev_t;
+
+void hwpe_reg_write(hwpe_dev_t *dev, int reg, uint32_t value);
+uint32_t hwpe_reg_read(hwpe_dev_t *dev, int reg);
+void hwpe_task_reg_write(hwpe_dev_t *dev, int reg, uint32_t value);
+uint32_t hwpe_task_reg_read(hwpe_dev_t *dev, int reg);
+void hwpe_soft_clear(hwpe_dev_t *dev);
+uint32_t hwpe_task_queue_status(hwpe_dev_t *dev);
+int hwpe_task_queue_acquire_task(hwpe_dev_t *dev, uint8_t *id);
+void hwpe_task_queue_write_task(hwpe_dev_t *dev, uint32_t *data, int len);
+void hwpe_task_queue_release_and_run(hwpe_dev_t *dev);
+void hwpe_task_queue_release(hwpe_dev_t *dev);
+uint8_t hwpe_last_task_id(hwpe_dev_t *dev);
+
+#endif // !__HWPE_H__
diff --git a/neureka/app/util/pulp_nnx_util.c b/neureka/app/util/pulp_nnx_util.c
new file mode 100644
index 0000000..0107fc1
--- /dev/null
+++ b/neureka/app/util/pulp_nnx_util.c
@@ -0,0 +1,35 @@
+/*
+ * Luka Macan <luka.macan@unibo.it>
+ *
+ * Copyright 2023 ETH Zurich and University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ *
+ * SPDX-License-Identifier: Apache-2.0
+ */
+
+#include "pulp_nnx_util.h"
+
+inline int nnx_calculate_number_of_tiles(const int dim_size,
+                                         const int tile_size) {
+  return ((dim_size - 1) / tile_size) + 1;
+}
+
+inline int nnx_calculate_last_tile_size(const int dim_size,
+                                        const int tile_size) {
+  return ((dim_size - 1) % tile_size) + 1;
+}
+
+inline uint32_t nnx_concat_half(const uint16_t high, const uint16_t low) {
+  return ((uint32_t)high << 16) | low;
+}
diff --git a/neureka/app/util/pulp_nnx_util.h b/neureka/app/util/pulp_nnx_util.h
new file mode 100644
index 0000000..d167f6d
--- /dev/null
+++ b/neureka/app/util/pulp_nnx_util.h
@@ -0,0 +1,51 @@
+/*
+ * Luka Macan <luka.macan@unibo.it>
+ *
+ * Copyright 2023 ETH Zurich and University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ *
+ * SPDX-License-Identifier: Apache-2.0
+ */
+
+#ifndef __NNX_UTIL_H__
+#define __NNX_UTIL_H__
+
+#include <stdint.h>
+
+/**
+ * nnx_calculate_number_of_iterations
+ *
+ * Calculates the number of iterations to go through a dimension.
+ * It does it by dividing the dimension with the tile size and doing a ceiling
+ * the result.
+ */
+int nnx_calculate_number_of_tiles(const int dim_size, const int tile_size);
+
+/**
+ * nnx_calculate_last_tile_size
+ *
+ * Calculates the size of the last executed tile by calculating the remainder of
+ * the dim_size and the tile_size. In case the remainder is 0, it returns the
+ * full tile_size.
+ */
+int nnx_calculate_last_tile_size(const int dim_size, const int tile_size);
+
+/**
+ * concat_half
+ *
+ * Concatenate 2 16-bit numbers into a 32-bit number.
+ */
+uint32_t nnx_concat_half(const uint16_t high, const uint16_t low);
+
+#endif // __NNX_UTIL_H__
diff --git a/neureka/dense/Makefile b/neureka/dense/Makefile
new file mode 100644
index 0000000..edb77cb
--- /dev/null
+++ b/neureka/dense/Makefile
@@ -0,0 +1,3 @@
+include ../app/Makefile
+
+STIM_DIR := ../dense/
\ No newline at end of file
diff --git a/neureka/dense/inc/bias.h b/neureka/dense/inc/bias.h
new file mode 100644
index 0000000..0cc75d8
--- /dev/null
+++ b/neureka/dense/inc/bias.h
@@ -0,0 +1,9 @@
+#ifndef __BIAS_H__
+#define __BIAS_H__
+
+#include <pmsis.h>
+
+#define BIAS_SIZE (35)
+extern int32_t bias[BIAS_SIZE];
+
+#endif  // __BIAS_H__
diff --git a/neureka/dense/inc/input.h b/neureka/dense/inc/input.h
new file mode 100644
index 0000000..ddb3845
--- /dev/null
+++ b/neureka/dense/inc/input.h
@@ -0,0 +1,9 @@
+#ifndef __INPUT_H__
+#define __INPUT_H__
+
+#include <pmsis.h>
+
+#define INPUT_SIZE (1407)
+extern uint8_t input[INPUT_SIZE];
+
+#endif  // __INPUT_H__
diff --git a/neureka/dense/inc/layer_conf.h b/neureka/dense/inc/layer_conf.h
new file mode 100644
index 0000000..d005132
--- /dev/null
+++ b/neureka/dense/inc/layer_conf.h
@@ -0,0 +1,42 @@
+#ifndef __LAYER_CONF_H__
+#define __LAYER_CONF_H__
+
+#define TEST_NAME "test"
+#define INPUT_HEIGHT (7)
+#define INPUT_WIDTH (3)
+#define INPUT_CHANNEL (67)
+#define INPUT_SIGNED (0)
+#define INPUT_BITS (8)
+
+#define OUTPUT_HEIGHT (5)
+#define OUTPUT_WIDTH (1)
+#define OUTPUT_CHANNEL (35)
+#define OUTPUT_BITS (8)
+
+#define WEIGHT_HEIGHT (3)
+#define WEIGHT_WIDTH (3)
+#define WEIGHT_CHANNEL_IN (67)
+#define WEIGHT_CHANNEL_OUT (35)
+#define WEIGHT_BITS (8)
+#define WEIGHT_OFFSET (-128)
+
+#define SCALE_BITS (8)
+
+#define BIAS_BITS (32)
+
+#define PADDING_TOP (0)
+#define PADDING_BOTTOM (0)
+#define PADDING_LEFT (0)
+#define PADDING_RIGHT (0)
+#define PADDING_VALUE (0)
+
+#define STRIDE_HEIGHT (1)
+#define STRIDE_WIDTH (1)
+
+#define GROUPS (1)
+#define OUTSHIFT (12)
+#define HAS_NORM_QUANT (1)
+#define HAS_BIAS (1)
+#define HAS_RELU (1)
+
+#endif  // __LAYER_CONF_H__
diff --git a/neureka/dense/inc/output.h b/neureka/dense/inc/output.h
new file mode 100644
index 0000000..cb7abf6
--- /dev/null
+++ b/neureka/dense/inc/output.h
@@ -0,0 +1,14 @@
+#ifndef __OUTPUT_H__
+#define __OUTPUT_H__
+
+#include <pmsis.h>
+
+#define OUTPUT_SIZE (175)
+extern uint8_t output[OUTPUT_SIZE];
+
+#define GOLDEN_OUTPUT_SIZE (175)
+extern uint8_t golden_output[GOLDEN_OUTPUT_SIZE];
+
+int check_output();
+
+#endif  // __OUTPUT_H__
diff --git a/neureka/dense/inc/scale.h b/neureka/dense/inc/scale.h
new file mode 100644
index 0000000..3c546a5
--- /dev/null
+++ b/neureka/dense/inc/scale.h
@@ -0,0 +1,9 @@
+#ifndef __SCALE_H__
+#define __SCALE_H__
+
+#include <pmsis.h>
+
+#define SCALE_SIZE (35)
+extern uint8_t scale[SCALE_SIZE];
+
+#endif  // __SCALE_H__
diff --git a/neureka/dense/inc/weight.h b/neureka/dense/inc/weight.h
new file mode 100644
index 0000000..635a7b4
--- /dev/null
+++ b/neureka/dense/inc/weight.h
@@ -0,0 +1,9 @@
+#ifndef __WEIGHT_H__
+#define __WEIGHT_H__
+
+#include <pmsis.h>
+
+#define WEIGHT_SIZE (30240)
+extern uint8_t weight[WEIGHT_SIZE];
+
+#endif  // __WEIGHT_H__
diff --git a/neureka/dense/src/bias.c b/neureka/dense/src/bias.c
new file mode 100644
index 0000000..f6deb22
--- /dev/null
+++ b/neureka/dense/src/bias.c
@@ -0,0 +1,10 @@
+#include "bias.h"
+
+#define BIAS_SIZE (35)
+PI_L1 int32_t bias[BIAS_SIZE] = {
+    0x34f1, 0x26e29, -0x13522, -0x16519, -0x29e74, -0x26137, -0xa2d1, 0x224b4, -0x2b112, -0x5691, 
+    0x288e, -0x2eec7, 0x31b00, -0x13b14, 0x19292, 0x10f19, 0x2e986, 0x3492f, 0x6d7c, 0x19cbd, 
+    0x11dd9, -0xecc1, -0x9d88, 0xa5fb, -0x28ff8, -0x24d17, -0x1c933, 0x32120, -0x2ad82, -0x2f262, 
+    -0x14bbf, -0x155b3, 0x18b39, 0x1e105, 0x3041c
+};
+
diff --git a/neureka/dense/src/input.c b/neureka/dense/src/input.c
new file mode 100644
index 0000000..7d83843
--- /dev/null
+++ b/neureka/dense/src/input.c
@@ -0,0 +1,147 @@
+#include "input.h"
+
+#define INPUT_SIZE (1407)
+PI_L1 uint8_t input[INPUT_SIZE] = {
+    0xc2, 0x2c, 0xe2, 0x81, 0x81, 0x2d, 0xa9, 0xb3, 0x4c, 0x8a, 
+    0x32, 0x75, 0x40, 0xc1, 0xed, 0x7a, 0x9a, 0x7d, 0x5e, 0xdf, 
+    0xc8, 0x44, 0x83, 0xe8, 0x1e, 0x1b, 0xd2, 0x84, 0x39, 0xc5, 
+    0x2c, 0x27, 0x9b, 0x78, 0xb3, 0x70, 0x5a, 0x5b, 0xb5, 0x95, 
+    0x3c, 0x8f, 0xe4, 0x32, 0xfa, 0x12, 0xb7, 0xc7, 0x48, 0x91, 
+    0x02, 0x33, 0xd0, 0xbe, 0x57, 0xae, 0x61, 0xd0, 0x8a, 0x5d, 
+    0xd1, 0x72, 0x25, 0x0a, 0x81, 0x35, 0xd7, 0x36, 0xec, 0x04, 
+    0x37, 0xda, 0x99, 0xf4, 0x28, 0x46, 0xc8, 0xdd, 0x25, 0xae, 
+    0x37, 0x78, 0xf2, 0x15, 0x48, 0x57, 0xc2, 0x5b, 0xe1, 0x63, 
+    0x07, 0xab, 0xac, 0x7c, 0x2a, 0xe0, 0xbc, 0x75, 0xa8, 0xf0, 
+    0x48, 0x74, 0xe1, 0xad, 0xce, 0x17, 0x21, 0xf1, 0x72, 0x3d, 
+    0x08, 0x00, 0x44, 0x04, 0x46, 0xc7, 0x35, 0x39, 0x5f, 0xf9, 
+    0xdc, 0xad, 0x25, 0x23, 0xd5, 0x0f, 0xf0, 0x5c, 0x23, 0x08, 
+    0x25, 0x82, 0xd9, 0x76, 0xe9, 0xd1, 0x10, 0x9b, 0x13, 0x38, 
+    0x67, 0x8b, 0x96, 0x80, 0x7f, 0xe8, 0x58, 0xbf, 0x9e, 0xf3, 
+    0x34, 0x20, 0x06, 0xb6, 0xcd, 0x80, 0x3f, 0xf5, 0x2b, 0x95, 
+    0x90, 0x66, 0xa7, 0xc3, 0xf0, 0x01, 0x13, 0x11, 0x09, 0xbd, 
+    0x94, 0xfe, 0x52, 0x98, 0xda, 0x19, 0x19, 0x1a, 0x47, 0xca, 
+    0x5f, 0x7b, 0x64, 0xd8, 0x15, 0xb1, 0xcd, 0x3a, 0xe0, 0x29, 
+    0x61, 0x93, 0xa6, 0xb9, 0x99, 0x3b, 0x66, 0x46, 0x37, 0x4f, 
+    0x64, 0x87, 0xb3, 0xe3, 0x1c, 0xdf, 0x20, 0xc4, 0x48, 0xca, 
+    0x9c, 0xa8, 0x48, 0xd0, 0x52, 0x06, 0x3d, 0x87, 0x29, 0x23, 
+    0x99, 0x68, 0xed, 0x0a, 0x5c, 0xee, 0x5a, 0xec, 0xd6, 0x6e, 
+    0x65, 0x75, 0xa7, 0x0f, 0x17, 0x4b, 0xa0, 0xcf, 0xc4, 0xdc, 
+    0x28, 0xc2, 0x69, 0x4a, 0x61, 0x24, 0x55, 0xa2, 0x0c, 0xcd, 
+    0x6c, 0x79, 0xf8, 0xa1, 0xc9, 0x39, 0xc0, 0xe2, 0xc8, 0x00, 
+    0xe5, 0xbe, 0xf0, 0x81, 0x2c, 0xc4, 0xa9, 0x0c, 0xdf, 0xce, 
+    0xee, 0xe8, 0xb8, 0x62, 0x49, 0x07, 0xaa, 0x3c, 0xed, 0x30, 
+    0xd9, 0x77, 0x10, 0xd7, 0x32, 0x5e, 0x3d, 0xef, 0x27, 0x13, 
+    0x50, 0xb5, 0xb5, 0x21, 0x76, 0x14, 0x8f, 0x77, 0x7d, 0xb1, 
+    0xf8, 0xc3, 0x45, 0xfd, 0x35, 0x59, 0xa4, 0xca, 0xca, 0xbd, 
+    0x8d, 0xea, 0x83, 0x78, 0xa2, 0x6b, 0x11, 0xef, 0xae, 0x9d, 
+    0x1f, 0x9d, 0xcc, 0x98, 0x89, 0x22, 0x8b, 0x7d, 0xe8, 0xa7, 
+    0x61, 0x9e, 0x0c, 0xe3, 0x7e, 0xcc, 0x6d, 0x19, 0xbc, 0x71, 
+    0x8c, 0x01, 0x2a, 0x03, 0xa7, 0x81, 0x99, 0x12, 0x93, 0xc7, 
+    0x18, 0x3d, 0x66, 0x37, 0x98, 0x08, 0xf2, 0xb6, 0x0b, 0xa2, 
+    0x89, 0x65, 0x34, 0x07, 0x6e, 0x09, 0x84, 0xfe, 0x73, 0xf8, 
+    0x96, 0xbd, 0x09, 0x5c, 0x47, 0x1e, 0x0e, 0xa9, 0x58, 0xe7, 
+    0x5d, 0xc1, 0xdb, 0xe8, 0x67, 0x40, 0x21, 0x2b, 0x6a, 0x00, 
+    0x49, 0x57, 0xd1, 0x67, 0x18, 0xfa, 0x79, 0x87, 0xd1, 0x45, 
+    0x5a, 0xbb, 0x43, 0x3b, 0x2f, 0xd9, 0xbe, 0x8b, 0x61, 0x1f, 
+    0xc9, 0xa9, 0xe9, 0x10, 0xcb, 0x5b, 0x24, 0x82, 0x30, 0x5a, 
+    0x77, 0xe4, 0x2f, 0x40, 0x67, 0x55, 0xd1, 0x84, 0x29, 0x91, 
+    0x6f, 0x4b, 0x21, 0x94, 0xaa, 0x1f, 0x50, 0x1c, 0xc0, 0xb9, 
+    0x19, 0x0a, 0xd2, 0xe3, 0x7f, 0x91, 0x4d, 0x26, 0x93, 0x3b, 
+    0x01, 0xfd, 0x69, 0xba, 0x5d, 0xd0, 0x2b, 0x53, 0x6e, 0xd6, 
+    0x0f, 0x95, 0xde, 0x25, 0xbb, 0x3a, 0x6b, 0x36, 0x50, 0xa3, 
+    0xf9, 0x3d, 0x4d, 0xb2, 0x59, 0x49, 0xd1, 0xf3, 0x5c, 0x4e, 
+    0xd4, 0xb9, 0x6d, 0x1a, 0x72, 0x5e, 0x42, 0x92, 0x5b, 0xc8, 
+    0xdc, 0x89, 0x28, 0xe2, 0xae, 0xee, 0x61, 0xe5, 0x79, 0x3e, 
+    0x10, 0x7c, 0x62, 0x2b, 0x9b, 0xf6, 0x0c, 0x6e, 0x61, 0xc1, 
+    0xf1, 0x34, 0x77, 0x52, 0x0f, 0xa2, 0xce, 0x21, 0x27, 0x62, 
+    0xc6, 0xb3, 0xc1, 0x74, 0x43, 0xeb, 0xcb, 0x74, 0x82, 0x91, 
+    0xf9, 0x38, 0x81, 0x76, 0x91, 0x50, 0x8e, 0x96, 0x73, 0x14, 
+    0x61, 0x62, 0x95, 0x6d, 0x01, 0x16, 0x24, 0xb2, 0x66, 0x30, 
+    0x0e, 0x33, 0x47, 0x37, 0xa4, 0xc8, 0x0f, 0xc7, 0x1d, 0xed, 
+    0xc4, 0x01, 0x54, 0xa2, 0xbb, 0xfe, 0x15, 0x0c, 0x0d, 0x64, 
+    0xd5, 0x09, 0x59, 0xf4, 0x3d, 0x46, 0xfb, 0x0c, 0xc6, 0x6d, 
+    0x46, 0x2e, 0x84, 0x28, 0x4a, 0xed, 0x33, 0xad, 0xfd, 0x9b, 
+    0xfd, 0x99, 0xd8, 0x17, 0x89, 0xe0, 0x42, 0x71, 0x39, 0xba, 
+    0xa1, 0xfe, 0x23, 0xf3, 0x88, 0x95, 0xc3, 0x3a, 0xb8, 0x5b, 
+    0x0b, 0x5c, 0x0c, 0x23, 0x64, 0x4b, 0xe9, 0x0d, 0x0f, 0x33, 
+    0xd2, 0x20, 0x0d, 0x1f, 0x7a, 0x39, 0x2d, 0x8f, 0xc5, 0xd1, 
+    0x48, 0x85, 0xba, 0x06, 0x4c, 0xbe, 0xd4, 0xfb, 0x8d, 0x6a, 
+    0xd3, 0xe5, 0x3b, 0xeb, 0xa0, 0xfa, 0x1b, 0x30, 0xd3, 0x08, 
+    0x26, 0x2e, 0x0b, 0x11, 0x43, 0x46, 0xba, 0x8d, 0x4d, 0xf2, 
+    0x6e, 0xc0, 0x99, 0x63, 0xc8, 0x54, 0x0f, 0xbe, 0xcd, 0xd0, 
+    0x65, 0xc0, 0xad, 0x19, 0xcd, 0xdb, 0x9b, 0x29, 0x3a, 0xef, 
+    0xc6, 0x0a, 0x0d, 0xd1, 0xe7, 0x0a, 0xbe, 0x6d, 0xb9, 0x46, 
+    0x76, 0xdc, 0x51, 0x63, 0xbf, 0x08, 0x41, 0x47, 0xbc, 0x30, 
+    0x50, 0x00, 0xf8, 0xc6, 0x92, 0x30, 0xec, 0xf0, 0x31, 0xa2, 
+    0x4e, 0xac, 0xc8, 0x63, 0x14, 0x05, 0x62, 0x22, 0x5e, 0x29, 
+    0x6b, 0x5e, 0x73, 0xdb, 0x7e, 0x5a, 0xf8, 0x9f, 0xd6, 0xfc, 
+    0x05, 0x50, 0x14, 0xa9, 0xf3, 0x66, 0xe5, 0x92, 0xad, 0x1f, 
+    0xe4, 0x65, 0x8b, 0x3b, 0xf5, 0x46, 0x13, 0x71, 0x19, 0x2a, 
+    0xc2, 0xb4, 0x51, 0xb8, 0xd7, 0x99, 0x23, 0xe5, 0x2b, 0xb2, 
+    0xfc, 0xb9, 0x1e, 0xd9, 0x5a, 0x3c, 0x7d, 0x4d, 0x33, 0x61, 
+    0x2a, 0xdf, 0x1a, 0xb5, 0x1f, 0x6a, 0x82, 0x9b, 0xcc, 0xe0, 
+    0x90, 0x6a, 0x6d, 0x41, 0x11, 0xe7, 0x5c, 0xf1, 0x09, 0x3d, 
+    0x04, 0xe0, 0x61, 0x64, 0xba, 0x01, 0x89, 0xa4, 0xee, 0x49, 
+    0x6f, 0x16, 0x8d, 0x72, 0x51, 0x87, 0x1a, 0x56, 0x88, 0x5a, 
+    0x3c, 0x8f, 0xc6, 0xe9, 0xc8, 0xb2, 0xf4, 0x3b, 0x56, 0x19, 
+    0x8c, 0x44, 0x04, 0xe4, 0xa6, 0xb2, 0x05, 0xdd, 0x2b, 0xc0, 
+    0x26, 0x92, 0x6a, 0xe0, 0xe9, 0x43, 0xb4, 0xd6, 0x31, 0x03, 
+    0xfc, 0xd8, 0x60, 0xd8, 0xbe, 0x5d, 0x09, 0x5b, 0xea, 0x2e, 
+    0x3d, 0x19, 0x6a, 0x0b, 0xdb, 0xfc, 0x1a, 0x4c, 0xe7, 0xc4, 
+    0xe0, 0x3a, 0xed, 0xa3, 0xf3, 0x05, 0x29, 0x59, 0xf7, 0x49, 
+    0x45, 0x09, 0x82, 0xb9, 0x54, 0x64, 0x61, 0x5e, 0x73, 0x03, 
+    0xa5, 0xe1, 0x5a, 0x7f, 0x27, 0x3f, 0x9c, 0xf6, 0x11, 0x38, 
+    0xa3, 0x09, 0xc1, 0x2e, 0x2b, 0xb4, 0xed, 0x5a, 0xc9, 0x4b, 
+    0x9b, 0x4b, 0x9a, 0xfa, 0x73, 0xe5, 0xa1, 0xc2, 0x9c, 0xfb, 
+    0x7f, 0x15, 0x54, 0xc0, 0xfe, 0x13, 0xa6, 0x1d, 0x0a, 0xed, 
+    0xb2, 0x79, 0xcb, 0xcc, 0x04, 0xb8, 0x05, 0x99, 0x62, 0x98, 
+    0x43, 0xe8, 0x7d, 0xd0, 0xd1, 0xb1, 0xc1, 0x81, 0xb2, 0xe1, 
+    0x17, 0x2e, 0xd4, 0x54, 0x90, 0x65, 0xf9, 0x5b, 0xc4, 0xc8, 
+    0xee, 0x9a, 0x14, 0x0a, 0x72, 0x65, 0xa5, 0x3c, 0xc5, 0x2d, 
+    0xd5, 0xc1, 0x7d, 0x3f, 0x48, 0xd5, 0x41, 0xe6, 0x1e, 0x55, 
+    0x53, 0xa6, 0x7e, 0x3d, 0xd0, 0xe5, 0xf9, 0xb2, 0x55, 0x2b, 
+    0x21, 0xad, 0xac, 0x18, 0xdf, 0x7a, 0x3a, 0xc4, 0x5e, 0x62, 
+    0x94, 0x0b, 0x0d, 0xfd, 0x33, 0x0b, 0x59, 0x53, 0xaa, 0xbd, 
+    0xc6, 0x26, 0x6f, 0xae, 0xb8, 0x29, 0x43, 0xc4, 0x6a, 0x61, 
+    0x2d, 0x7f, 0x3e, 0x39, 0x81, 0xd0, 0xad, 0x34, 0xa9, 0x50, 
+    0x69, 0x46, 0x07, 0x69, 0xf3, 0xd3, 0x74, 0x74, 0x65, 0xb1, 
+    0x5d, 0x90, 0x8c, 0xb6, 0x39, 0xd5, 0xd1, 0x1b, 0x73, 0xd7, 
+    0xd4, 0x5d, 0xbb, 0x81, 0x37, 0x86, 0x29, 0x6b, 0x81, 0x1b, 
+    0x56, 0xb2, 0x90, 0x35, 0xc1, 0xc1, 0x8c, 0x64, 0x59, 0x3a, 
+    0x6b, 0xa2, 0x35, 0x33, 0x77, 0xf4, 0x14, 0xda, 0xbd, 0x92, 
+    0x3b, 0xa0, 0x9c, 0x68, 0xce, 0xb9, 0x6e, 0x58, 0x5a, 0x56, 
+    0x93, 0x26, 0x41, 0x0b, 0x23, 0x18, 0x35, 0x56, 0x6e, 0x1e, 
+    0x87, 0x16, 0x81, 0xca, 0x03, 0x5e, 0x43, 0x94, 0xec, 0xc5, 
+    0xdd, 0xb9, 0x81, 0xcd, 0x43, 0x54, 0x91, 0x1a, 0x1f, 0x99, 
+    0x83, 0x21, 0xb2, 0xf2, 0x13, 0x3e, 0x23, 0x1b, 0x43, 0xd8, 
+    0x51, 0x51, 0xe2, 0x06, 0x10, 0xc8, 0xb9, 0x46, 0x8e, 0x20, 
+    0x50, 0xa2, 0x6a, 0x5e, 0x2d, 0xa7, 0xf1, 0xa5, 0x47, 0x6c, 
+    0x65, 0x2f, 0xd4, 0xc9, 0x76, 0xe1, 0xef, 0xd7, 0x53, 0x21, 
+    0x52, 0x28, 0xc8, 0x3d, 0x61, 0x77, 0x11, 0x01, 0x71, 0x3b, 
+    0xdc, 0xac, 0x7f, 0xf8, 0xb0, 0xc0, 0x69, 0xda, 0xb1, 0x61, 
+    0x94, 0x61, 0x37, 0x96, 0xc7, 0xc7, 0x5b, 0xdd, 0xfb, 0x7f, 
+    0x00, 0xe0, 0x1f, 0xb4, 0x80, 0x74, 0xf9, 0x30, 0x97, 0xc9, 
+    0x12, 0xc5, 0x55, 0x82, 0x9b, 0xbf, 0xf0, 0x1d, 0xc7, 0x93, 
+    0x5a, 0x32, 0xc6, 0xf5, 0x7c, 0xbd, 0xfe, 0x67, 0x8b, 0x68, 
+    0x3d, 0x70, 0xd8, 0x24, 0x71, 0xc2, 0x01, 0x2a, 0x44, 0x93, 
+    0x84, 0x9d, 0x91, 0x69, 0xb4, 0x6d, 0xa7, 0x7a, 0xab, 0x12, 
+    0xcf, 0xab, 0xaf, 0x7c, 0x7b, 0x75, 0x53, 0xc9, 0x14, 0xb0, 
+    0x04, 0x9f, 0x40, 0x72, 0x96, 0xf8, 0x95, 0xb4, 0xda, 0x3e, 
+    0x1b, 0xf7, 0x3c, 0x91, 0x8d, 0x18, 0x52, 0x62, 0x2a, 0x4f, 
+    0xf4, 0x55, 0x7b, 0x4d, 0x15, 0x59, 0x71, 0xed, 0xf4, 0x25, 
+    0xf8, 0x26, 0x3f, 0xc6, 0x8f, 0x5b, 0x1d, 0x96, 0x3a, 0x13, 
+    0x24, 0x08, 0x5f, 0x5b, 0x69, 0x27, 0x5a, 0xbb, 0xb0, 0x86, 
+    0xdc, 0x84, 0x9d, 0x4a, 0x6b, 0x41, 0xd3, 0x87, 0x57, 0x08, 
+    0x1f, 0x22, 0x32, 0xc1, 0x7f, 0x2f, 0x84, 0xcb, 0xba, 0x86, 
+    0x27, 0x87, 0xa6, 0xa4, 0x1a, 0x99, 0xb5, 0xa5, 0xa1, 0xc3, 
+    0xa9, 0xa4, 0x03, 0x45, 0xc8, 0x2a, 0x94, 0xdf, 0x5c, 0x3b, 
+    0xc0, 0x72, 0x85, 0xf3, 0xe0, 0x1f, 0xad, 0x44, 0x66, 0x85, 
+    0xdd, 0x7b, 0xf3, 0x3a, 0x59, 0xd7, 0x81, 0x27, 0x65, 0xcf, 
+    0xba, 0xf0, 0x74, 0xfd, 0xf5, 0xf5, 0x12, 0xce, 0x31, 0xdf, 
+    0xf1, 0xdf, 0x22, 0x93, 0x6d, 0x23, 0x7f, 0xa3, 0xa7, 0x54, 
+    0xa0, 0x05, 0x10, 0xe5, 0x2d, 0x49, 0x14, 0x8f, 0x99, 0xe1, 
+    0xf5, 0x8e, 0x51, 0xe0, 0xe3, 0xcf, 0x56, 0x2f, 0x6e, 0xae, 
+    0xa5, 0x91, 0x4e, 0xe2, 0x73, 0x4c, 0xea, 0xc2, 0x25, 0x6a, 
+    0x8c, 0x6a, 0xb6, 0x26, 0x88, 0x28, 0xb5
+};
+
diff --git a/neureka/dense/src/output.c b/neureka/dense/src/output.c
new file mode 100644
index 0000000..ab9b752
--- /dev/null
+++ b/neureka/dense/src/output.c
@@ -0,0 +1,46 @@
+#include "output.h"
+
+#define OUTPUT_SIZE (175)
+PI_L1 uint8_t output[OUTPUT_SIZE];
+
+#define GOLDEN_OUTPUT_SIZE (175)
+PI_L2 uint8_t golden_output[GOLDEN_OUTPUT_SIZE] = {
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x10, 0x00, 0xa0, 0x00, 0xdc, 
+    0x08, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0xff, 0x00, 0x43, 
+    0x2d, 0x28, 0x00, 0xba, 0xad, 0x00, 0x00, 0x00, 0x05, 0x80, 
+    0x00, 0x1b, 0x00, 0x1a, 0xa1, 0x00, 0x6f, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x6b, 0x00, 0x96, 0x00, 0x00, 0x00, 0x00, 0x28, 
+    0x00, 0x1e, 0xff, 0x51, 0x35, 0x0f, 0x00, 0x00, 0x00, 0xff, 
+    0x00, 0x00, 0x00, 0x00, 0x5b, 0x00, 0x00, 0x00, 0x1f, 0x00, 
+    0x00, 0x00, 0x00, 0x52, 0x08, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0xba, 0x06, 0x1b, 
+    0x13, 0x00, 0x00, 0xa6, 0x00, 0x00, 0x00, 0x00, 0x00, 0xc3, 
+    0x00, 0x00, 0x22, 0x26, 0xa2, 0x00, 0x44, 0x00, 0x0c, 0x00, 
+    0x3a, 0x00, 0x00, 0x00, 0xf7, 0x00, 0x00, 0x84, 0x00, 0x00, 
+    0x00, 0x00, 0xec, 0x03, 0x62, 0x2c, 0x00, 0x00, 0x00, 0xa9, 
+    0x00, 0x00, 0x00, 0x01, 0x09, 0x00, 0x0b, 0x00, 0x25, 0xa0, 
+    0x00, 0x00, 0x00, 0x13, 0x00, 0x17, 0x00, 0x00, 0x00, 0xff, 
+    0x00, 0x00, 0x00, 0x47, 0x00, 0x00, 0x00, 0xff, 0x00, 0x78, 
+    0x17, 0x04, 0x00, 0xdc, 0x76, 0x00, 0x00, 0x00, 0x00, 0xcd, 
+    0x00, 0x00, 0x06, 0x1b, 0x48
+};
+
+int check_output() {
+        printf("Checking the output vector:\n");
+
+        int n_err = 0;
+        for (int i = 0; i < OUTPUT_SIZE; i++) {
+            if (output[i] != golden_output[i]) {
+                printf("ERROR: wrong value of output @ %d: %d vs. golden: %d\n", i, output[i], golden_output[i]);
+                n_err++;
+            }
+        }
+
+        if (n_err == 0)
+            printf("> Success! No errors found.\n");
+        else
+            printf("> Failure! Found %d/%d errors.\n", n_err, OUTPUT_SIZE);
+        return n_err;
+    }
+
+    
\ No newline at end of file
diff --git a/neureka/dense/src/scale.c b/neureka/dense/src/scale.c
new file mode 100644
index 0000000..33769cd
--- /dev/null
+++ b/neureka/dense/src/scale.c
@@ -0,0 +1,10 @@
+#include "scale.h"
+
+#define SCALE_SIZE (35)
+PI_L1 uint8_t scale[SCALE_SIZE] = {
+    0x06, 0x07, 0x14, 0x06, 0x10, 0x0b, 0x1d, 0x14, 0x01, 0x1b, 
+    0x03, 0x04, 0x17, 0x1c, 0x05, 0x13, 0x06, 0x0f, 0x0c, 0x0c, 
+    0x02, 0x04, 0x0e, 0x19, 0x0b, 0x1a, 0x12, 0x10, 0x06, 0x0b, 
+    0x09, 0x16, 0x06, 0x01, 0x0c
+};
+
diff --git a/neureka/dense/src/weight.c b/neureka/dense/src/weight.c
new file mode 100644
index 0000000..811eb11
--- /dev/null
+++ b/neureka/dense/src/weight.c
@@ -0,0 +1,3030 @@
+#include "weight.h"
+
+#define WEIGHT_SIZE (30240)
+PI_L1 uint8_t weight[WEIGHT_SIZE] = {
+    0xfa, 0x5e, 0x7d, 0x0e, 0x0d, 0x0d, 0x03, 0xc1, 0x56, 0x6a, 
+    0x95, 0x9e, 0x1e, 0xdf, 0x88, 0x4d, 0xbe, 0x88, 0xc4, 0x67, 
+    0x6e, 0x2e, 0x9e, 0x60, 0xcc, 0xb4, 0x25, 0x04, 0xe5, 0xfc, 
+    0x82, 0x6b, 0x1e, 0x34, 0x8a, 0x08, 0x47, 0xda, 0xe8, 0xeb, 
+    0x39, 0x54, 0x10, 0x81, 0x36, 0x76, 0x53, 0xa3, 0x87, 0x45, 
+    0x86, 0xd4, 0x9b, 0xb1, 0x58, 0xb8, 0xad, 0xa8, 0x47, 0x0a, 
+    0x0c, 0xde, 0x20, 0xc5, 0xae, 0xf5, 0xaa, 0xaf, 0x29, 0x92, 
+    0x48, 0x40, 0xd2, 0x79, 0x9d, 0x5c, 0xa5, 0x8e, 0x6d, 0x1a, 
+    0x0b, 0x12, 0xf4, 0x03, 0xa3, 0x1f, 0x45, 0x3c, 0x16, 0xd5, 
+    0xb6, 0xe1, 0x25, 0xce, 0x4a, 0x1d, 0xe4, 0xc8, 0xd6, 0xda, 
+    0x54, 0xfa, 0x6c, 0x63, 0x35, 0x4b, 0x0b, 0x3a, 0x84, 0xe5, 
+    0x8c, 0x89, 0x18, 0xb9, 0x94, 0x6d, 0x01, 0xf2, 0x5d, 0x39, 
+    0xc0, 0x2d, 0x4d, 0x97, 0x42, 0xaa, 0x49, 0x6b, 0x2b, 0xa0, 
+    0xd4, 0x4b, 0x06, 0xeb, 0x48, 0xab, 0xc5, 0x11, 0x2b, 0xf8, 
+    0xb9, 0xc8, 0xa7, 0xb0, 0xe5, 0xe8, 0x69, 0x5b, 0xe8, 0xaf, 
+    0xdf, 0xa2, 0x32, 0x3a, 0xc6, 0x20, 0x86, 0xc9, 0x2d, 0x81, 
+    0xa3, 0xd2, 0x07, 0xd2, 0x43, 0x24, 0xd8, 0x02, 0x0c, 0x37, 
+    0x48, 0x6f, 0x5f, 0x83, 0x47, 0x76, 0xd5, 0xf6, 0xe6, 0xb1, 
+    0xef, 0xcb, 0x5a, 0x39, 0xc0, 0x85, 0x0e, 0xa9, 0x30, 0x39, 
+    0x75, 0x01, 0x06, 0xc7, 0x29, 0x35, 0xa3, 0x98, 0x56, 0xd9, 
+    0x40, 0x54, 0xe8, 0x0d, 0x4f, 0xb7, 0x68, 0xfb, 0x57, 0x82, 
+    0x2f, 0x56, 0x91, 0x70, 0xe2, 0xab, 0xe7, 0xcb, 0x5a, 0x19, 
+    0xc0, 0xa5, 0x0f, 0xa9, 0xb0, 0x39, 0x75, 0x01, 0x06, 0xcf, 
+    0x2d, 0xb5, 0xa3, 0x98, 0x56, 0xd9, 0x49, 0x54, 0xe8, 0x0b, 
+    0x4c, 0x37, 0x68, 0xfb, 0x57, 0x83, 0x2f, 0x56, 0x91, 0x70, 
+    0xe2, 0xbb, 0x18, 0x34, 0xa5, 0xe6, 0x3f, 0x5a, 0xf0, 0x56, 
+    0xcf, 0xc6, 0x8a, 0xfe, 0xf9, 0x30, 0xd2, 0x4a, 0x5c, 0x67, 
+    0xa9, 0x26, 0xb6, 0xab, 0x17, 0xf4, 0xb3, 0xc8, 0x97, 0x04, 
+    0xa8, 0x7c, 0xd0, 0xa9, 0x6e, 0x8f, 0x1d, 0x44, 0xe0, 0x3e, 
+    0x07, 0xbc, 0x00, 0x64, 0x58, 0x14, 0x26, 0xc1, 0x52, 0x73, 
+    0x88, 0xf8, 0x22, 0xa9, 0x3c, 0x42, 0x0d, 0x4a, 0xec, 0xf8, 
+    0xc2, 0xd7, 0x29, 0x6d, 0x02, 0x04, 0xc4, 0x15, 0x40, 0x12, 
+    0xe1, 0xc4, 0x5c, 0x34, 0x8f, 0x46, 0x86, 0xbc, 0x93, 0xf4, 
+    0xeb, 0x2d, 0xcb, 0x67, 0xb6, 0x56, 0x32, 0x8c, 0xe0, 0x87, 
+    0x3e, 0x17, 0x06, 0xe6, 0x57, 0xf9, 0xfc, 0x7f, 0x6d, 0x3d, 
+    0x08, 0x83, 0x1c, 0x36, 0x47, 0x2b, 0x55, 0xf5, 0x89, 0x66, 
+    0xcf, 0xf5, 0x5a, 0x34, 0x95, 0x7b, 0x6f, 0xfa, 0x1a, 0x5b, 
+    0x1b, 0x05, 0x62, 0x8d, 0x0a, 0x21, 0x83, 0x83, 0x22, 0x3c, 
+    0x39, 0x47, 0x24, 0x17, 0x8f, 0xea, 0xce, 0x2d, 0xdf, 0x47, 
+    0x86, 0x49, 0x4f, 0xba, 0x0e, 0x2f, 0xa5, 0xa5, 0x1b, 0x3d, 
+    0xc6, 0x20, 0x4f, 0xf0, 0x86, 0x62, 0x47, 0xed, 0x22, 0x19, 
+    0x71, 0xad, 0x27, 0xaf, 0xe9, 0x56, 0x93, 0xf5, 0x5c, 0x47, 
+    0xff, 0xce, 0xbf, 0x2a, 0xa8, 0x95, 0x9a, 0xb5, 0x17, 0xc1, 
+    0xa6, 0x3b, 0x6e, 0xd9, 0x6a, 0xde, 0xf3, 0x7d, 0xd8, 0x33, 
+    0xca, 0x0d, 0xa9, 0xd2, 0x48, 0x38, 0x3b, 0x84, 0x48, 0x60, 
+    0xb7, 0x2c, 0x8a, 0x85, 0xb8, 0xcb, 0x01, 0xe5, 0xe5, 0x94, 
+    0xbf, 0xb4, 0x5b, 0x60, 0x98, 0xfb, 0xe6, 0xae, 0x45, 0xf1, 
+    0x6b, 0xbe, 0x65, 0x49, 0x54, 0x33, 0xd7, 0x32, 0x2f, 0xc0, 
+    0x5c, 0x2a, 0x31, 0xa7, 0x0c, 0xe6, 0xaf, 0x2a, 0x9f, 0x64, 
+    0x7c, 0x57, 0xab, 0xbb, 0xa7, 0x15, 0xbc, 0x92, 0x84, 0x2f, 
+    0x81, 0x87, 0xe3, 0x8f, 0x45, 0xf1, 0x6b, 0xbe, 0x65, 0x69, 
+    0x5c, 0x33, 0xc6, 0x32, 0x2f, 0xc0, 0x5c, 0x28, 0x39, 0xa5, 
+    0x0c, 0xe4, 0xaf, 0x2e, 0x9b, 0x64, 0x78, 0x47, 0xab, 0xbf, 
+    0xe7, 0x15, 0xbc, 0x94, 0xc0, 0x2f, 0x90, 0x87, 0xe3, 0x8f, 
+    0xba, 0x0e, 0x94, 0x41, 0x9a, 0x96, 0xa3, 0xcc, 0x39, 0xcd, 
+    0xd0, 0x3f, 0xa3, 0xd7, 0xc6, 0x5a, 0xf3, 0x1b, 0x50, 0xd1, 
+    0x64, 0x9b, 0x87, 0xb8, 0x54, 0x40, 0x18, 0xea, 0x43, 0x6b, 
+    0x3f, 0xd0, 0x6f, 0x78, 0x1c, 0x70, 0x07, 0x00, 0x00, 0x00, 
+    0x03, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x03, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 
+    0x06, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x06, 0x00, 
+    0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x03, 0x00, 
+    0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x05, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x06, 0x00, 
+    0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 
+    0x04, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 
+    0x02, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x07, 0x00, 
+    0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x06, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 
+    0x04, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x07, 0x00, 
+    0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 
+    0x02, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x07, 0x00, 
+    0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x02, 0x00, 
+    0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x06, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x05, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x06, 0x00, 
+    0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 
+    0x03, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x05, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x04, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0xfd, 0x34, 0x94, 0x6f, 0x49, 0x64, 
+    0x1b, 0xa1, 0xe6, 0x63, 0x32, 0x22, 0xb6, 0xbd, 0xf8, 0x2e, 
+    0xe9, 0x90, 0x26, 0xe4, 0xd1, 0x61, 0x06, 0x16, 0x02, 0x9c, 
+    0x3f, 0xb4, 0xc2, 0x90, 0x37, 0x4c, 0xdc, 0xbf, 0x0e, 0xf0, 
+    0x93, 0x83, 0x4c, 0x08, 0xfd, 0x5a, 0x0f, 0x36, 0xec, 0x00, 
+    0x02, 0x9d, 0xb8, 0xcf, 0xfa, 0x1e, 0x66, 0xfc, 0x57, 0x69, 
+    0x08, 0xaa, 0x8a, 0x14, 0x6d, 0x0b, 0x05, 0xc9, 0x9a, 0xb5, 
+    0xd9, 0xf6, 0x16, 0x84, 0xb7, 0xae, 0x0e, 0xd9, 0x94, 0x25, 
+    0xce, 0x28, 0xde, 0xbf, 0x96, 0xb2, 0xa5, 0x89, 0x31, 0xa3, 
+    0xd8, 0x32, 0xe6, 0x63, 0x4c, 0x89, 0x37, 0xd7, 0x84, 0x2e, 
+    0x4a, 0x5f, 0x4b, 0xcb, 0x0c, 0x0c, 0x51, 0x4d, 0x94, 0x97, 
+    0x6a, 0xc4, 0x02, 0xf8, 0x74, 0xee, 0xaa, 0x26, 0x41, 0xc2, 
+    0xbd, 0xc2, 0xe5, 0x68, 0x98, 0xfd, 0x8e, 0xe2, 0x90, 0x24, 
+    0xe4, 0x1b, 0x25, 0x4a, 0xbe, 0xd3, 0xde, 0xad, 0x63, 0x39, 
+    0x76, 0x6b, 0x4a, 0xbe, 0xbf, 0x16, 0xc9, 0xb2, 0xcb, 0x32, 
+    0x3d, 0x82, 0xf6, 0xaa, 0xec, 0x36, 0x21, 0x4d, 0x0b, 0x87, 
+    0x87, 0x90, 0x2c, 0x82, 0x47, 0xbb, 0x5c, 0xd0, 0x9b, 0xc5, 
+    0x19, 0xb8, 0x3b, 0xa6, 0xfd, 0x00, 0x16, 0x9a, 0x84, 0xa6, 
+    0xcc, 0x84, 0xd6, 0x44, 0x1a, 0xe8, 0x0d, 0xfb, 0xb7, 0x8b, 
+    0x6c, 0x32, 0x20, 0x4c, 0x03, 0x3f, 0x83, 0xc8, 0xa2, 0xa2, 
+    0x65, 0x2f, 0x2d, 0x72, 0xbf, 0xc5, 0x19, 0xaa, 0x7b, 0x8c, 
+    0xae, 0x01, 0x5b, 0x9e, 0xcc, 0xce, 0x9c, 0xb6, 0xc4, 0xf4, 
+    0x9a, 0xe8, 0x0d, 0xea, 0xb7, 0x8a, 0x6c, 0x32, 0x20, 0x4c, 
+    0x03, 0x3e, 0x87, 0xd8, 0xa2, 0xa2, 0x65, 0x2f, 0x2c, 0x72, 
+    0xbf, 0xc5, 0x19, 0xba, 0x7b, 0x8c, 0xef, 0x01, 0x5b, 0x9a, 
+    0xcc, 0xce, 0x9c, 0xb6, 0xc4, 0xf4, 0x65, 0x17, 0xf2, 0x15, 
+    0x48, 0x75, 0x93, 0xcd, 0xdf, 0xb3, 0xfc, 0xc1, 0x78, 0x27, 
+    0x5d, 0x5d, 0x9a, 0xd0, 0xd3, 0x8d, 0x40, 0x3a, 0xe6, 0x45, 
+    0x84, 0x73, 0x10, 0xfe, 0xa4, 0x65, 0x33, 0x31, 0x63, 0x49, 
+    0x3b, 0x0b, 0xd1, 0xe4, 0x92, 0xa8, 0x5e, 0x6f, 0x63, 0xa3, 
+    0x75, 0xb2, 0xa0, 0x2d, 0xaf, 0x91, 0x08, 0x93, 0x07, 0xff, 
+    0xd8, 0x41, 0xe7, 0x59, 0x5d, 0xad, 0xd4, 0x72, 0xc4, 0x0d, 
+    0x48, 0x2b, 0xfc, 0xc2, 0xa0, 0x27, 0xdf, 0x1c, 0x6b, 0xcb, 
+    0x67, 0xc3, 0x6e, 0x6f, 0xf7, 0xf5, 0x85, 0x76, 0x87, 0xec, 
+    0xd9, 0xb0, 0x1f, 0xdd, 0xe5, 0xa6, 0x32, 0x8a, 0x6c, 0xdb, 
+    0x74, 0xe0, 0x58, 0x5e, 0xce, 0x0e, 0xa2, 0x69, 0xa4, 0x5d, 
+    0x7e, 0x26, 0x1f, 0x16, 0x22, 0xbd, 0x01, 0x57, 0x83, 0x76, 
+    0x5c, 0x81, 0x35, 0x06, 0x15, 0xcd, 0x59, 0xf3, 0x3b, 0xbf, 
+    0x03, 0xaa, 0x51, 0xd7, 0x74, 0xf8, 0x8c, 0x8c, 0xd7, 0x5e, 
+    0x4a, 0xa1, 0x3f, 0x3b, 0x3f, 0xca, 0x38, 0xd5, 0xf3, 0x7b, 
+    0x65, 0x34, 0xf7, 0x9a, 0x58, 0x66, 0xe8, 0xc6, 0xb1, 0x3b, 
+    0x4d, 0x19, 0xfe, 0xcb, 0xd7, 0xb2, 0xa2, 0xdf, 0x51, 0x59, 
+    0xb1, 0x73, 0x2e, 0xd5, 0x92, 0xfc, 0x05, 0x60, 0xe0, 0x6a, 
+    0x2e, 0x7d, 0x12, 0x2d, 0x66, 0xfc, 0x28, 0xf2, 0x44, 0x64, 
+    0xd4, 0xac, 0x8a, 0x94, 0x4f, 0x80, 0xc7, 0x0d, 0xaa, 0xcd, 
+    0xc1, 0x90, 0xd5, 0x43, 0x2a, 0xbc, 0x91, 0xd6, 0x30, 0x89, 
+    0xea, 0x79, 0x48, 0xcb, 0xa0, 0xb0, 0x4f, 0x20, 0xc1, 0x7d, 
+    0xd1, 0x70, 0x65, 0xfa, 0xc6, 0x70, 0xca, 0x26, 0x93, 0x1a, 
+    0x3a, 0x33, 0xc2, 0x49, 0x3a, 0x01, 0x11, 0xb2, 0x9e, 0x4b, 
+    0xa4, 0x78, 0x7c, 0xf5, 0x3c, 0xd4, 0xc6, 0x71, 0x40, 0xc2, 
+    0x62, 0xea, 0xcf, 0x63, 0x11, 0xfd, 0x8a, 0xf8, 0x65, 0xfa, 
+    0x46, 0x70, 0xca, 0xa6, 0x92, 0x12, 0x3a, 0x22, 0xc6, 0x49, 
+    0x2a, 0x01, 0x11, 0xb2, 0x9e, 0x4b, 0xa0, 0x78, 0x7c, 0xf5, 
+    0x3c, 0xd4, 0xc6, 0x71, 0x40, 0xc2, 0xe2, 0xea, 0xcf, 0x63, 
+    0x11, 0xfd, 0x82, 0xf8, 0x9a, 0x05, 0xb9, 0x8f, 0x35, 0x59, 
+    0x6d, 0xed, 0xc5, 0xdd, 0x39, 0xb6, 0xd5, 0xfe, 0xee, 0x4d, 
+    0x61, 0xb4, 0x5f, 0x87, 0x83, 0x0a, 0xc3, 0x2b, 0x39, 0x8e, 
+    0xbf, 0x3d, 0x1d, 0x15, 0x30, 0x9c, 0xee, 0x02, 0x7d, 0x07, 
+    0x06, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 
+    0x06, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x04, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 
+    0x03, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x07, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 
+    0x06, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x03, 0x00, 
+    0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x04, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x03, 0x00, 
+    0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 
+    0x02, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x02, 0x00, 
+    0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x03, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x06, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x06, 0x00, 
+    0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 
+    0x03, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x07, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x03, 0x00, 
+    0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 
+    0x07, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0xe7, 0xf1, 
+    0xe5, 0x0b, 0xa8, 0xa0, 0x23, 0x89, 0xee, 0xe0, 0x4c, 0x00, 
+    0x99, 0xe0, 0xeb, 0xdd, 0x41, 0x87, 0x83, 0x61, 0x4c, 0x85, 
+    0xcb, 0x5d, 0x7a, 0xfe, 0xe6, 0x2c, 0x49, 0x58, 0xc0, 0xc1, 
+    0xb3, 0xbe, 0xdc, 0xd4, 0xce, 0xbe, 0x81, 0x22, 0xc2, 0x79, 
+    0x13, 0x7c, 0xd4, 0x38, 0xaa, 0x8d, 0x91, 0x64, 0xbb, 0x29, 
+    0xd4, 0xff, 0x36, 0x9c, 0x66, 0xc0, 0x19, 0x0d, 0x36, 0x82, 
+    0x49, 0x39, 0xfd, 0xd8, 0xf7, 0xa0, 0xd4, 0xa8, 0xa7, 0x44, 
+    0xdd, 0x66, 0x47, 0xd9, 0x12, 0x58, 0x50, 0x11, 0x5c, 0x9b, 
+    0x0c, 0xd5, 0x06, 0xb0, 0xab, 0xac, 0x44, 0xb3, 0xb5, 0xb9, 
+    0x7f, 0x03, 0x67, 0x35, 0x7b, 0xed, 0xe2, 0xb0, 0xf4, 0xbc, 
+    0x42, 0xf1, 0x60, 0x64, 0x26, 0x22, 0xc3, 0xf3, 0x20, 0x89, 
+    0x94, 0xcc, 0x12, 0xcf, 0x3d, 0x25, 0x3c, 0xc3, 0x57, 0xa1, 
+    0xd9, 0x8e, 0xfc, 0x32, 0xd8, 0xc8, 0x95, 0x72, 0xb2, 0xd9, 
+    0xf0, 0xbc, 0x72, 0x61, 0x54, 0x64, 0xed, 0xce, 0x34, 0x7b, 
+    0x38, 0xf8, 0xd6, 0x11, 0xce, 0x37, 0x0a, 0xd0, 0x6e, 0x84, 
+    0xae, 0xa6, 0xfa, 0x88, 0xb2, 0x77, 0x28, 0x57, 0x0c, 0x80, 
+    0xc2, 0x7e, 0xe1, 0xc9, 0x03, 0xd5, 0x6f, 0xa6, 0xfc, 0x1a, 
+    0xee, 0x64, 0x7b, 0x27, 0x6f, 0xf0, 0x73, 0x98, 0x33, 0xc3, 
+    0xe3, 0x1c, 0xba, 0xd8, 0x24, 0xd4, 0x7c, 0xb7, 0xfc, 0x8b, 
+    0xf2, 0xe0, 0x81, 0x5d, 0x26, 0x9c, 0xf6, 0xf9, 0x39, 0x43, 
+    0x13, 0xdc, 0x7f, 0xa2, 0x75, 0x81, 0xfc, 0xa4, 0x79, 0x45, 
+    0x2f, 0x83, 0x76, 0x74, 0x33, 0xc3, 0xe7, 0x1c, 0xba, 0xd8, 
+    0x24, 0x54, 0x7c, 0xa7, 0xfc, 0x8b, 0xf2, 0xe0, 0x81, 0x5d, 
+    0x26, 0x92, 0xf6, 0xf8, 0x31, 0x43, 0x13, 0xdc, 0x7f, 0xa2, 
+    0x74, 0x81, 0xfc, 0x24, 0x79, 0x45, 0x2f, 0xb3, 0x76, 0xf4, 
+    0xcc, 0x3c, 0x18, 0xe3, 0x45, 0x27, 0xdb, 0x2b, 0x83, 0x58, 
+    0x03, 0x74, 0x0d, 0x1f, 0x7e, 0xa2, 0xd9, 0x6d, 0x09, 0x07, 
+    0xce, 0xbc, 0xec, 0x23, 0x80, 0x5d, 0x8b, 0x7e, 0x03, 0xdb, 
+    0x86, 0xba, 0xd0, 0x4c, 0x89, 0x0b, 0x15, 0xd9, 0x6c, 0x4d, 
+    0xa7, 0xc9, 0x3c, 0x21, 0xe9, 0x81, 0xdb, 0xba, 0x86, 0xbf, 
+    0xb7, 0x72, 0x99, 0xf6, 0x0c, 0xe0, 0xf9, 0x8b, 0x46, 0x87, 
+    0x1a, 0xdd, 0xd5, 0x0e, 0x1c, 0x00, 0x19, 0x56, 0xbe, 0xfe, 
+    0xa6, 0x6c, 0xdb, 0xe8, 0x2c, 0x4e, 0x09, 0x19, 0x79, 0x66, 
+    0x9b, 0x44, 0xe7, 0xb3, 0x7d, 0x70, 0x3f, 0x2b, 0x7b, 0xe1, 
+    0x65, 0xf3, 0x3f, 0x30, 0x0b, 0x42, 0x5e, 0x90, 0x23, 0x62, 
+    0x98, 0x4d, 0x10, 0x90, 0x92, 0x62, 0xb8, 0x31, 0x2c, 0x9b, 
+    0x15, 0xb8, 0x38, 0x53, 0x89, 0x9f, 0xdd, 0xff, 0x79, 0x16, 
+    0xd7, 0x54, 0xc5, 0x7c, 0x51, 0xd5, 0xdf, 0x9b, 0xa0, 0x1a, 
+    0xf5, 0x15, 0x95, 0x2d, 0x0c, 0xc4, 0xcb, 0x48, 0x5c, 0xf0, 
+    0x2b, 0x13, 0xa8, 0x00, 0xda, 0x48, 0xf8, 0x96, 0x3a, 0x81, 
+    0x7b, 0x53, 0x91, 0x49, 0xe8, 0x83, 0xba, 0x8c, 0x59, 0xca, 
+    0x12, 0xe0, 0x23, 0xf4, 0x5b, 0xd7, 0xb5, 0xff, 0xc2, 0xc5, 
+    0xa0, 0xb9, 0xed, 0xa3, 0x5e, 0x9c, 0x9e, 0x37, 0x66, 0xa1, 
+    0x8b, 0xd5, 0x3e, 0xfd, 0x0a, 0x15, 0x3f, 0xb5, 0x71, 0x3c, 
+    0x3b, 0xbf, 0xff, 0xde, 0xe4, 0x8b, 0x6a, 0xf1, 0x34, 0x35, 
+    0x9a, 0x32, 0xe5, 0xd2, 0x6b, 0x10, 0xdf, 0xe7, 0x8b, 0xaf, 
+    0x5a, 0x5c, 0x23, 0xc9, 0x07, 0xa6, 0x73, 0x71, 0xe2, 0x96, 
+    0x9f, 0x14, 0x9a, 0xf5, 0xb0, 0xa4, 0x1d, 0x13, 0xef, 0xff, 
+    0x64, 0x3b, 0xf4, 0xf9, 0xb8, 0x31, 0x9b, 0x71, 0xf1, 0x52, 
+    0xfe, 0x92, 0xdf, 0xe7, 0x81, 0xe7, 0xf6, 0x71, 0x2e, 0xca, 
+    0x1f, 0xa2, 0x53, 0x71, 0xfa, 0x96, 0x9f, 0x14, 0x9a, 0xf5, 
+    0xb1, 0xa4, 0x1d, 0x13, 0xef, 0xfe, 0x64, 0x3a, 0xf0, 0xf9, 
+    0x38, 0x31, 0x9b, 0x71, 0xf5, 0x52, 0xfa, 0x92, 0xdf, 0xe3, 
+    0x81, 0xa7, 0x76, 0x75, 0x2a, 0xc2, 0x1f, 0xa6, 0xac, 0x8e, 
+    0x05, 0x69, 0x60, 0xeb, 0x65, 0x0a, 0x4e, 0x5b, 0xe2, 0xec, 
+    0x10, 0x01, 0x9b, 0xc5, 0x0f, 0x06, 0xc7, 0xce, 0x64, 0x8e, 
+    0x0a, 0xad, 0x05, 0x6d, 0x20, 0x1c, 0x7e, 0x58, 0x89, 0x8a, 
+    0xd5, 0x3d, 0xe0, 0x59, 0x05, 0x00, 0x00, 0x00, 0x04, 0x00, 
+    0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x06, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 
+    0x05, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x06, 0x00, 
+    0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x04, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 
+    0x02, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x03, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x06, 0x00, 
+    0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x03, 0x00, 
+    0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 
+    0x05, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 
+    0x03, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 
+    0x05, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 
+    0x03, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x06, 0x00, 
+    0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x02, 0x00, 
+    0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 
+    0x02, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x02, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 
+    0x06, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0xc4, 0x5f, 0xfb, 0xa9, 0x84, 0x01, 0x52, 0x16, 
+    0x8c, 0xdd, 0x09, 0x32, 0xb6, 0x6b, 0xbb, 0x8f, 0xda, 0x5e, 
+    0x28, 0x68, 0x75, 0xbe, 0x9e, 0xcb, 0x7e, 0xf3, 0x20, 0x46, 
+    0xdf, 0x94, 0xfe, 0x87, 0x86, 0xd0, 0xf0, 0x44, 0x53, 0xa8, 
+    0x66, 0x05, 0x1c, 0xf5, 0x14, 0xfb, 0xae, 0xff, 0x1b, 0xac, 
+    0xfd, 0xe5, 0x68, 0x14, 0x46, 0x87, 0x11, 0x9c, 0x72, 0x2a, 
+    0xa9, 0x3e, 0xd9, 0x2b, 0x54, 0xfc, 0x3f, 0x82, 0x84, 0xfb, 
+    0xc3, 0x27, 0x8e, 0x93, 0xff, 0x3d, 0x67, 0x9e, 0x1c, 0xbc, 
+    0x05, 0x23, 0xf7, 0xc3, 0x16, 0x4f, 0x1e, 0x38, 0x72, 0x0c, 
+    0x83, 0x05, 0x1e, 0xb1, 0x64, 0x9f, 0x80, 0xd2, 0x58, 0xde, 
+    0x02, 0xd9, 0x6f, 0x7b, 0x8c, 0x3f, 0x74, 0x86, 0x90, 0xcf, 
+    0x54, 0xa1, 0x90, 0x64, 0x3b, 0x31, 0xef, 0x18, 0x11, 0xee, 
+    0x18, 0xb2, 0xf6, 0x85, 0x03, 0x2f, 0xb3, 0x61, 0x79, 0xb7, 
+    0x84, 0x2d, 0xef, 0xdc, 0xb8, 0xdb, 0x0a, 0xc9, 0xae, 0x8e, 
+    0x2f, 0x9b, 0x61, 0x01, 0xd0, 0x0f, 0x5f, 0xb2, 0x41, 0x00, 
+    0x30, 0x2c, 0x43, 0x29, 0x64, 0x47, 0xe0, 0x2c, 0x3f, 0x77, 
+    0xd0, 0xb3, 0x26, 0x43, 0x13, 0x00, 0x88, 0x03, 0x0e, 0x37, 
+    0x9b, 0x47, 0xfd, 0x50, 0xe3, 0x62, 0x6b, 0x63, 0x98, 0x53, 
+    0xdc, 0xd4, 0x5c, 0xb1, 0x71, 0x03, 0x1a, 0x2d, 0x8f, 0x19, 
+    0x27, 0xdd, 0x78, 0xa8, 0x03, 0xf5, 0x9c, 0x26, 0x80, 0xc1, 
+    0x11, 0x07, 0x8a, 0x2f, 0x6e, 0xf2, 0x83, 0xf1, 0x23, 0x50, 
+    0x61, 0x6e, 0x62, 0x61, 0x3c, 0x43, 0xd6, 0xc6, 0x5c, 0xb1, 
+    0x71, 0x00, 0x1a, 0x2c, 0xcf, 0x19, 0x27, 0xdf, 0x78, 0xa8, 
+    0x33, 0xf5, 0x90, 0x27, 0x02, 0xc1, 0x11, 0x07, 0x8a, 0x2b, 
+    0x4e, 0xf2, 0x83, 0xd1, 0x23, 0x50, 0x63, 0x6a, 0x26, 0x61, 
+    0x3c, 0x43, 0xd6, 0xc6, 0xa3, 0x4e, 0x8e, 0xff, 0xe5, 0xd3, 
+    0x30, 0xe6, 0xd8, 0x20, 0x87, 0x57, 0xcc, 0x0a, 0x6f, 0xd8, 
+    0xfd, 0x3e, 0xee, 0xf8, 0x75, 0xd4, 0xb1, 0x0d, 0x7c, 0x2e, 
+    0xdc, 0xaf, 0x9c, 0x95, 0xd9, 0x9e, 0xc3, 0xbc, 0x29, 0x39, 
+    0x6d, 0xb7, 0x3d, 0x1e, 0x6e, 0x69, 0xa5, 0xf0, 0xbf, 0x8b, 
+    0xe6, 0x03, 0x2f, 0x9f, 0x76, 0xe9, 0xb9, 0x2a, 0x5b, 0x16, 
+    0x0c, 0xee, 0x65, 0xca, 0x7f, 0x9d, 0x44, 0xc1, 0x39, 0x2e, 
+    0x8e, 0xbe, 0xae, 0xa6, 0x16, 0x8b, 0xe9, 0x61, 0x8a, 0xe8, 
+    0x93, 0xaf, 0x0f, 0x08, 0x9b, 0xbf, 0xa4, 0xec, 0xae, 0x62, 
+    0x3d, 0xe4, 0xab, 0x70, 0x3a, 0x2b, 0x0d, 0xd6, 0x80, 0xa8, 
+    0x5f, 0xf7, 0x83, 0x3a, 0xfd, 0x41, 0xdf, 0xec, 0xc6, 0x5c, 
+    0x32, 0xd7, 0xce, 0x7d, 0xe2, 0x8d, 0x8d, 0xb0, 0xec, 0xdc, 
+    0xfc, 0x16, 0x4f, 0x58, 0x5e, 0x6b, 0x34, 0x38, 0x9f, 0xf9, 
+    0x64, 0xdd, 0x43, 0x35, 0xdf, 0x01, 0x04, 0xa9, 0xb9, 0xd7, 
+    0x43, 0x89, 0x7d, 0x78, 0x05, 0x65, 0x1a, 0x13, 0x37, 0x0f, 
+    0x22, 0xb7, 0x8e, 0xe8, 0xe2, 0x3a, 0x42, 0xa3, 0x3e, 0x11, 
+    0xae, 0x4f, 0x7e, 0x15, 0xcd, 0x34, 0xc6, 0xd4, 0x53, 0xa7, 
+    0xdb, 0xfe, 0x0c, 0xe2, 0x82, 0x82, 0x9d, 0x52, 0x8d, 0xb1, 
+    0x5a, 0x61, 0xfb, 0xba, 0x77, 0xc3, 0x4f, 0xcf, 0x20, 0x12, 
+    0x99, 0x56, 0x67, 0x0f, 0x56, 0x8c, 0x0c, 0xbc, 0x0c, 0x25, 
+    0x2a, 0xbe, 0x77, 0x22, 0xc6, 0xb4, 0x50, 0xb7, 0xe2, 0x10, 
+    0xca, 0x2c, 0xff, 0x97, 0xe6, 0xc8, 0x58, 0x75, 0x07, 0x5c, 
+    0x55, 0x77, 0xd2, 0xdb, 0x30, 0x2a, 0x5e, 0x18, 0x4f, 0x5b, 
+    0x4e, 0x1f, 0x2d, 0x6d, 0x1e, 0x3c, 0x8e, 0x9a, 0x73, 0x4e, 
+    0x57, 0x85, 0x59, 0x86, 0x84, 0x30, 0x08, 0x8e, 0xf9, 0x97, 
+    0xca, 0x91, 0x58, 0x6d, 0x93, 0x0f, 0x75, 0x77, 0xd2, 0xdf, 
+    0x30, 0x2a, 0x1f, 0x18, 0x4f, 0x4b, 0x4e, 0x1f, 0x2d, 0x6d, 
+    0x1c, 0x3c, 0x8e, 0x9a, 0x73, 0x4e, 0x57, 0x85, 0x59, 0x86, 
+    0x44, 0x10, 0x0a, 0x8e, 0xfb, 0x97, 0xce, 0x91, 0x58, 0x65, 
+    0x13, 0x0f, 0x8a, 0x88, 0x2d, 0x20, 0xcf, 0xd5, 0xe0, 0xe7, 
+    0xb0, 0xb4, 0xb1, 0xe0, 0xd2, 0x92, 0xe3, 0xc3, 0x71, 0x65, 
+    0x8c, 0xb1, 0xa8, 0x7a, 0xa6, 0x79, 0xbb, 0xef, 0xf5, 0x71, 
+    0x04, 0x68, 0x31, 0x6e, 0xa7, 0x9a, 0xec, 0xf0, 0x04, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 
+    0x04, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x03, 0x00, 
+    0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x07, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x02, 0x00, 
+    0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 
+    0x04, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x03, 0x00, 
+    0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 
+    0x06, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x04, 0x00, 
+    0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 
+    0x02, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x02, 0x00, 
+    0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x04, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x07, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 
+    0x04, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x02, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 
+    0x02, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x07, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x04, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x02, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x02, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x03, 0x00, 
+    0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 
+    0x07, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x06, 0x49, 0x9e, 0x07, 
+    0xe9, 0x01, 0x2d, 0x8a, 0xb5, 0xeb, 0x9d, 0x21, 0x31, 0xbb, 
+    0x4c, 0x26, 0xee, 0xda, 0xae, 0x11, 0x8f, 0x95, 0x1a, 0xa9, 
+    0x74, 0xae, 0xda, 0x53, 0x58, 0xd8, 0x92, 0x81, 0x55, 0xaf, 
+    0x14, 0x79, 0x72, 0x01, 0xff, 0xb4, 0xac, 0xc4, 0x33, 0xa8, 
+    0xc0, 0xea, 0xee, 0xe4, 0x8f, 0xe7, 0xfc, 0xbf, 0x3a, 0xd5, 
+    0x17, 0x01, 0x0d, 0xb3, 0x21, 0x52, 0x58, 0x58, 0x53, 0x44, 
+    0x71, 0x0b, 0xf0, 0x6d, 0x90, 0xb6, 0x1d, 0x86, 0x3c, 0x9d, 
+    0xa1, 0xa3, 0xa6, 0xcd, 0xfc, 0x03, 0xd3, 0x50, 0x63, 0xf0, 
+    0xd0, 0x0c, 0x31, 0x6e, 0x9d, 0x8c, 0xdc, 0x4d, 0xde, 0x0d, 
+    0x0c, 0xb2, 0x30, 0xdf, 0xe2, 0x05, 0x56, 0x09, 0x3c, 0x37, 
+    0x81, 0x81, 0x7c, 0x37, 0x94, 0x6f, 0x2e, 0x96, 0x18, 0xa6, 
+    0xe7, 0x7e, 0x50, 0xc8, 0x49, 0x51, 0x6d, 0x25, 0x1f, 0x47, 
+    0xae, 0x18, 0xf8, 0xc1, 0x9f, 0x12, 0xdb, 0x04, 0xe9, 0x45, 
+    0xde, 0x93, 0x20, 0x63, 0xa9, 0xaa, 0x22, 0xf8, 0x42, 0x0d, 
+    0x84, 0x73, 0x47, 0xc5, 0xbc, 0x83, 0xf9, 0x53, 0x05, 0xdb, 
+    0x7c, 0xbe, 0x9c, 0xb0, 0xd5, 0x4a, 0x1c, 0xcd, 0xba, 0xa8, 
+    0xb2, 0xf6, 0x0a, 0x36, 0x70, 0x17, 0x1b, 0x9d, 0x4e, 0x8c, 
+    0xb0, 0xae, 0xdf, 0x75, 0x35, 0xdb, 0xf4, 0x93, 0x7f, 0x9f, 
+    0x3a, 0xaa, 0xa0, 0x59, 0x61, 0xdf, 0x6b, 0x23, 0xff, 0x36, 
+    0x4d, 0x2a, 0xbc, 0x80, 0x9a, 0x08, 0x8e, 0x67, 0x68, 0x72, 
+    0x3c, 0x14, 0x1b, 0xbd, 0x0c, 0x08, 0xd1, 0xcb, 0xfa, 0xf8, 
+    0x64, 0x9f, 0xf4, 0x93, 0x6f, 0x9f, 0x3a, 0xaa, 0xa0, 0x59, 
+    0x61, 0xdf, 0x69, 0x23, 0xbe, 0x36, 0x5d, 0x4a, 0xbc, 0x80, 
+    0xba, 0x48, 0x8e, 0x76, 0x68, 0x72, 0x3c, 0x17, 0x1b, 0xbd, 
+    0x0c, 0x88, 0xf1, 0xaa, 0xfa, 0xf8, 0x64, 0x9f, 0x0b, 0x6c, 
+    0x90, 0x60, 0xc5, 0x55, 0x5f, 0xa6, 0x9e, 0x20, 0x96, 0xdc, 
+    0x41, 0xc9, 0xa2, 0xb5, 0x43, 0x7f, 0x45, 0xb7, 0x71, 0x89, 
+    0x97, 0x8d, 0xc3, 0xe8, 0xe4, 0x42, 0xf3, 0x77, 0x0e, 0x55, 
+    0x05, 0x07, 0x9b, 0x60, 0x3b, 0xc3, 0x20, 0x74, 0xc0, 0x66, 
+    0xcf, 0xb0, 0x7c, 0xde, 0x76, 0x5a, 0xef, 0xaa, 0x0d, 0x0f, 
+    0xe7, 0xb8, 0x08, 0x58, 0xc1, 0x83, 0x67, 0x7e, 0x5b, 0x56, 
+    0x69, 0x45, 0x8b, 0xa6, 0x5f, 0xa5, 0x0c, 0xd4, 0x59, 0x3b, 
+    0xe7, 0xd5, 0x8a, 0xae, 0xd8, 0xe5, 0x09, 0xfc, 0xbf, 0xe6, 
+    0x48, 0x83, 0x51, 0x7e, 0xbb, 0x65, 0x2d, 0x3a, 0xc3, 0xfb, 
+    0xee, 0x3c, 0x9b, 0x34, 0x52, 0xed, 0x53, 0x3c, 0xeb, 0x3c, 
+    0x7f, 0x12, 0xc9, 0xf7, 0xb3, 0x6c, 0x44, 0x19, 0x41, 0x57, 
+    0xa8, 0x34, 0x45, 0x50, 0x01, 0x73, 0x85, 0x1c, 0x34, 0x7f, 
+    0x05, 0xf5, 0x0d, 0x13, 0x6f, 0xee, 0x7c, 0xe4, 0x7e, 0xbd, 
+    0xae, 0xf4, 0xc5, 0x12, 0xa4, 0xa2, 0xc6, 0x82, 0x3f, 0x4f, 
+    0x9c, 0xe4, 0xd8, 0xfa, 0x83, 0xa5, 0xaa, 0xf3, 0x56, 0x6e, 
+    0xe1, 0x36, 0x4e, 0x44, 0x8e, 0x94, 0xbf, 0x39, 0xb2, 0xfb, 
+    0x43, 0xc4, 0x28, 0xa1, 0x79, 0xad, 0x14, 0x94, 0xc9, 0xd9, 
+    0x32, 0x2d, 0x67, 0xc1, 0x39, 0x8e, 0x1e, 0xea, 0x0a, 0xf6, 
+    0x52, 0x72, 0x50, 0x2e, 0xea, 0xc5, 0x26, 0x8a, 0xd0, 0x1f, 
+    0x58, 0xb6, 0xb0, 0x63, 0x9f, 0x80, 0xf1, 0x75, 0xfc, 0xb0, 
+    0x0b, 0xf3, 0xfb, 0xf4, 0x3c, 0xc1, 0xa7, 0xf2, 0x97, 0xe8, 
+    0x7d, 0x73, 0x2e, 0xdd, 0x4f, 0x9e, 0xd3, 0x6c, 0x7a, 0x2c, 
+    0xe2, 0xc4, 0xb4, 0x72, 0xe6, 0x08, 0xb8, 0x3c, 0xfa, 0x10, 
+    0xbf, 0xd1, 0x77, 0x78, 0xfe, 0xa2, 0x9b, 0xfd, 0x3c, 0xf5, 
+    0xcd, 0x40, 0xa6, 0xe3, 0x16, 0xc9, 0x34, 0xb6, 0x17, 0xcb, 
+    0x4f, 0xde, 0xd2, 0x6c, 0x7a, 0x2c, 0xe2, 0xc4, 0xb4, 0x12, 
+    0xe6, 0x0b, 0xb8, 0x3c, 0xfa, 0x70, 0x9f, 0xd1, 0x75, 0x70, 
+    0xfe, 0xa2, 0x1b, 0xfd, 0x3c, 0xf4, 0xcd, 0x40, 0xa6, 0xe3, 
+    0x16, 0xc9, 0x3c, 0xb6, 0x1f, 0xcb, 0xb0, 0x21, 0x2d, 0x93, 
+    0x85, 0xd3, 0x1d, 0x3b, 0x4b, 0xed, 0x19, 0xf4, 0x47, 0xc3, 
+    0x05, 0x8f, 0x60, 0x2e, 0x8a, 0x8f, 0x01, 0x5d, 0xe4, 0x02, 
+    0xc3, 0x0b, 0x32, 0xbf, 0x59, 0x1c, 0xe9, 0x36, 0xc3, 0x49, 
+    0xe0, 0x34, 0x02, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 
+    0x05, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x04, 0x00, 
+    0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x04, 0x00, 
+    0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 
+    0x07, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x05, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x02, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 
+    0x03, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x06, 0x00, 
+    0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x06, 0x00, 
+    0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x07, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x05, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x07, 0x00, 
+    0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 
+    0x04, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0xe8, 0x14, 0x48, 0x7f, 0x8b, 0xeb, 0x9f, 0x7c, 0xca, 0xb0, 
+    0x52, 0x6e, 0x5f, 0x99, 0x14, 0xd0, 0xb5, 0x63, 0x02, 0x98, 
+    0x38, 0x27, 0xae, 0x2e, 0xe6, 0x53, 0xc4, 0x0b, 0x29, 0x25, 
+    0x51, 0xc9, 0x66, 0x59, 0x26, 0x57, 0x90, 0xd1, 0x8e, 0x40, 
+    0xf1, 0x46, 0x29, 0x9f, 0xce, 0xd6, 0xa1, 0xb5, 0x35, 0x3e, 
+    0xd6, 0x11, 0x29, 0x73, 0xba, 0xeb, 0x85, 0x88, 0x07, 0x27, 
+    0x45, 0x47, 0xdd, 0xac, 0xcc, 0xe8, 0x0b, 0x7e, 0x94, 0x62, 
+    0xbd, 0x9a, 0x4f, 0x9c, 0xe8, 0x8f, 0xa1, 0x8a, 0x0d, 0x8c, 
+    0xbb, 0x30, 0x7a, 0x90, 0xcb, 0x26, 0x0f, 0xd8, 0x5b, 0x4f, 
+    0x2b, 0x3a, 0xb3, 0xee, 0x82, 0x4c, 0xe8, 0xd1, 0x04, 0xcd, 
+    0xa4, 0x34, 0xbd, 0x08, 0xd9, 0x4e, 0x8b, 0xf2, 0x3b, 0x17, 
+    0xb9, 0x7d, 0x9a, 0xc5, 0xa8, 0x8c, 0xd3, 0xf7, 0x47, 0x26, 
+    0xa0, 0xb2, 0x1b, 0xb9, 0x6b, 0x33, 0xbe, 0x76, 0x5a, 0x6d, 
+    0xdb, 0x31, 0x29, 0x79, 0xd5, 0x8c, 0x9b, 0x5c, 0xd1, 0xce, 
+    0xc0, 0x60, 0xee, 0x26, 0x8a, 0xd1, 0x76, 0x1e, 0x0e, 0xcf, 
+    0xa9, 0xe0, 0xc7, 0xf8, 0x36, 0x5f, 0x58, 0xaa, 0xd4, 0xcd, 
+    0x4c, 0x21, 0x60, 0xaf, 0xaf, 0x69, 0x16, 0x92, 0x3a, 0xf3, 
+    0x55, 0xf5, 0x5f, 0x39, 0xec, 0x41, 0x1f, 0x0b, 0xf4, 0xc3, 
+    0x4c, 0x17, 0x60, 0x1c, 0xb6, 0xcf, 0x09, 0x9e, 0x56, 0x9c, 
+    0xba, 0x86, 0xfa, 0xab, 0xd4, 0x9e, 0x5e, 0x9a, 0xa1, 0x97, 
+    0xaf, 0xa3, 0x24, 0xb5, 0x1a, 0xfb, 0x65, 0x9d, 0xe9, 0x2f, 
+    0xb8, 0x08, 0x8d, 0x08, 0x45, 0x43, 0x4c, 0x17, 0x60, 0x1c, 
+    0xb6, 0xcf, 0x89, 0x8e, 0x77, 0xd8, 0xba, 0x86, 0xfa, 0xaa, 
+    0xd6, 0x9e, 0x5e, 0x3a, 0xa0, 0x97, 0xaf, 0xa3, 0x14, 0xb5, 
+    0x1a, 0xfb, 0x45, 0x9d, 0xf9, 0x2f, 0xb8, 0x08, 0x8f, 0x0a, 
+    0xe5, 0x43, 0xb3, 0xe8, 0x9f, 0xe3, 0x49, 0x30, 0x76, 0x71, 
+    0x88, 0x27, 0x45, 0x79, 0x05, 0x55, 0x29, 0x61, 0xa1, 0xc5, 
+    0x5f, 0x68, 0x50, 0x5c, 0xeb, 0x4a, 0xe5, 0x04, 0xba, 0x62, 
+    0x06, 0xd0, 0x47, 0xf7, 0x70, 0xf5, 0x1a, 0xbc, 0xa2, 0x30, 
+    0x75, 0x19, 0x16, 0x9a, 0x6f, 0x93, 0x72, 0xeb, 0x7a, 0x9b, 
+    0x3c, 0xa4, 0xd0, 0xb3, 0x5f, 0x0f, 0x74, 0xa4, 0x89, 0xab, 
+    0xe4, 0x58, 0x80, 0x28, 0x1c, 0xb6, 0x1e, 0x71, 0xd8, 0x1f, 
+    0xac, 0xe1, 0x14, 0x0b, 0x7d, 0xce, 0xdd, 0x0d, 0xf3, 0x9e, 
+    0xb7, 0x29, 0x53, 0xc4, 0x92, 0x28, 0xfc, 0xb8, 0x43, 0x55, 
+    0x31, 0xee, 0x42, 0x18, 0xb9, 0x6d, 0xb2, 0xe0, 0x74, 0x7a, 
+    0x86, 0x52, 0xf7, 0x71, 0x2c, 0x21, 0x6f, 0x4e, 0xad, 0x80, 
+    0xf4, 0xa9, 0xc8, 0xd7, 0xbf, 0x8b, 0x95, 0x27, 0xfd, 0x2e, 
+    0x88, 0x84, 0xeb, 0x0b, 0x45, 0xbb, 0xf0, 0x0c, 0x16, 0x00, 
+    0x89, 0x96, 0xa5, 0x39, 0xee, 0x7c, 0xbe, 0x7b, 0xbf, 0xec, 
+    0xfa, 0xe4, 0xa4, 0xde, 0x0d, 0x00, 0xf4, 0xa1, 0xc6, 0xb4, 
+    0x38, 0xc5, 0xa8, 0xd5, 0x26, 0x5c, 0xc5, 0x89, 0x78, 0x0c, 
+    0x64, 0xdb, 0x9b, 0x3b, 0x7f, 0x9d, 0x30, 0x9f, 0xe0, 0xa9, 
+    0x8a, 0xdb, 0xa3, 0x66, 0x5f, 0x47, 0x6e, 0x25, 0x6e, 0x07, 
+    0x1d, 0x3f, 0x0e, 0xae, 0x85, 0x10, 0x2c, 0x79, 0x84, 0xa7, 
+    0xe7, 0xc0, 0xf4, 0x04, 0xc0, 0xae, 0xc1, 0xe0, 0x5b, 0x20, 
+    0x64, 0x30, 0xe2, 0x3c, 0x85, 0x8b, 0xeb, 0x0f, 0x9d, 0xcc, 
+    0x60, 0xe8, 0xc0, 0x7e, 0xd0, 0x82, 0x2b, 0xd3, 0x8c, 0xa7, 
+    0xc4, 0x5d, 0x2e, 0xd1, 0xa4, 0xeb, 0x83, 0xa0, 0xa4, 0x17, 
+    0x89, 0x9f, 0xdd, 0x61, 0xcb, 0x11, 0x62, 0x31, 0xb2, 0x8d, 
+    0x90, 0x89, 0xb0, 0x8a, 0x88, 0xe8, 0xe3, 0xfb, 0xc5, 0x3e, 
+    0x48, 0x96, 0x2f, 0xd8, 0x8c, 0xa3, 0xc4, 0x5d, 0x2e, 0xd1, 
+    0xa4, 0xa3, 0xc3, 0x80, 0xa4, 0x05, 0xc8, 0x8f, 0xd5, 0x61, 
+    0xcb, 0x11, 0x66, 0x31, 0xb0, 0x8d, 0x90, 0x89, 0xb8, 0x0a, 
+    0x88, 0xc8, 0xe3, 0xfb, 0xc5, 0x3e, 0x48, 0xb6, 0x2f, 0xda, 
+    0x73, 0x5c, 0x3b, 0xa2, 0xd1, 0x2e, 0x5b, 0x5c, 0x3c, 0x7f, 
+    0x5b, 0xfa, 0x37, 0x70, 0x2a, 0x9e, 0x34, 0xee, 0x99, 0xce, 
+    0x4f, 0x72, 0x6f, 0x76, 0x47, 0xf5, 0x77, 0x37, 0x1c, 0x04, 
+    0x3a, 0xc1, 0xb7, 0x69, 0xd0, 0x25, 0x07, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 
+    0x03, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x07, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x04, 0x00, 
+    0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x04, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x07, 0x00, 
+    0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 
+    0x03, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x06, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x06, 0x00, 
+    0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x02, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x02, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x04, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x04, 0x00, 
+    0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x02, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 
+    0x07, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x04, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x02, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x06, 0x00, 
+    0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x07, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 
+    0x07, 0x00, 0x00, 0x00, 0x6d, 0x7e, 0xe7, 0x0c, 0xf1, 0xbe, 
+    0xc7, 0x85, 0x6a, 0xb1, 0x59, 0xe4, 0xba, 0x42, 0x1d, 0xfb, 
+    0x49, 0xf7, 0xb7, 0x21, 0x84, 0x1b, 0x90, 0xd9, 0x66, 0xa9, 
+    0x39, 0x79, 0xab, 0x89, 0xfb, 0x16, 0x68, 0x18, 0xd1, 0x2a, 
+    0x23, 0xda, 0xf3, 0xb2, 0xb5, 0xae, 0x1c, 0xd5, 0xed, 0x91, 
+    0xf7, 0xe3, 0x67, 0x09, 0x0a, 0x11, 0xf0, 0xf1, 0x92, 0x9d, 
+    0xde, 0xc4, 0x9a, 0x32, 0xa4, 0xb5, 0xa2, 0xa6, 0x68, 0x73, 
+    0xd3, 0xe3, 0x12, 0x0c, 0x93, 0x8f, 0x6f, 0x4f, 0x5b, 0xac, 
+    0xd0, 0x65, 0xb7, 0x72, 0xc7, 0x23, 0xb4, 0xa8, 0x10, 0x30, 
+    0x58, 0x32, 0xad, 0x3c, 0x7d, 0x86, 0xd0, 0xbe, 0xc6, 0x1d, 
+    0x3f, 0x04, 0xdb, 0xbc, 0x56, 0x74, 0x20, 0x4f, 0xa4, 0xc6, 
+    0x9e, 0x7a, 0xc5, 0xea, 0x03, 0xa1, 0x21, 0x5e, 0x2e, 0xe9, 
+    0xd3, 0xad, 0xc2, 0xd2, 0xf4, 0xd2, 0xe7, 0x33, 0xe3, 0x1c, 
+    0xb5, 0xbc, 0x26, 0x4b, 0x0c, 0x8f, 0x3d, 0xbd, 0x6c, 0x21, 
+    0xc2, 0x98, 0xc7, 0x5d, 0x65, 0x7a, 0x46, 0x71, 0x7b, 0xe3, 
+    0xa5, 0xe4, 0xd2, 0x2d, 0x10, 0xe7, 0x29, 0x36, 0x44, 0x2e, 
+    0x56, 0x79, 0x52, 0xad, 0xb4, 0x60, 0x7f, 0xb6, 0x5a, 0x48, 
+    0x03, 0x09, 0x98, 0x74, 0x73, 0x7c, 0xab, 0xc1, 0x09, 0x21, 
+    0x21, 0x7c, 0x48, 0x6c, 0x81, 0x42, 0x89, 0x94, 0xf4, 0x7d, 
+    0x1e, 0xdf, 0x8b, 0xd3, 0x80, 0xbb, 0x79, 0x96, 0xd9, 0x15, 
+    0x38, 0x41, 0x2f, 0x21, 0xc8, 0x09, 0x04, 0x88, 0xb8, 0x24, 
+    0x20, 0x78, 0x92, 0x65, 0x7d, 0x75, 0x1f, 0xfb, 0xfa, 0x6d, 
+    0x81, 0xe2, 0xa9, 0xd4, 0xf4, 0x7d, 0x1e, 0xdf, 0x8b, 0xf7, 
+    0x84, 0xbb, 0x79, 0xd6, 0xdb, 0x15, 0xa8, 0x60, 0x2f, 0x30, 
+    0x5a, 0x09, 0x0c, 0x88, 0xb8, 0x24, 0x28, 0x78, 0x9a, 0x65, 
+    0x6d, 0x75, 0x0d, 0xff, 0xfa, 0x6d, 0x7e, 0x1d, 0x56, 0x2b, 
+    0x0b, 0x82, 0xe1, 0x20, 0x74, 0x08, 0x7b, 0x44, 0x86, 0x29, 
+    0x24, 0xea, 0x57, 0x9f, 0xd0, 0xcf, 0xa5, 0xf6, 0xf3, 0x77, 
+    0x47, 0xdb, 0xd7, 0x87, 0x65, 0x9a, 0x92, 0x8a, 0xf2, 0x00, 
+    0x05, 0x92, 0x19, 0x2a, 0x53, 0x46, 0x71, 0x4d, 0x03, 0xf2, 
+    0x9c, 0x90, 0x95, 0x6a, 0x26, 0x2b, 0x66, 0xae, 0x2d, 0x89, 
+    0x0e, 0x1e, 0x95, 0x73, 0x1c, 0xbe, 0x89, 0xa0, 0x3d, 0x24, 
+    0x04, 0x0c, 0x06, 0xc0, 0x6e, 0x21, 0x92, 0x1c, 0xd4, 0x81, 
+    0x43, 0xd5, 0xbe, 0xdb, 0xda, 0x11, 0xf5, 0x7d, 0xaf, 0x66, 
+    0xe2, 0xd7, 0x14, 0xc6, 0x1a, 0x47, 0x1c, 0xe0, 0x12, 0x4b, 
+    0xb9, 0x3b, 0xe6, 0xe5, 0x6b, 0x3b, 0x63, 0x11, 0x9a, 0x48, 
+    0xd3, 0xe7, 0xd4, 0xb5, 0xd9, 0xdf, 0x5c, 0xb3, 0x8e, 0x17, 
+    0x04, 0xbc, 0x54, 0x55, 0x02, 0xb8, 0x28, 0x6f, 0x54, 0x4b, 
+    0x9b, 0x67, 0x6a, 0x98, 0xe7, 0xb3, 0x1c, 0x2f, 0xf5, 0x5a, 
+    0x10, 0x18, 0x48, 0x77, 0x67, 0x9f, 0xbe, 0x88, 0xdc, 0xa3, 
+    0xb4, 0xcb, 0x64, 0x20, 0x98, 0xfb, 0x64, 0x78, 0x11, 0xaa, 
+    0x31, 0x7a, 0x3c, 0x2b, 0x25, 0x9a, 0x15, 0xc6, 0xc8, 0x69, 
+    0x5e, 0xc3, 0x9b, 0x4c, 0xe4, 0xec, 0x58, 0xfa, 0x08, 0x71, 
+    0xa1, 0x79, 0x8f, 0xeb, 0xd8, 0x41, 0xf2, 0x1f, 0xe0, 0x32, 
+    0x0a, 0x08, 0x1c, 0x90, 0x7f, 0xf7, 0x6c, 0xf5, 0x31, 0x4d, 
+    0x55, 0x5c, 0xc0, 0x9c, 0xdb, 0x94, 0x14, 0xac, 0x32, 0xd2, 
+    0xbf, 0xf8, 0xb4, 0x29, 0x4a, 0xec, 0x1e, 0xc4, 0x3a, 0xfc, 
+    0x34, 0x74, 0x36, 0x0e, 0xe0, 0xb1, 0x3a, 0x09, 0x82, 0xfa, 
+    0x5d, 0x6d, 0x33, 0x99, 0x65, 0x0f, 0x79, 0x0e, 0xf2, 0x27, 
+    0x03, 0xa7, 0x14, 0xa0, 0x18, 0xe3, 0x8d, 0xe4, 0xb0, 0x2b, 
+    0x68, 0x99, 0x09, 0xa6, 0xdb, 0xd2, 0x12, 0x3a, 0xb6, 0x0f, 
+    0xe0, 0xb1, 0x1a, 0x09, 0x80, 0xfa, 0x5d, 0xed, 0x31, 0x99, 
+    0x65, 0x4f, 0x7d, 0x0e, 0xf1, 0x27, 0x03, 0xb5, 0x14, 0xa0, 
+    0x18, 0xe3, 0x8d, 0xe4, 0xb0, 0x2b, 0x68, 0x99, 0x0b, 0xa4, 
+    0xdb, 0xd8, 0x12, 0x3a, 0x49, 0xf0, 0x1f, 0x4e, 0xe5, 0xf6, 
+    0x7f, 0x05, 0xa2, 0x12, 0xce, 0x66, 0x9a, 0xb0, 0x82, 0xf1, 
+    0x0e, 0xd8, 0xfc, 0x4a, 0xeb, 0x5f, 0xe7, 0x1c, 0x72, 0x1b, 
+    0x4f, 0xd4, 0x97, 0x66, 0xf4, 0x5b, 0x24, 0x27, 0xed, 0xc5, 
+    0x00, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x07, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x07, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 
+    0x05, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 
+    0x02, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x07, 0x00, 
+    0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x05, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x06, 0x00, 
+    0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x04, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 
+    0x05, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 
+    0x03, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x04, 0x00, 
+    0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 
+    0x03, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x04, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 
+    0x02, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0xd2, 0xc2, 
+    0x63, 0x6b, 0x40, 0x0d, 0x57, 0xed, 0x9b, 0xb6, 0x8f, 0x44, 
+    0xcb, 0x86, 0x60, 0x0a, 0xc3, 0xcd, 0xc2, 0x66, 0x23, 0xbf, 
+    0xa1, 0x17, 0x66, 0x3e, 0xb9, 0xb6, 0x91, 0xc0, 0x53, 0x8a, 
+    0xf6, 0x70, 0x08, 0x76, 0xb0, 0x07, 0x46, 0xe9, 0xec, 0x5d, 
+    0x6e, 0x02, 0x4e, 0x89, 0x69, 0x25, 0x9c, 0x53, 0xd7, 0x41, 
+    0x04, 0x51, 0x55, 0xfa, 0x4b, 0x8e, 0x37, 0x44, 0xbf, 0xe5, 
+    0x14, 0x86, 0xfd, 0x64, 0x5b, 0x61, 0x5c, 0x65, 0x06, 0x48, 
+    0x3d, 0x26, 0x86, 0x06, 0x34, 0x34, 0x61, 0x70, 0x7b, 0x09, 
+    0x1b, 0xb1, 0xf2, 0xaa, 0xd8, 0x0d, 0xa1, 0x23, 0x73, 0xdb, 
+    0x28, 0x32, 0xac, 0x3d, 0x81, 0x5e, 0x1f, 0x9d, 0xa0, 0x98, 
+    0xb3, 0xf9, 0xf3, 0xba, 0xbd, 0xff, 0x74, 0x50, 0x3a, 0x55, 
+    0x6d, 0xae, 0x0e, 0xbc, 0xc8, 0x92, 0x1f, 0x02, 0x78, 0xe2, 
+    0xd3, 0xbe, 0xcf, 0xc7, 0xa2, 0x19, 0x11, 0x2c, 0x47, 0x97, 
+    0x28, 0xc8, 0x38, 0xc6, 0xd8, 0x1e, 0xb2, 0x7e, 0xdf, 0x70, 
+    0xde, 0x20, 0x62, 0x29, 0x96, 0x83, 0x8c, 0x32, 0x7f, 0x45, 
+    0x18, 0x6f, 0xff, 0xae, 0x49, 0xb8, 0x2d, 0x0d, 0xe1, 0xd3, 
+    0xaf, 0x66, 0x7e, 0x79, 0x59, 0x0e, 0xa3, 0xfb, 0x07, 0x13, 
+    0x6d, 0x8c, 0xa0, 0x6f, 0x7a, 0x4c, 0x81, 0x4d, 0x24, 0x41, 
+    0xff, 0x02, 0x49, 0x1e, 0x6d, 0x4e, 0x8b, 0x17, 0xf6, 0xac, 
+    0x40, 0x8b, 0xfd, 0x2d, 0x04, 0xf0, 0x07, 0x49, 0x01, 0x78, 
+    0x53, 0x09, 0x09, 0xeb, 0x1d, 0x13, 0xa9, 0xc9, 0xa0, 0x3c, 
+    0x56, 0xfc, 0x25, 0x4f, 0x24, 0x41, 0xbe, 0x03, 0x4d, 0x1e, 
+    0x6d, 0x4c, 0x89, 0x17, 0xf6, 0xad, 0x40, 0x8b, 0xfd, 0x2d, 
+    0x05, 0xf0, 0x27, 0x69, 0x01, 0x78, 0x53, 0x09, 0x09, 0xeb, 
+    0x1d, 0x13, 0xe9, 0xcc, 0xa0, 0x3c, 0x5e, 0xfc, 0xa1, 0x4f, 
+    0xdb, 0xbe, 0x41, 0xfc, 0xb2, 0xe1, 0x92, 0xb3, 0x76, 0xe8, 
+    0x09, 0x52, 0xbf, 0x74, 0x02, 0xd2, 0xfa, 0x0f, 0xd8, 0x96, 
+    0xfe, 0x87, 0xac, 0xf6, 0xf6, 0x14, 0xe2, 0xec, 0x16, 0x33, 
+    0x5f, 0xc3, 0xa1, 0x03, 0x5e, 0xb0, 0x92, 0xd5, 0x27, 0x71, 
+    0x63, 0x8b, 0x06, 0xbe, 0x52, 0xdb, 0x4b, 0x28, 0xb4, 0x83, 
+    0xee, 0x62, 0xd4, 0xdc, 0x85, 0xfc, 0x1f, 0xcb, 0xb5, 0x3a, 
+    0x39, 0x42, 0x2e, 0xa4, 0x6a, 0x78, 0xe5, 0x10, 0xe6, 0x6b, 
+    0xe2, 0x7f, 0x03, 0xcf, 0xbd, 0x9f, 0x0c, 0x9a, 0x6e, 0x07, 
+    0x4b, 0x42, 0xb8, 0x5f, 0x12, 0x6e, 0x90, 0x2e, 0xf4, 0xce, 
+    0x22, 0xbe, 0x18, 0x05, 0x85, 0x0d, 0xb4, 0xe7, 0x53, 0xbb, 
+    0x9b, 0x4f, 0x52, 0x29, 0xbf, 0x72, 0x14, 0x24, 0x28, 0x36, 
+    0x20, 0xe6, 0x9f, 0x5a, 0xb2, 0xa4, 0xdf, 0x4a, 0xf7, 0xf9, 
+    0xb8, 0xd7, 0x0c, 0x42, 0x99, 0xad, 0x58, 0xae, 0x70, 0x06, 
+    0x97, 0xde, 0x39, 0x3d, 0x20, 0xca, 0xbc, 0x36, 0xd4, 0x74, 
+    0x81, 0xa6, 0x1c, 0x47, 0x85, 0x65, 0xde, 0xb5, 0xe1, 0x92, 
+    0x8c, 0x4f, 0x67, 0x39, 0x61, 0x83, 0xff, 0x36, 0x25, 0xee, 
+    0x15, 0x45, 0x78, 0x52, 0xca, 0xdc, 0x9a, 0xee, 0x10, 0xda, 
+    0xd9, 0x09, 0x1a, 0xc6, 0x7a, 0xfb, 0x57, 0xc1, 0x95, 0x5c, 
+    0x1b, 0x35, 0xc6, 0x95, 0xd3, 0x37, 0xed, 0xe6, 0xd9, 0xf9, 
+    0x4f, 0x9a, 0x7e, 0x9d, 0x0f, 0x2c, 0x6b, 0xde, 0x8c, 0x39, 
+    0x6d, 0x7e, 0x74, 0xe1, 0x4b, 0x7b, 0xaf, 0xbd, 0x08, 0xa9, 
+    0x9d, 0x1b, 0x85, 0x60, 0x19, 0xac, 0x54, 0x55, 0x43, 0x1d, 
+    0x53, 0x76, 0xf2, 0xae, 0xbd, 0x59, 0x66, 0x92, 0x99, 0x66, 
+    0x52, 0xe8, 0x55, 0x9f, 0x2d, 0x2c, 0x0b, 0xbf, 0xa4, 0x60, 
+    0x56, 0xf9, 0x6f, 0x6f, 0x40, 0xa4, 0xb5, 0x5b, 0xb3, 0xc1, 
+    0x23, 0x85, 0x54, 0x51, 0x42, 0x15, 0x53, 0x36, 0xfe, 0xae, 
+    0xbf, 0x59, 0x6f, 0x92, 0x9f, 0x74, 0x46, 0xa8, 0x55, 0x9f, 
+    0x2d, 0x2c, 0x0b, 0x7f, 0xb4, 0xe0, 0x12, 0xf9, 0x6f, 0x2f, 
+    0x48, 0xa4, 0xb5, 0x5b, 0xb3, 0xc0, 0x13, 0x85, 0xab, 0xae, 
+    0xbd, 0xea, 0xac, 0xc9, 0x01, 0x51, 0x40, 0xa6, 0x90, 0x6d, 
+    0x60, 0x8b, 0xb9, 0x57, 0xaa, 0x60, 0xd2, 0xd3, 0xf4, 0x80, 
+    0x4b, 0x1f, 0xed, 0x06, 0x90, 0xd0, 0xb7, 0x5b, 0x4a, 0xa4, 
+    0x4c, 0x3f, 0xec, 0x7a, 0x03, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x02, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x07, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x06, 0x00, 
+    0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 
+    0x02, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x04, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x06, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 
+    0x05, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x06, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 
+    0x02, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x04, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 
+    0x02, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x07, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 
+    0x04, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x07, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x05, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 
+    0x05, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x02, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x03, 0x00, 
+    0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x62, 0x0f, 0xc7, 0x45, 0x3e, 0x48, 0xc7, 0x20, 
+    0xaf, 0x96, 0x33, 0x26, 0xd7, 0xfc, 0x52, 0x2f, 0x0d, 0x11, 
+    0xc7, 0x22, 0x77, 0xb8, 0x15, 0x45, 0x6c, 0x66, 0x0e, 0x3b, 
+    0x26, 0x6e, 0xd1, 0x3c, 0x09, 0xa2, 0x73, 0xc4, 0x59, 0x82, 
+    0xe4, 0xb3, 0xae, 0x9f, 0x02, 0xd3, 0x05, 0xb2, 0xef, 0xdd, 
+    0x24, 0x38, 0xc0, 0x21, 0xf4, 0x84, 0xd7, 0x8b, 0xa3, 0xcb, 
+    0x3a, 0x55, 0x86, 0xa8, 0x18, 0xab, 0xc0, 0x22, 0xcb, 0x11, 
+    0x10, 0x5b, 0x61, 0x5b, 0xfa, 0x8b, 0x12, 0x0c, 0x82, 0x01, 
+    0xab, 0x48, 0x78, 0x53, 0x88, 0x71, 0x6a, 0x84, 0x11, 0x29, 
+    0xc5, 0x19, 0x70, 0x48, 0xae, 0x6b, 0x30, 0xcc, 0xfe, 0xd9, 
+    0xe1, 0xf2, 0x02, 0xab, 0x28, 0x15, 0x87, 0x30, 0xb7, 0xba, 
+    0xa7, 0x80, 0xb4, 0x52, 0x08, 0x9f, 0xf6, 0xf1, 0xbc, 0x9a, 
+    0xfc, 0x33, 0xac, 0xa1, 0x35, 0x92, 0xb9, 0x72, 0xb2, 0x5a, 
+    0xab, 0x89, 0x5c, 0xfc, 0x39, 0xac, 0x88, 0xac, 0x87, 0xb6, 
+    0xbf, 0x55, 0x95, 0xa9, 0x01, 0xcf, 0x76, 0xc3, 0x6b, 0x8e, 
+    0xbd, 0x4b, 0x6f, 0x31, 0x57, 0xd9, 0x0a, 0x25, 0x48, 0xfd, 
+    0xe6, 0x88, 0x8b, 0xc1, 0xfd, 0xd2, 0xfd, 0xc4, 0x49, 0x8c, 
+    0xf3, 0x84, 0xfb, 0xfe, 0x46, 0x8f, 0xf6, 0xea, 0x0d, 0xc3, 
+    0x60, 0xd4, 0xb7, 0xc7, 0x6b, 0x4e, 0xc0, 0x03, 0xd2, 0xa3, 
+    0x9d, 0xaa, 0x1a, 0x26, 0x2e, 0xdd, 0xbd, 0x08, 0xab, 0xdc, 
+    0xf5, 0x56, 0xba, 0xc8, 0xda, 0xcc, 0xf3, 0xad, 0xfd, 0xed, 
+    0xa6, 0xfe, 0xf1, 0x38, 0x4f, 0x64, 0x20, 0xd2, 0xb7, 0xc3, 
+    0x6b, 0x4e, 0xd4, 0x03, 0xf2, 0x21, 0x9d, 0x8a, 0x1a, 0x27, 
+    0x2e, 0xdd, 0x2d, 0x08, 0xab, 0xd8, 0xf5, 0x56, 0xba, 0xc8, 
+    0x58, 0xcc, 0xf3, 0xac, 0xfd, 0xee, 0xa6, 0xbe, 0xf1, 0x38, 
+    0x4d, 0xe9, 0x20, 0xd2, 0x48, 0x3c, 0x94, 0xb1, 0x2b, 0xfc, 
+    0x0d, 0xde, 0x62, 0x75, 0xe5, 0xd8, 0xd1, 0x22, 0xd2, 0xf7, 
+    0x54, 0x27, 0x0a, 0xa9, 0x45, 0x37, 0xa7, 0x33, 0x0c, 0x53, 
+    0x02, 0x11, 0x59, 0x41, 0x0e, 0xc7, 0xb2, 0x16, 0xdf, 0x2d, 
+    0x69, 0xe7, 0x0a, 0x57, 0x1c, 0x6d, 0xd2, 0x4f, 0xd9, 0xaf, 
+    0x9e, 0x36, 0xf2, 0x12, 0x55, 0x57, 0x3b, 0x05, 0x04, 0x1c, 
+    0xad, 0xc6, 0x81, 0xd1, 0x14, 0x65, 0x28, 0x4c, 0x1a, 0x9d, 
+    0x7a, 0x13, 0x56, 0x12, 0xd3, 0x30, 0x15, 0x82, 0xcf, 0x21, 
+    0x4f, 0x8e, 0x2c, 0x6e, 0x02, 0x3f, 0x0a, 0x1e, 0xd3, 0x08, 
+    0x47, 0x51, 0x1a, 0x30, 0x88, 0xcd, 0xba, 0xbb, 0x8e, 0xde, 
+    0x2c, 0xbf, 0x02, 0xd6, 0xa8, 0xcc, 0x17, 0xe8, 0xc7, 0xb0, 
+    0xf9, 0x06, 0x8e, 0xe5, 0xcc, 0xfb, 0x7b, 0x02, 0xb7, 0x5a, 
+    0x7a, 0x81, 0x57, 0x40, 0x74, 0x6a, 0xe1, 0x60, 0xf5, 0x94, 
+    0x90, 0x32, 0x2b, 0xba, 0x6f, 0x67, 0x05, 0xa6, 0xe9, 0xc5, 
+    0x12, 0xb1, 0x62, 0xa7, 0xe5, 0xdb, 0xa9, 0xb0, 0x79, 0xf4, 
+    0x93, 0x4b, 0x38, 0xd4, 0x59, 0x72, 0xba, 0x97, 0x67, 0x9c, 
+    0x47, 0x65, 0xe2, 0x06, 0x19, 0x4c, 0x22, 0xfc, 0x66, 0x4e, 
+    0x4e, 0x17, 0xb1, 0x64, 0x3c, 0x99, 0x7d, 0xce, 0xd9, 0x80, 
+    0xb1, 0xde, 0xb0, 0x70, 0xb6, 0x1a, 0x36, 0x91, 0x7c, 0x88, 
+    0xe4, 0x49, 0x18, 0x7d, 0x0f, 0xfc, 0x56, 0xb3, 0x25, 0xb1, 
+    0x12, 0x00, 0xf3, 0x77, 0x7a, 0xe7, 0xe4, 0xd2, 0xd8, 0x99, 
+    0xdb, 0x21, 0x91, 0x0a, 0x0e, 0xef, 0xb2, 0x44, 0x61, 0x6d, 
+    0x26, 0xb2, 0x3a, 0xca, 0x28, 0x50, 0x84, 0x6d, 0x0b, 0xb8, 
+    0x47, 0x58, 0x19, 0x86, 0xf3, 0x65, 0x12, 0x96, 0x55, 0xef, 
+    0xe2, 0xd3, 0x8c, 0xe0, 0xe1, 0x53, 0xbd, 0x00, 0x93, 0xc3, 
+    0x4a, 0x6f, 0xb2, 0xd8, 0x80, 0x6d, 0x26, 0xb2, 0x32, 0xdb, 
+    0x28, 0xd0, 0x84, 0x69, 0x1b, 0x39, 0x47, 0xdc, 0x14, 0xa6, 
+    0xf3, 0x65, 0x12, 0x14, 0x55, 0xff, 0x62, 0xf3, 0x8c, 0xc2, 
+    0xe1, 0x51, 0xbd, 0x00, 0x93, 0x8b, 0x0a, 0x6f, 0xb2, 0xd8, 
+    0x00, 0x6d, 0xd9, 0x4d, 0xcd, 0x24, 0xd7, 0x2f, 0x7b, 0x96, 
+    0xe4, 0xc6, 0xb8, 0x23, 0xeb, 0x59, 0x0c, 0x9a, 0xed, 0xeb, 
+    0xaa, 0x00, 0x9d, 0x0c, 0x73, 0x3d, 0x1e, 0xae, 0x42, 0xff, 
+    0x6c, 0x74, 0xf5, 0x90, 0x4d, 0x27, 0xff, 0x92, 0x04, 0x00, 
+    0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 
+    0x05, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x04, 0x00, 
+    0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x06, 0x00, 
+    0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x06, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x04, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x07, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x06, 0x00, 
+    0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x06, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x04, 0x00, 
+    0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x02, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x07, 0x00, 
+    0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x04, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x04, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x02, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x02, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x07, 0x00, 
+    0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 
+    0x02, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0xff, 0x2a, 0x32, 0x76, 
+    0x13, 0x21, 0x2b, 0x93, 0x80, 0xbd, 0xb3, 0x32, 0x28, 0xa0, 
+    0x66, 0x6d, 0x12, 0x27, 0xce, 0xb5, 0x07, 0x1c, 0x6c, 0x5d, 
+    0x87, 0x15, 0x2c, 0xf5, 0x4d, 0x69, 0xd3, 0x9c, 0x4b, 0xb8, 
+    0x54, 0xa5, 0x1b, 0xcc, 0x35, 0x97, 0x07, 0xb8, 0x7e, 0xb9, 
+    0xca, 0xa6, 0xb8, 0x31, 0x84, 0x7f, 0x18, 0x5c, 0x4a, 0x62, 
+    0xfe, 0x8b, 0xe3, 0x8d, 0x7b, 0x9d, 0x97, 0x62, 0xc3, 0x53, 
+    0xf0, 0x8d, 0x8d, 0x98, 0x37, 0x96, 0xd2, 0xea, 0x5e, 0x2d, 
+    0xe9, 0xb8, 0x10, 0xc2, 0x40, 0x1b, 0xac, 0x2e, 0xb5, 0x88, 
+    0x7c, 0x5a, 0x9a, 0x06, 0xcf, 0x6b, 0x91, 0xd1, 0x9f, 0xc5, 
+    0x3c, 0x62, 0x6c, 0x84, 0xc2, 0x59, 0xe9, 0xad, 0x2d, 0x31, 
+    0xf3, 0x38, 0x36, 0xaf, 0xd4, 0x11, 0x37, 0x09, 0xee, 0x89, 
+    0x0e, 0x57, 0x00, 0xca, 0xb4, 0x98, 0x1d, 0x94, 0xad, 0x8c, 
+    0x62, 0x8b, 0x53, 0x40, 0xf5, 0x27, 0x59, 0x82, 0x58, 0x63, 
+    0xc2, 0x63, 0x56, 0x6f, 0x16, 0x89, 0x63, 0xd6, 0x21, 0xbc, 
+    0xe7, 0xf4, 0xef, 0x86, 0x04, 0xff, 0xb2, 0xcf, 0x97, 0x03, 
+    0xa3, 0xc1, 0xc9, 0xb3, 0x93, 0xeb, 0xc9, 0x49, 0x38, 0x12, 
+    0xb9, 0x73, 0x5a, 0x7c, 0xf9, 0x7a, 0xc8, 0x05, 0x30, 0x64, 
+    0x3d, 0xf8, 0x2f, 0xc2, 0x85, 0xe2, 0xe4, 0xb0, 0x66, 0x8b, 
+    0xb4, 0xda, 0x6f, 0xde, 0x98, 0x63, 0xab, 0x89, 0x9b, 0x3a, 
+    0x16, 0xda, 0x79, 0x0f, 0x59, 0x12, 0x03, 0x57, 0xd1, 0x94, 
+    0x1a, 0x7a, 0xc2, 0x61, 0x20, 0xe5, 0xb1, 0x9e, 0x26, 0x9e, 
+    0x76, 0xe8, 0xe4, 0xb0, 0x67, 0x8b, 0x94, 0x9a, 0x6e, 0xde, 
+    0x98, 0x23, 0xab, 0x81, 0x99, 0xb2, 0x16, 0xda, 0x79, 0x0f, 
+    0x79, 0x12, 0x03, 0x57, 0xd9, 0x94, 0x5a, 0x7a, 0xca, 0x61, 
+    0x30, 0xe5, 0xb1, 0x9c, 0x26, 0x9e, 0x75, 0xe8, 0x1b, 0x4f, 
+    0x98, 0x74, 0x6b, 0x65, 0x91, 0x21, 0x67, 0xdc, 0x54, 0x7e, 
+    0x66, 0x4d, 0xe9, 0x25, 0x86, 0xf0, 0x86, 0xed, 0xfc, 0xa8, 
+    0x26, 0x6b, 0xa5, 0x85, 0x35, 0x9e, 0xcf, 0x1a, 0x4e, 0x63, 
+    0xd9, 0x61, 0x8a, 0x17, 0x86, 0x89, 0x67, 0x5a, 0x21, 0x61, 
+    0xee, 0xf3, 0xd1, 0x93, 0x30, 0x01, 0x34, 0xeb, 0xd1, 0x76, 
+    0x6b, 0x92, 0x6f, 0x67, 0x13, 0xdd, 0x5f, 0x33, 0x15, 0xda, 
+    0x42, 0x4c, 0x5b, 0xae, 0x0f, 0xa7, 0xb2, 0x5e, 0x1c, 0x71, 
+    0xfa, 0x4c, 0xe9, 0x5f, 0x35, 0xcb, 0x64, 0xe4, 0x2b, 0x3d, 
+    0xc8, 0xc1, 0xc5, 0xa1, 0x6a, 0x32, 0xd7, 0x48, 0x9f, 0xe3, 
+    0xd6, 0xba, 0xdc, 0x7d, 0x8d, 0x3f, 0x82, 0x6f, 0xbf, 0x42, 
+    0x57, 0xa1, 0x10, 0xf7, 0x98, 0x3e, 0x37, 0x3c, 0xcb, 0x03, 
+    0x9d, 0x83, 0x3f, 0xe3, 0xae, 0x92, 0x50, 0x49, 0xc6, 0x01, 
+    0x40, 0xd2, 0xbc, 0xb7, 0xb5, 0xce, 0xac, 0x3e, 0x79, 0x2b, 
+    0xe4, 0x3d, 0x16, 0xec, 0x3f, 0x69, 0xa9, 0xe9, 0x3f, 0xd7, 
+    0x9e, 0x79, 0xa7, 0xe7, 0x12, 0x01, 0xae, 0x3b, 0x70, 0x5b, 
+    0x98, 0x13, 0x53, 0xe6, 0x92, 0x69, 0xcc, 0xe3, 0x4c, 0xe4, 
+    0x01, 0x47, 0x75, 0x90, 0x2a, 0x69, 0xef, 0x5a, 0xfa, 0xa4, 
+    0x3d, 0x3b, 0x0c, 0x2f, 0xa9, 0xdf, 0xdc, 0x8f, 0x9a, 0x8f, 
+    0x2c, 0x9b, 0x7a, 0x1d, 0x91, 0xd0, 0x63, 0x1a, 0xfe, 0xce, 
+    0xf9, 0x9b, 0xf7, 0x05, 0xf6, 0xef, 0x82, 0xc3, 0xe1, 0x19, 
+    0x10, 0x16, 0x13, 0xe8, 0x8e, 0x71, 0x88, 0xfc, 0xe6, 0x00, 
+    0x18, 0xe2, 0xdf, 0xec, 0x8e, 0xed, 0x4f, 0x03, 0xff, 0x8f, 
+    0x90, 0xfb, 0x2e, 0x22, 0xc5, 0x04, 0x9d, 0x38, 0xb6, 0x66, 
+    0xb6, 0x37, 0x91, 0xd5, 0x51, 0xf3, 0x38, 0x22, 0xae, 0x60, 
+    0x92, 0xa0, 0xc8, 0x5c, 0x0e, 0x25, 0x3a, 0x52, 0x54, 0x5c, 
+    0x8e, 0xed, 0x0e, 0x03, 0xfe, 0x8f, 0x90, 0xda, 0x2b, 0x32, 
+    0xc5, 0x04, 0x9d, 0x39, 0xb6, 0xe6, 0xb6, 0x37, 0x91, 0xd5, 
+    0x51, 0xf9, 0x38, 0x22, 0xae, 0x60, 0x9a, 0xa1, 0xc8, 0x5c, 
+    0x0e, 0x25, 0x3a, 0x52, 0xd4, 0x6c, 0x71, 0x12, 0xf1, 0xfc, 
+    0x01, 0x70, 0x6f, 0x25, 0xd4, 0xcd, 0x3a, 0xfb, 0x62, 0xc6, 
+    0x49, 0x19, 0x49, 0xc8, 0x6e, 0x2a, 0xae, 0x06, 0xc7, 0xdd, 
+    0x51, 0x9f, 0x65, 0x5e, 0x37, 0xa3, 0xf1, 0xda, 0xc5, 0xad, 
+    0x2b, 0x93, 0x07, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x02, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x04, 0x00, 
+    0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 
+    0x04, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 
+    0x04, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x02, 0x00, 
+    0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 
+    0x05, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x07, 0x00, 
+    0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x02, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 
+    0x03, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x04, 0x00, 
+    0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 
+    0x07, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x03, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x02, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 
+    0x03, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x05, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x04, 0x00, 
+    0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 
+    0x07, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x02, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x04, 0x00, 
+    0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x07, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x07, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 
+    0x74, 0x73, 0xb9, 0x06, 0xb5, 0x76, 0x7e, 0x6b, 0x7f, 0xb0, 
+    0xef, 0x45, 0xc3, 0x06, 0x30, 0x8c, 0x41, 0x38, 0xa6, 0x8a, 
+    0xec, 0xb8, 0x23, 0xef, 0xa0, 0xec, 0x44, 0x5f, 0x61, 0xbc, 
+    0xa9, 0xe5, 0x94, 0x5a, 0xba, 0x8f, 0x1e, 0xc6, 0xaf, 0x7f, 
+    0x39, 0xc0, 0xc2, 0x07, 0xd9, 0x84, 0x3f, 0x77, 0x43, 0xb5, 
+    0xbc, 0xce, 0x03, 0x25, 0x76, 0x4b, 0x98, 0xd0, 0x0e, 0x53, 
+    0x79, 0x78, 0x81, 0x7c, 0xcc, 0x99, 0x7f, 0x1d, 0x5c, 0x13, 
+    0x7a, 0xda, 0x43, 0xd3, 0xda, 0x1a, 0x2f, 0xe8, 0x6b, 0x3d, 
+    0x21, 0xde, 0xf5, 0x92, 0xcc, 0xe1, 0x12, 0xfd, 0xd9, 0x40, 
+    0xd2, 0x25, 0xee, 0x85, 0x06, 0x13, 0xd7, 0x64, 0xfb, 0xda, 
+    0xdd, 0xba, 0xb3, 0xfc, 0x7f, 0x17, 0x2f, 0xe9, 0x26, 0xf2, 
+    0x20, 0x78, 0xcb, 0xfd, 0x6b, 0x01, 0x18, 0x93, 0x79, 0xf4, 
+    0xb1, 0x6b, 0x4f, 0xdd, 0x06, 0x0f, 0x84, 0x55, 0x6e, 0xa4, 
+    0xe9, 0x2e, 0x15, 0xe5, 0x36, 0xf6, 0x99, 0x9d, 0x68, 0x0e, 
+    0xd0, 0xf1, 0xd1, 0xaa, 0x2b, 0x99, 0x89, 0x25, 0x2b, 0xa6, 
+    0xcc, 0xa8, 0x1d, 0x91, 0x3d, 0x5f, 0x2c, 0xae, 0x22, 0xfc, 
+    0x1d, 0x03, 0xd2, 0x34, 0xc3, 0x60, 0xa6, 0x8b, 0x25, 0x8b, 
+    0x21, 0x0b, 0xc1, 0x85, 0x00, 0xf7, 0x7a, 0xaa, 0x60, 0x66, 
+    0x3a, 0x7a, 0x3c, 0x39, 0xeb, 0x66, 0x5f, 0x39, 0x5f, 0x47, 
+    0x1f, 0x5d, 0xfb, 0xbf, 0x34, 0xfc, 0x18, 0x0a, 0xd6, 0x3c, 
+    0xa2, 0xf4, 0x08, 0x1b, 0x5e, 0x48, 0x61, 0x72, 0xa1, 0xb3, 
+    0x08, 0x55, 0x7a, 0xfa, 0x72, 0xf0, 0x3a, 0xfa, 0x30, 0x3d, 
+    0xab, 0x26, 0x5f, 0x39, 0x5f, 0x07, 0x1d, 0x5d, 0xed, 0xbf, 
+    0x34, 0xfc, 0x18, 0x0a, 0xd6, 0x3c, 0xa6, 0xe4, 0x08, 0x0b, 
+    0x54, 0x48, 0x61, 0x72, 0xa1, 0xb3, 0x08, 0x55, 0x7a, 0xfa, 
+    0x72, 0xf0, 0xc5, 0x05, 0xcf, 0xc2, 0x54, 0xd9, 0xa0, 0xc6, 
+    0xa0, 0xf8, 0xe2, 0xa2, 0x12, 0x40, 0xcb, 0x03, 0xe7, 0xf5, 
+    0x29, 0xc3, 0x59, 0x1b, 0xf7, 0xf4, 0xab, 0xb7, 0x9e, 0x8d, 
+    0x5e, 0x4c, 0xf7, 0xaa, 0x85, 0x05, 0x8d, 0x0f, 0x14, 0x0b, 
+    0x10, 0x78, 0x94, 0x13, 0x6f, 0xe0, 0xe2, 0x54, 0xe6, 0x89, 
+    0xdb, 0x06, 0x47, 0xb6, 0x59, 0x03, 0x86, 0xb5, 0xf0, 0x63, 
+    0xc8, 0xc1, 0x57, 0xb8, 0x70, 0xd7, 0x43, 0xda, 0xe3, 0x1c, 
+    0x61, 0xb1, 0xd3, 0xeb, 0xa3, 0x85, 0x5c, 0x88, 0xe0, 0x7e, 
+    0xca, 0x55, 0x6d, 0x13, 0x9f, 0x04, 0xb5, 0xd5, 0xc3, 0x30, 
+    0xb0, 0x5e, 0xe4, 0x3e, 0xaf, 0xbc, 0x50, 0xa2, 0x70, 0xcd, 
+    0x93, 0x3a, 0x19, 0x1e, 0x2b, 0x26, 0xcb, 0x83, 0xcb, 0xb2, 
+    0x74, 0xc6, 0xa9, 0x1b, 0xca, 0x13, 0x1a, 0x6e, 0x9c, 0x46, 
+    0x4c, 0x1d, 0x6a, 0xd3, 0x06, 0x83, 0xb6, 0x19, 0x60, 0xc2, 
+    0x2e, 0xbb, 0x70, 0x4c, 0xd2, 0x4b, 0x34, 0x33, 0xb6, 0xb1, 
+    0x6a, 0x7b, 0xb5, 0x05, 0x68, 0x0d, 0xf1, 0x93, 0xd9, 0x71, 
+    0x53, 0x8d, 0x59, 0x87, 0x33, 0xe1, 0xab, 0xec, 0x2f, 0x7f, 
+    0x04, 0x48, 0x3d, 0xf5, 0xe4, 0x5b, 0x07, 0x84, 0xd3, 0x1e, 
+    0x77, 0xe8, 0x37, 0x18, 0xcb, 0xf8, 0x60, 0x19, 0xc7, 0x01, 
+    0x94, 0xc9, 0xa7, 0x22, 0x59, 0x4b, 0x91, 0x13, 0xcd, 0x88, 
+    0x7d, 0x84, 0xb1, 0x42, 0xa0, 0x21, 0xca, 0x78, 0x52, 0x3c, 
+    0x4c, 0x7f, 0x16, 0x57, 0xcf, 0x91, 0xe5, 0x7b, 0x52, 0x56, 
+    0x31, 0xa0, 0x7e, 0x16, 0x9b, 0xcb, 0xea, 0x57, 0xa5, 0x33, 
+    0x18, 0x28, 0xb5, 0x06, 0x09, 0xd5, 0x7d, 0x90, 0x9c, 0x13, 
+    0xa8, 0x7b, 0x9a, 0x5e, 0x1d, 0x6d, 0xc5, 0xce, 0x07, 0xd2, 
+    0x59, 0xfd, 0x65, 0xf9, 0x7b, 0x01, 0xed, 0x0f, 0x4d, 0x36, 
+    0xd3, 0xce, 0xf0, 0x01, 0xa5, 0x33, 0x58, 0x29, 0x95, 0x06, 
+    0x09, 0xdd, 0x7d, 0x90, 0x9d, 0x13, 0xa8, 0x31, 0x9a, 0x5c, 
+    0x1d, 0x2d, 0xc4, 0xcf, 0x07, 0xf2, 0x59, 0xdd, 0x65, 0xf9, 
+    0x73, 0x11, 0xe9, 0xa3, 0x6f, 0x16, 0xd3, 0xcb, 0xf0, 0x03, 
+    0x5a, 0xcc, 0xa7, 0xd6, 0x6a, 0xf9, 0xf6, 0x22, 0x82, 0x6f, 
+    0x62, 0xec, 0x57, 0xce, 0x65, 0xa3, 0xe2, 0xd2, 0x3b, 0x30, 
+    0xf8, 0x0d, 0xa6, 0x22, 0x9a, 0x06, 0x8c, 0xee, 0x16, 0x5c, 
+    0x90, 0xe9, 0x2c, 0x34, 0x0f, 0xfc, 0x01, 0x00, 0x00, 0x00, 
+    0x03, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 
+    0x04, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x07, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x07, 0x00, 
+    0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 
+    0x02, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x07, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x04, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x02, 0x00, 
+    0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x03, 0x00, 
+    0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 
+    0x05, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 
+    0x06, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x02, 0x00, 
+    0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x07, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x04, 0x00, 
+    0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x02, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 
+    0x06, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x04, 0x00, 
+    0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 
+    0x03, 0x00, 0x00, 0x00, 0xfd, 0xac, 0x75, 0x65, 0xb1, 0x2c, 
+    0x0f, 0xdc, 0x3e, 0x6e, 0xce, 0xd6, 0x91, 0x02, 0x81, 0x8a, 
+    0x5f, 0xb3, 0xb4, 0xae, 0x73, 0x6d, 0xe5, 0x3e, 0xf5, 0xb5, 
+    0xe2, 0xa1, 0x35, 0x8f, 0x0f, 0xdf, 0x10, 0x68, 0x8d, 0xc3, 
+    0x48, 0x7d, 0xae, 0x89, 0x4d, 0xc2, 0x61, 0x2c, 0xfe, 0x69, 
+    0x4d, 0x59, 0x3e, 0x62, 0xe5, 0x51, 0x72, 0x6d, 0xa1, 0xb9, 
+    0xa3, 0xe2, 0xed, 0x80, 0xfd, 0xff, 0xdd, 0xb2, 0x4a, 0x4f, 
+    0xff, 0xb6, 0xab, 0x9e, 0xac, 0x93, 0xf5, 0x8c, 0xf4, 0x69, 
+    0xa8, 0xdb, 0xed, 0x4e, 0x1f, 0x2f, 0xf3, 0x6c, 0x7e, 0xe5, 
+    0xd8, 0xa1, 0x81, 0x1c, 0x1a, 0x4e, 0xe3, 0xcf, 0xe0, 0x39, 
+    0xda, 0x5d, 0xd1, 0x74, 0x65, 0x3b, 0x21, 0x91, 0x7f, 0xd7, 
+    0x2d, 0x19, 0x84, 0x9c, 0x6d, 0x47, 0x80, 0x9a, 0x28, 0xa2, 
+    0x60, 0x7c, 0x32, 0x59, 0xb2, 0x18, 0x4d, 0x29, 0x51, 0x14, 
+    0x2b, 0xab, 0x1f, 0x6c, 0xe2, 0x5e, 0x0b, 0x3d, 0x01, 0x39, 
+    0x2d, 0xb7, 0x5d, 0xd9, 0xf5, 0xcb, 0x6e, 0x64, 0x6d, 0xed, 
+    0x32, 0xdd, 0x81, 0x08, 0x87, 0xcf, 0x5c, 0x49, 0xf3, 0xee, 
+    0xa5, 0x08, 0x13, 0xc6, 0x97, 0x37, 0x22, 0x54, 0x25, 0xc0, 
+    0x15, 0x42, 0x8a, 0x33, 0x3a, 0x0b, 0x3b, 0xda, 0xc6, 0xd1, 
+    0x0d, 0x6d, 0x80, 0x6a, 0x0c, 0xac, 0xaa, 0x2f, 0xd6, 0x90, 
+    0x25, 0xac, 0x5e, 0xf2, 0x73, 0xe2, 0x27, 0x98, 0x01, 0xe2, 
+    0xd7, 0x16, 0x22, 0x30, 0x23, 0xb5, 0x17, 0x42, 0x8c, 0x73, 
+    0x0f, 0xd8, 0x33, 0x86, 0xcc, 0x9d, 0x29, 0xa9, 0xfa, 0x7a, 
+    0x0c, 0xac, 0xaa, 0x2f, 0x96, 0x90, 0x25, 0x8e, 0x5e, 0x71, 
+    0x73, 0xe2, 0x27, 0x98, 0x01, 0xe2, 0xd7, 0x16, 0x22, 0x30, 
+    0x23, 0xb5, 0x17, 0x42, 0x8e, 0x73, 0x0f, 0xc8, 0x33, 0x86, 
+    0xcc, 0xdd, 0x29, 0xa9, 0xea, 0x7a, 0xf3, 0x53, 0x55, 0xd0, 
+    0x69, 0x6f, 0xda, 0x71, 0xa1, 0x8e, 0x8c, 0x1d, 0xd8, 0x67, 
+    0xfe, 0x1d, 0x28, 0xe9, 0xdd, 0xcf, 0xdc, 0x4a, 0xe8, 0xbd, 
+    0x71, 0x8c, 0xf0, 0x37, 0xcc, 0x79, 0x33, 0x22, 0xd6, 0x56, 
+    0x15, 0x85, 0x7b, 0xd1, 0xb0, 0x41, 0xac, 0x2b, 0xf2, 0x96, 
+    0x64, 0xca, 0x03, 0xf7, 0x91, 0xa3, 0x2e, 0x39, 0xe3, 0x29, 
+    0x10, 0x60, 0xa3, 0x96, 0x48, 0x67, 0x3c, 0xd1, 0x45, 0x98, 
+    0x9f, 0x62, 0x01, 0xa9, 0x63, 0xaa, 0x0c, 0xe4, 0x4c, 0x52, 
+    0xbc, 0xe9, 0xfe, 0x6c, 0x59, 0x0f, 0x9a, 0x31, 0x72, 0x3f, 
+    0xd6, 0xba, 0x1b, 0xd8, 0xff, 0x2e, 0xec, 0xeb, 0x82, 0xcb, 
+    0x57, 0xa5, 0x89, 0x7a, 0xdb, 0x8c, 0x18, 0xff, 0x1e, 0xf8, 
+    0x47, 0x05, 0x9f, 0x48, 0x07, 0xe8, 0xac, 0x14, 0xcd, 0xcc, 
+    0xd8, 0xac, 0x21, 0xaf, 0x22, 0x81, 0x2b, 0x81, 0x8a, 0xb3, 
+    0x51, 0x14, 0xc3, 0x2d, 0xc9, 0xa9, 0x5b, 0xd2, 0xf5, 0x74, 
+    0x52, 0xeb, 0x4f, 0x99, 0x18, 0x17, 0x87, 0xfa, 0x16, 0x88, 
+    0xa3, 0xd5, 0x17, 0xe8, 0xd1, 0x52, 0x59, 0x10, 0xb3, 0xa8, 
+    0xa6, 0xdc, 0x3d, 0x55, 0xa3, 0x89, 0x35, 0x91, 0x3b, 0x3c, 
+    0x45, 0x32, 0xfe, 0x33, 0xcc, 0x6a, 0x1d, 0x0f, 0x24, 0x21, 
+    0x7b, 0xab, 0x59, 0x6b, 0x92, 0x38, 0x9b, 0x65, 0x82, 0x8f, 
+    0xfa, 0x91, 0x5e, 0x2c, 0xeb, 0xb0, 0x89, 0x2f, 0xd3, 0x1e, 
+    0x06, 0x52, 0x0b, 0xdf, 0x16, 0x66, 0x1f, 0xe2, 0x96, 0xcf, 
+    0x28, 0x29, 0x0e, 0xad, 0xe2, 0x88, 0x0c, 0x81, 0xa9, 0x69, 
+    0x8c, 0x0c, 0x8e, 0xf5, 0xae, 0xeb, 0x76, 0x80, 0xf8, 0x86, 
+    0xbe, 0xa8, 0x24, 0x2f, 0xa3, 0xff, 0x80, 0xea, 0x0d, 0xbf, 
+    0x9d, 0x6d, 0xdf, 0xa2, 0x73, 0x43, 0x09, 0x44, 0x3b, 0x66, 
+    0x64, 0x8f, 0x2c, 0xc4, 0x77, 0x69, 0xcc, 0x3d, 0x8f, 0xe5, 
+    0xae, 0xca, 0x76, 0x80, 0xd8, 0x84, 0xbf, 0xa8, 0xa1, 0x2f, 
+    0x83, 0xff, 0x82, 0xea, 0x0d, 0x9f, 0x9d, 0x6d, 0xcf, 0xa2, 
+    0xf3, 0x53, 0x09, 0x60, 0x3f, 0x64, 0x64, 0x8f, 0x2c, 0x84, 
+    0x7b, 0x69, 0x8c, 0x3c, 0x70, 0x1a, 0x51, 0x35, 0x89, 0x7f, 
+    0x27, 0x7b, 0x40, 0x57, 0x5e, 0xd0, 0x7c, 0x00, 0x7d, 0x15, 
+    0xf2, 0x60, 0x62, 0x92, 0x30, 0x5d, 0x0c, 0xac, 0xf6, 0x9f, 
+    0xc0, 0x9b, 0x9b, 0x70, 0xd3, 0x7b, 0x84, 0x96, 0x73, 0xc3, 
+    0x00, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x07, 0x00, 
+    0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x05, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x03, 0x00, 
+    0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 
+    0x07, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x03, 0x00, 
+    0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x06, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x06, 0x00, 
+    0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x03, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x07, 0x00, 
+    0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x05, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x06, 0x00, 
+    0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 
+    0x06, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x03, 0x00, 
+    0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 
+    0x03, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x02, 0x00, 
+    0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 
+    0x07, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x07, 0x00, 
+    0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x07, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x03, 0x00, 
+    0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 
+    0x03, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x07, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 
+    0x05, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x07, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x06, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x04, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x96, 0x8f, 
+    0xda, 0xe7, 0x48, 0x5f, 0xd5, 0xda, 0x90, 0x1b, 0x02, 0x32, 
+    0x10, 0x49, 0xd0, 0x0d, 0xcd, 0xee, 0xd8, 0x87, 0xeb, 0xc0, 
+    0x98, 0x49, 0x3b, 0x22, 0x39, 0x13, 0x41, 0xb4, 0x1b, 0x84, 
+    0xdb, 0x75, 0xca, 0xf6, 0xd3, 0x69, 0xe2, 0x6d, 0x10, 0xb7, 
+    0xca, 0xf3, 0xc3, 0xbc, 0xfa, 0xfb, 0x64, 0x4b, 0x91, 0xc3, 
+    0x7a, 0xa0, 0xb8, 0xd3, 0x2a, 0x15, 0x5e, 0x5b, 0x48, 0x80, 
+    0x7f, 0xea, 0xee, 0x96, 0x31, 0x6a, 0x31, 0x60, 0x3d, 0x94, 
+    0xb3, 0x55, 0x82, 0x3b, 0x77, 0xe3, 0x19, 0xdb, 0x6a, 0x0e, 
+    0x55, 0x73, 0xba, 0x5f, 0xfb, 0xae, 0xdf, 0x52, 0x7e, 0xd6, 
+    0xf4, 0xb8, 0x3c, 0x81, 0x60, 0x08, 0xcb, 0xdc, 0x68, 0xb8, 
+    0xc2, 0xd2, 0x02, 0x67, 0x6f, 0x8a, 0x5f, 0xd1, 0x15, 0x42, 
+    0x02, 0xc3, 0x9e, 0x35, 0x66, 0xaa, 0x11, 0xe2, 0xfe, 0xca, 
+    0x46, 0x55, 0x25, 0x3b, 0xfe, 0x9c, 0x2c, 0xa3, 0xd7, 0xe1, 
+    0x70, 0x77, 0xa3, 0xbb, 0x52, 0xdd, 0x9d, 0xb7, 0x32, 0x4f, 
+    0x36, 0xd2, 0x91, 0x13, 0x34, 0x9b, 0x06, 0x29, 0xb4, 0x81, 
+    0xa3, 0xab, 0x24, 0xfd, 0x95, 0x06, 0x7d, 0x25, 0x33, 0x11, 
+    0x17, 0x98, 0x3d, 0xea, 0xf5, 0xda, 0xe8, 0x40, 0xc8, 0xe7, 
+    0x3a, 0xe2, 0x1a, 0x26, 0xdf, 0xbe, 0xdb, 0x29, 0x13, 0x33, 
+    0x70, 0x17, 0x42, 0x2b, 0x70, 0x44, 0x0f, 0xae, 0x24, 0x67, 
+    0xf5, 0x24, 0x76, 0x6d, 0x06, 0xbb, 0x14, 0x90, 0x58, 0xbe, 
+    0x55, 0x08, 0xbd, 0x00, 0x8f, 0x89, 0x7f, 0xc6, 0x03, 0x6b, 
+    0xba, 0x98, 0x35, 0xea, 0x13, 0xb3, 0x30, 0x17, 0x42, 0x2b, 
+    0x70, 0x45, 0x07, 0xae, 0x24, 0x67, 0xf5, 0x04, 0x76, 0x65, 
+    0x06, 0x39, 0x16, 0x90, 0x7c, 0xba, 0x55, 0x48, 0xfd, 0x00, 
+    0x8e, 0xcb, 0x7f, 0xc6, 0x03, 0x63, 0xba, 0x9e, 0x37, 0xea, 
+    0xec, 0x4c, 0xcf, 0xe8, 0xbd, 0xd4, 0x8f, 0xba, 0xf8, 0x51, 
+    0xdb, 0x98, 0x0a, 0xfb, 0x89, 0x9a, 0xf9, 0xc6, 0xe9, 0x6f, 
+    0x83, 0x45, 0xaa, 0xb7, 0x02, 0xff, 0x71, 0x34, 0x80, 0x39, 
+    0xfc, 0x9c, 0x45, 0x61, 0xc8, 0x15, 0x11, 0x4b, 0xa5, 0xdf, 
+    0xa7, 0xd1, 0x37, 0xdb, 0x89, 0x22, 0x9d, 0x72, 0xb9, 0x15, 
+    0xc0, 0xa8, 0xe4, 0xb1, 0x1b, 0x5f, 0xf3, 0xf5, 0x01, 0xd1, 
+    0x3f, 0x1b, 0x5c, 0x29, 0x4b, 0x59, 0x55, 0x06, 0xfe, 0xf2, 
+    0xcf, 0x60, 0x8f, 0xf9, 0x49, 0x88, 0xea, 0x3d, 0xbc, 0x86, 
+    0xcc, 0x70, 0x0f, 0x3f, 0xd3, 0x98, 0x43, 0x32, 0x16, 0x9c, 
+    0x13, 0x85, 0x4c, 0xcf, 0xff, 0x2f, 0xe4, 0xd4, 0x65, 0x3f, 
+    0x1a, 0xa2, 0xd5, 0xdc, 0x79, 0x30, 0xb9, 0x89, 0xb0, 0x55, 
+    0x59, 0xdf, 0xbc, 0xd8, 0x86, 0x63, 0x79, 0x22, 0x4b, 0xa3, 
+    0x76, 0xf1, 0x96, 0xd9, 0x42, 0x11, 0x4f, 0x6e, 0xac, 0x60, 
+    0x8a, 0x0d, 0xe2, 0xd3, 0x96, 0xfc, 0xb9, 0xf6, 0x48, 0x51, 
+    0xb8, 0x5f, 0xe3, 0x86, 0x13, 0x6d, 0x3c, 0x92, 0xdd, 0xb3, 
+    0xea, 0x49, 0xf0, 0x7c, 0x33, 0x2b, 0x9f, 0x0b, 0x95, 0xc9, 
+    0x4b, 0xe7, 0x0f, 0x3c, 0x27, 0xff, 0xf5, 0xde, 0xc4, 0xfc, 
+    0x61, 0x1f, 0xb0, 0x53, 0x6c, 0x9b, 0x0a, 0x87, 0x05, 0xb1, 
+    0xf7, 0x82, 0x9f, 0x19, 0xec, 0x77, 0x1c, 0xbe, 0x9c, 0x98, 
+    0x23, 0xeb, 0xbf, 0xf5, 0x2a, 0x8b, 0x44, 0x82, 0x82, 0xa2, 
+    0x3a, 0x4a, 0xf5, 0xd7, 0x7f, 0xe3, 0xa7, 0xab, 0xd7, 0xa5, 
+    0xa2, 0x25, 0x4e, 0x49, 0x0c, 0x4a, 0x1b, 0xdc, 0x9f, 0x19, 
+    0xdc, 0x75, 0x98, 0x98, 0x78, 0x54, 0xab, 0xeb, 0x8f, 0x9e, 
+    0x7a, 0xea, 0x66, 0xe7, 0xc7, 0xe8, 0x6f, 0xfe, 0x72, 0xdf, 
+    0xf3, 0x27, 0xa3, 0xeb, 0xee, 0xf6, 0x20, 0xb7, 0x50, 0xe7, 
+    0x16, 0x19, 0x13, 0xcc, 0x1f, 0x19, 0xdc, 0x35, 0x98, 0x98, 
+    0xf8, 0x54, 0xa3, 0xeb, 0x9e, 0x9e, 0x7a, 0xca, 0x46, 0xe7, 
+    0xc7, 0xa8, 0x6f, 0xfe, 0xf6, 0xdf, 0xf3, 0xe7, 0xa3, 0xeb, 
+    0xc6, 0xf6, 0x20, 0xb7, 0x50, 0x47, 0x16, 0x09, 0xec, 0x33, 
+    0xe0, 0xe6, 0x23, 0x8a, 0x67, 0x67, 0x07, 0xab, 0x5c, 0x14, 
+    0x61, 0x61, 0x85, 0x35, 0xb9, 0x18, 0x38, 0x57, 0x90, 0x01, 
+    0x09, 0x20, 0x0c, 0x18, 0x5c, 0x14, 0x39, 0x09, 0xdf, 0x48, 
+    0xaf, 0xb8, 0xe9, 0xf6, 0x01, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x04, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x03, 0x00, 
+    0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 
+    0x04, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x06, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 
+    0x04, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 
+    0x03, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x07, 0x00, 
+    0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 
+    0x04, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x03, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x06, 0x00, 
+    0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 
+    0x03, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x05, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x05, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x02, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 
+    0x06, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x04, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x03, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x02, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x7b, 0x74, 0x95, 0x3b, 0xd3, 0x34, 0x8d, 0x2b, 
+    0x05, 0x8a, 0x6e, 0x6c, 0x05, 0x69, 0xbe, 0xc1, 0xe7, 0xf0, 
+    0x22, 0x2f, 0xaa, 0x5c, 0x4c, 0xb6, 0x21, 0x28, 0x43, 0x4c, 
+    0x31, 0x06, 0x88, 0xcb, 0x47, 0x6d, 0x2d, 0x26, 0x28, 0x37, 
+    0x81, 0xc1, 0xeb, 0x2a, 0x46, 0x8e, 0x76, 0x0d, 0xa4, 0x4a, 
+    0xd9, 0x0b, 0xe5, 0x82, 0x98, 0x35, 0xdd, 0xc8, 0x75, 0xec, 
+    0x8d, 0x35, 0xa4, 0xea, 0x73, 0x48, 0x7f, 0xdf, 0xaf, 0xf2, 
+    0xa7, 0x50, 0x60, 0x2b, 0x9b, 0x7c, 0xb8, 0xc5, 0xd4, 0x8b, 
+    0x7e, 0x7e, 0x64, 0x47, 0x8b, 0x99, 0xfa, 0x74, 0xff, 0x36, 
+    0xfa, 0x97, 0x5f, 0x24, 0x5d, 0x2b, 0x90, 0x72, 0x14, 0xac, 
+    0x3e, 0xc4, 0xf2, 0xda, 0x37, 0x9b, 0xc0, 0x38, 0xbf, 0xdf, 
+    0xbe, 0x30, 0xe9, 0x06, 0xd2, 0x8a, 0x30, 0x3c, 0xd7, 0x4f, 
+    0x99, 0xa3, 0xc5, 0x63, 0x15, 0x42, 0xff, 0x1a, 0x14, 0x6b, 
+    0xea, 0x5a, 0x1a, 0xeb, 0xa6, 0x8f, 0x8a, 0x5b, 0x21, 0x15, 
+    0x67, 0xe5, 0x73, 0x2e, 0xb0, 0xb3, 0x7f, 0x52, 0x99, 0xda, 
+    0x84, 0x93, 0xf3, 0x9b, 0x9b, 0xea, 0x30, 0x16, 0x9d, 0x47, 
+    0x58, 0xcd, 0x14, 0x52, 0x7e, 0x14, 0x32, 0x56, 0x6b, 0x89, 
+    0xb6, 0x32, 0x6b, 0xba, 0xc5, 0x56, 0x6c, 0xc5, 0xde, 0x2e, 
+    0x0f, 0xfd, 0x6d, 0x76, 0x85, 0xba, 0x90, 0x7b, 0xf2, 0x39, 
+    0xd6, 0x4b, 0x93, 0x34, 0xdf, 0x6f, 0x7e, 0xdd, 0xd2, 0x70, 
+    0x2e, 0x31, 0x6b, 0x4d, 0x6a, 0xc1, 0xbe, 0x27, 0x2b, 0x90, 
+    0xa7, 0x7e, 0x39, 0xbd, 0xcd, 0x3a, 0xaf, 0x9d, 0x7f, 0x56, 
+    0x89, 0xba, 0x90, 0x7a, 0xf2, 0x39, 0xd6, 0x4b, 0x93, 0x34, 
+    0xdf, 0x67, 0x7e, 0xdd, 0x52, 0x50, 0x2e, 0x31, 0x6b, 0x4c, 
+    0x6a, 0xc1, 0xbe, 0x23, 0x2b, 0x90, 0xa7, 0x5e, 0x79, 0xb5, 
+    0xcd, 0x3a, 0xaf, 0xbd, 0x80, 0xa9, 0x76, 0x45, 0x6f, 0x85, 
+    0x0d, 0xc6, 0x29, 0xb4, 0x6c, 0xcb, 0x20, 0x98, 0x81, 0x22, 
+    0xad, 0xaf, 0xd1, 0xce, 0x94, 0xb3, 0x95, 0x3e, 0x41, 0xdc, 
+    0xd4, 0x6f, 0x58, 0xa1, 0x86, 0x4a, 0x32, 0xc5, 0x50, 0x42, 
+    0xf3, 0x49, 0x33, 0x45, 0x53, 0x14, 0xcf, 0x66, 0xde, 0xa9, 
+    0x44, 0xe2, 0x96, 0xff, 0xdc, 0xe2, 0x42, 0x25, 0x7e, 0x0f, 
+    0xb5, 0x4c, 0x16, 0xca, 0x47, 0x40, 0x04, 0x18, 0xf9, 0xb6, 
+    0x70, 0x56, 0x3c, 0x47, 0xa3, 0xe0, 0x6f, 0x27, 0x18, 0x56, 
+    0xe5, 0x6d, 0xea, 0x04, 0xfb, 0x04, 0x60, 0x9a, 0xcb, 0xa7, 
+    0x60, 0xda, 0xaf, 0xe4, 0xd2, 0x10, 0xca, 0xba, 0xd1, 0x4e, 
+    0x3d, 0xbb, 0x0b, 0xe6, 0x63, 0xce, 0x28, 0x4b, 0x49, 0xe2, 
+    0xfa, 0x19, 0xcd, 0xd1, 0xed, 0xc6, 0xa7, 0x73, 0x3f, 0xc4, 
+    0xe3, 0xf0, 0x54, 0xb7, 0xa8, 0x9d, 0xc4, 0x65, 0xf2, 0xcd, 
+    0x04, 0x49, 0xd7, 0x9b, 0xda, 0x9e, 0xb4, 0xeb, 0xa1, 0xe2, 
+    0xb5, 0x6a, 0x84, 0x0d, 0xb0, 0xc1, 0xd7, 0x9f, 0x2d, 0x08, 
+    0xdf, 0x69, 0x7f, 0xc0, 0xcb, 0xca, 0xf0, 0x0c, 0x33, 0x73, 
+    0x2f, 0xbe, 0xe6, 0xb9, 0x0d, 0xcc, 0x11, 0x9f, 0x86, 0xa6, 
+    0xbf, 0xa4, 0x68, 0xa4, 0x57, 0x55, 0x80, 0x2a, 0x26, 0x24, 
+    0x87, 0xf7, 0x55, 0x17, 0x9e, 0x41, 0x95, 0x09, 0xcc, 0x5a, 
+    0x17, 0xd0, 0x6f, 0xaa, 0xbb, 0x35, 0x22, 0x56, 0xb3, 0x8d, 
+    0xc9, 0xb6, 0xc5, 0x51, 0x5a, 0x3e, 0x15, 0x22, 0x56, 0x67, 
+    0x20, 0xf4, 0x09, 0x40, 0x8a, 0x65, 0x37, 0x60, 0xb6, 0x3c, 
+    0x27, 0x19, 0x55, 0x2d, 0xcc, 0x61, 0xcc, 0xd2, 0x62, 0x19, 
+    0x72, 0xf3, 0xa2, 0x9a, 0xeb, 0x22, 0x2d, 0x94, 0xd5, 0x19, 
+    0xd2, 0x7a, 0x55, 0x06, 0xae, 0xad, 0x75, 0xb4, 0xad, 0x60, 
+    0xb4, 0x52, 0xb5, 0x3d, 0x97, 0x96, 0x0f, 0x51, 0x55, 0x29, 
+    0xcc, 0x40, 0xcc, 0xd2, 0x62, 0x18, 0x72, 0xb3, 0xa2, 0x9e, 
+    0xe3, 0x29, 0x2d, 0x94, 0xd5, 0x59, 0xf2, 0x5a, 0x55, 0x06, 
+    0x2e, 0xa5, 0x75, 0xb4, 0xad, 0x60, 0xb4, 0x42, 0xb5, 0x65, 
+    0x97, 0x96, 0xf0, 0xae, 0xaa, 0xd6, 0x33, 0xbf, 0x33, 0x2d, 
+    0x9d, 0xe7, 0x8d, 0x4c, 0x5d, 0x61, 0x1c, 0xd6, 0xd2, 0x6b, 
+    0x2a, 0xa6, 0x0d, 0xa5, 0xaa, 0xf9, 0xd1, 0x5a, 0x8a, 0x4b, 
+    0x52, 0x9f, 0x4b, 0xbd, 0x4a, 0x9a, 0x68, 0x69, 0x00, 0x00, 
+    0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x04, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x03, 0x00, 
+    0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 
+    0x06, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x02, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x07, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x03, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x06, 0x00, 
+    0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 
+    0x07, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 
+    0x04, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x04, 0x00, 
+    0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 
+    0x02, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x02, 0x00, 
+    0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 
+    0x06, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x03, 0x00, 
+    0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x02, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x02, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x04, 0x00, 
+    0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 
+    0x04, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x73, 0x6c, 0xfd, 0x2c, 
+    0x58, 0xf8, 0x2a, 0x03, 0x14, 0xff, 0x58, 0xaf, 0x67, 0x3c, 
+    0x45, 0xe6, 0x67, 0x4a, 0xfe, 0x67, 0x79, 0x72, 0x8a, 0x8d, 
+    0x19, 0xf7, 0x7b, 0xc9, 0x0c, 0xa3, 0x88, 0xc2, 0x4a, 0x4e, 
+    0x1b, 0xc7, 0x76, 0xbe, 0xf8, 0xf0, 0x34, 0x5b, 0xbf, 0x88, 
+    0xef, 0xf8, 0x28, 0xd7, 0x37, 0xf5, 0x15, 0x06, 0x53, 0x06, 
+    0x0a, 0xbd, 0xa0, 0x30, 0x6e, 0x9e, 0x07, 0x62, 0x64, 0x04, 
+    0x9f, 0x85, 0x11, 0x9c, 0xa3, 0xfa, 0x63, 0xa9, 0x30, 0xcc, 
+    0x6f, 0x54, 0x8f, 0x44, 0x2d, 0x16, 0xaa, 0xbe, 0xef, 0x6f, 
+    0x26, 0x6c, 0x59, 0x00, 0x26, 0x7b, 0x84, 0x38, 0x39, 0x55, 
+    0xd9, 0xfc, 0xe2, 0xa8, 0x4a, 0xeb, 0x7c, 0xa4, 0xa7, 0xd5, 
+    0x4e, 0x82, 0xe8, 0x99, 0xbe, 0xe3, 0x7f, 0x14, 0xb3, 0x7a, 
+    0x89, 0x88, 0x9e, 0x16, 0x7e, 0x22, 0x46, 0xd8, 0x2a, 0xe1, 
+    0x29, 0x0b, 0x6c, 0xbc, 0xf8, 0x45, 0xb6, 0x95, 0xfd, 0x17, 
+    0x9e, 0xe8, 0x18, 0xa8, 0x29, 0xa3, 0xb8, 0x96, 0x0b, 0xa9, 
+    0xca, 0xb2, 0x92, 0xb9, 0xf0, 0x51, 0x95, 0x8c, 0x48, 0x91, 
+    0x55, 0x81, 0x56, 0x16, 0x5c, 0x0d, 0xae, 0xbc, 0x11, 0x06, 
+    0xaf, 0x2f, 0x23, 0x9f, 0x83, 0x5f, 0xa0, 0xd0, 0x51, 0xf4, 
+    0x47, 0x01, 0x0f, 0x09, 0xc0, 0xb8, 0xfe, 0x8b, 0x5b, 0xb4, 
+    0x50, 0x73, 0x9d, 0x88, 0xd8, 0xb9, 0x5f, 0x42, 0x87, 0x56, 
+    0x1e, 0x41, 0x99, 0x09, 0x10, 0x12, 0xe4, 0x49, 0xe8, 0x9f, 
+    0xd7, 0xdc, 0xcd, 0x60, 0x5d, 0xe0, 0x2a, 0x09, 0x0b, 0x30, 
+    0xcc, 0xb9, 0xee, 0xab, 0x5b, 0xb4, 0x50, 0x73, 0x9d, 0x88, 
+    0xd8, 0xb9, 0x5f, 0x02, 0x87, 0x56, 0x1e, 0x41, 0x09, 0x09, 
+    0x14, 0x12, 0xa4, 0x4d, 0xe2, 0x9f, 0xd7, 0xdc, 0xcd, 0xe0, 
+    0x59, 0xe0, 0x2b, 0x09, 0x0b, 0x31, 0xcc, 0xb9, 0x11, 0x54, 
+    0xa4, 0x4b, 0xaf, 0x8c, 0x62, 0x77, 0x27, 0x46, 0xa0, 0xfd, 
+    0x78, 0xa9, 0xe1, 0xbe, 0xf6, 0xf6, 0xeb, 0xed, 0x5b, 0xb2, 
+    0x1d, 0x60, 0x28, 0x23, 0x32, 0x1f, 0xa6, 0x1f, 0xd4, 0xf6, 
+    0xf4, 0xce, 0x33, 0x46, 0x45, 0xf5, 0x65, 0xb7, 0x61, 0xa5, 
+    0x71, 0x0b, 0x11, 0x01, 0x73, 0xd9, 0xb1, 0x1d, 0xa0, 0xb5, 
+    0xbe, 0x21, 0x9e, 0x05, 0xf1, 0x20, 0x7b, 0x3d, 0x8c, 0x85, 
+    0x23, 0x3b, 0x5f, 0x98, 0xc6, 0x31, 0xc9, 0x21, 0x91, 0xd7, 
+    0xb7, 0xf6, 0x87, 0xd0, 0x41, 0xf1, 0x28, 0xbe, 0xf5, 0x1c, 
+    0x1a, 0x70, 0x69, 0x81, 0x53, 0x69, 0xd8, 0xcf, 0xa9, 0x51, 
+    0x0d, 0x1d, 0xf8, 0x52, 0x94, 0x22, 0xf5, 0xf7, 0x1c, 0xe5, 
+    0xcb, 0x62, 0x37, 0x42, 0x1b, 0x9a, 0x61, 0xa2, 0xc8, 0xc5, 
+    0xe7, 0x62, 0xab, 0x14, 0x6a, 0x9f, 0xfc, 0x73, 0x2c, 0xa3, 
+    0x41, 0xe0, 0xee, 0x90, 0xed, 0xfd, 0xf0, 0xea, 0x06, 0x2d, 
+    0x37, 0x8e, 0x8d, 0xfb, 0xfc, 0xe1, 0x21, 0x90, 0x1c, 0x1a, 
+    0x26, 0x9f, 0x66, 0xde, 0xa0, 0x74, 0x3d, 0xb8, 0x26, 0x82, 
+    0x62, 0x2a, 0xc8, 0xa0, 0x04, 0x73, 0xe7, 0xd3, 0xa1, 0xfc, 
+    0x39, 0x58, 0x49, 0x1f, 0xb5, 0x7e, 0x41, 0xd6, 0x58, 0xb8, 
+    0x45, 0xed, 0x90, 0xea, 0x4d, 0xdf, 0x99, 0x47, 0x08, 0x10, 
+    0x00, 0xe4, 0x10, 0xfa, 0x67, 0x03, 0x5e, 0xef, 0xcd, 0xfe, 
+    0xfc, 0x1b, 0x61, 0xb6, 0x2d, 0x24, 0x6e, 0xfe, 0x7b, 0xfc, 
+    0x10, 0x46, 0xcd, 0x37, 0xc7, 0x02, 0xd8, 0x06, 0x7e, 0xf2, 
+    0xf1, 0x5c, 0xac, 0x91, 0xf0, 0x14, 0xc8, 0xc4, 0x93, 0x78, 
+    0x0e, 0x20, 0xa8, 0xbc, 0x3d, 0xee, 0xfc, 0x03, 0x4a, 0xbe, 
+    0x28, 0x18, 0xbd, 0xa0, 0xff, 0xfe, 0x59, 0x72, 0xcb, 0xb7, 
+    0x4f, 0x32, 0xf0, 0x34, 0x78, 0xf6, 0x78, 0x58, 0xb0, 0x97, 
+    0x60, 0x14, 0xc8, 0xc4, 0x13, 0x78, 0x2e, 0x22, 0x78, 0xac, 
+    0xbd, 0xee, 0xfc, 0x03, 0x4b, 0xbe, 0x29, 0x38, 0x3d, 0xe0, 
+    0x7f, 0xfe, 0x51, 0x72, 0xc3, 0xb7, 0x4f, 0x32, 0xf0, 0x34, 
+    0x78, 0xf6, 0x78, 0x58, 0xb0, 0x97, 0x9f, 0xeb, 0x37, 0x3b, 
+    0xec, 0x87, 0xd1, 0xdd, 0x87, 0x53, 0x42, 0x11, 0x03, 0xfc, 
+    0xb4, 0x41, 0xd6, 0xc7, 0xc2, 0x1f, 0x80, 0x01, 0xae, 0x8d, 
+    0x3c, 0x48, 0xb0, 0xcd, 0x0f, 0xcb, 0x87, 0x09, 0x87, 0xa7, 
+    0x4f, 0x68, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x04, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x04, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 
+    0x04, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x05, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x03, 0x00, 
+    0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 
+    0x07, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x04, 0x00, 
+    0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x06, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x04, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x02, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 
+    0x04, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x04, 0x00, 
+    0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 
+    0x02, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x02, 0x00, 
+    0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 
+    0x02, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x02, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x04, 0x00, 
+    0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x04, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x03, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x02, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 
+    0x05, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x07, 0x00, 
+    0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 
+    0x43, 0x48, 0x28, 0xff, 0x98, 0x33, 0x51, 0x26, 0x05, 0x65, 
+    0x55, 0x2f, 0x33, 0xf0, 0x8e, 0x61, 0xe5, 0xc1, 0x91, 0x02, 
+    0xb2, 0x8e, 0x82, 0x4b, 0xbc, 0xc7, 0xe5, 0x5e, 0x96, 0x25, 
+    0xe2, 0xeb, 0x5f, 0x79, 0x0f, 0x18, 0x52, 0x8c, 0x3d, 0x42, 
+    0xf7, 0x97, 0xa9, 0xd8, 0xf0, 0xcd, 0xaf, 0xe5, 0xa7, 0x3a, 
+    0x8b, 0x93, 0x26, 0x96, 0xc4, 0xb3, 0xd5, 0x2e, 0xbf, 0x9d, 
+    0x9b, 0xd6, 0x12, 0x60, 0xc9, 0x22, 0x3d, 0x93, 0xe3, 0xa0, 
+    0x3c, 0x8d, 0x78, 0xf3, 0x99, 0x27, 0x9f, 0xf3, 0xaf, 0xad, 
+    0x35, 0x4f, 0xd6, 0x5b, 0xf7, 0x78, 0xd5, 0xf5, 0x95, 0x0c, 
+    0xd5, 0x35, 0xbf, 0x67, 0x0b, 0xc2, 0x20, 0x5f, 0x6d, 0x87, 
+    0x2f, 0xbf, 0xf7, 0x8e, 0xe4, 0x70, 0x62, 0x2d, 0x1c, 0x45, 
+    0xbc, 0x04, 0x2d, 0x3f, 0x19, 0x6f, 0x14, 0xf3, 0xb6, 0x72, 
+    0x7a, 0x5e, 0x99, 0x48, 0x0e, 0xc3, 0x07, 0x2f, 0xcf, 0x91, 
+    0xc7, 0x21, 0x80, 0x75, 0x00, 0x94, 0xc5, 0xea, 0x5f, 0x27, 
+    0x97, 0x81, 0xda, 0xa0, 0x13, 0x88, 0x68, 0x0c, 0x82, 0x92, 
+    0xf0, 0x44, 0xff, 0xfa, 0xef, 0x52, 0xe2, 0x69, 0xe5, 0x9a, 
+    0x4e, 0x85, 0x31, 0x71, 0xa3, 0xbf, 0xe6, 0x8b, 0xe9, 0x03, 
+    0x00, 0x99, 0xd9, 0x8b, 0x74, 0xb1, 0x55, 0x8a, 0x75, 0x8f, 
+    0xb2, 0x8a, 0x70, 0x0c, 0x85, 0x32, 0xda, 0x53, 0xbd, 0xbb, 
+    0xf0, 0xfa, 0xd0, 0x79, 0xe9, 0x68, 0x56, 0xa5, 0x07, 0x75, 
+    0x92, 0xbe, 0xa8, 0x99, 0xbb, 0x40, 0x8e, 0x81, 0xcd, 0xf8, 
+    0xdc, 0x77, 0x1a, 0x30, 0x96, 0xce, 0x12, 0x8a, 0x78, 0x0c, 
+    0x84, 0xb2, 0xf8, 0x53, 0xbd, 0xfb, 0xf4, 0x7a, 0xf2, 0x79, 
+    0xe9, 0x68, 0x5e, 0xa5, 0x07, 0x75, 0x83, 0xbf, 0xaa, 0x9b, 
+    0xbb, 0x43, 0x86, 0x81, 0xcd, 0xf9, 0xdc, 0x77, 0x12, 0x20, 
+    0x96, 0xcf, 0xed, 0x75, 0x87, 0xf3, 0x7b, 0x4d, 0x07, 0xac, 
+    0x42, 0x04, 0x0b, 0x85, 0x0d, 0x86, 0x16, 0x97, 0xa1, 0x5a, 
+    0xf8, 0x8a, 0x7c, 0x40, 0x55, 0x64, 0x44, 0xbc, 0x79, 0x7e, 
+    0x32, 0x06, 0x23, 0x88, 0xed, 0xdf, 0x69, 0x30, 0x1a, 0xe2, 
+    0xe7, 0xea, 0x99, 0x16, 0x58, 0x84, 0x8e, 0x2e, 0x60, 0xe4, 
+    0x63, 0x5b, 0x41, 0x45, 0x47, 0x94, 0x08, 0x23, 0x72, 0xfb, 
+    0x96, 0x58, 0x29, 0xb6, 0x7c, 0x1e, 0xb1, 0x3b, 0x9f, 0x1b, 
+    0x2b, 0x28, 0x45, 0x1b, 0xf9, 0x9d, 0xf9, 0xb3, 0xd4, 0x0f, 
+    0x54, 0x94, 0x9d, 0x8f, 0xc1, 0x5c, 0x7e, 0x71, 0xb4, 0xd0, 
+    0xb4, 0x60, 0x84, 0xa2, 0x9d, 0x33, 0x32, 0xa9, 0x24, 0x97, 
+    0x62, 0xd0, 0xc6, 0x38, 0xdb, 0xe9, 0x5d, 0x52, 0xc7, 0xd5, 
+    0x06, 0x63, 0x1d, 0x18, 0x20, 0x36, 0x30, 0x7b, 0xf1, 0x0c, 
+    0x7f, 0xad, 0x8d, 0x9a, 0xfa, 0xfa, 0x2e, 0x56, 0x0a, 0xbc, 
+    0xb2, 0xcf, 0x38, 0x23, 0x1d, 0xf8, 0xb9, 0x12, 0x6e, 0x3f, 
+    0x44, 0x81, 0xab, 0x59, 0x87, 0x12, 0xe7, 0x9e, 0xfc, 0x9a, 
+    0x19, 0x97, 0x7c, 0x53, 0xda, 0xe7, 0xd0, 0x32, 0xea, 0xb6, 
+    0x36, 0x36, 0x84, 0xad, 0x54, 0x76, 0x14, 0x8d, 0x79, 0x15, 
+    0x2f, 0xe3, 0xae, 0x36, 0xbb, 0xa1, 0x83, 0xb7, 0xc6, 0xaf, 
+    0x81, 0x9d, 0x7b, 0x53, 0x52, 0xfc, 0x74, 0x8d, 0x2f, 0xb1, 
+    0x5c, 0x03, 0xa3, 0xe4, 0x1d, 0x06, 0x2d, 0x56, 0x98, 0x36, 
+    0x38, 0xfa, 0xe6, 0x96, 0x19, 0x7d, 0x27, 0x08, 0xae, 0x91, 
+    0xc3, 0x1c, 0xcd, 0x2e, 0x32, 0xa6, 0xfb, 0x77, 0x2f, 0x52, 
+    0xe2, 0xde, 0x13, 0x41, 0x61, 0x33, 0xed, 0xc3, 0xb3, 0x0c, 
+    0xcd, 0x02, 0x5c, 0xfe, 0x1c, 0x1d, 0xf7, 0x7a, 0x93, 0x8e, 
+    0xd1, 0x74, 0xa7, 0x58, 0x29, 0x94, 0x86, 0x15, 0x2f, 0x3d, 
+    0x62, 0x3e, 0xd3, 0x24, 0x2f, 0x52, 0x62, 0xde, 0x11, 0x41, 
+    0x61, 0x33, 0xdd, 0xc3, 0xb3, 0xac, 0xcd, 0x02, 0x5c, 0xf6, 
+    0x1c, 0x1d, 0xb0, 0x7a, 0x92, 0x8e, 0x19, 0x74, 0x27, 0x58, 
+    0xab, 0x94, 0x86, 0x15, 0xaf, 0xbd, 0x62, 0x3e, 0xc3, 0x34, 
+    0xd0, 0xad, 0x9d, 0x21, 0xee, 0xbe, 0x9e, 0xcc, 0x22, 0x3c, 
+    0x4c, 0x53, 0x32, 0xfd, 0xa3, 0x09, 0xe3, 0xe2, 0x4f, 0x85, 
+    0x6d, 0x71, 0xe6, 0x8b, 0xd8, 0xa7, 0x54, 0x6b, 0x79, 0xea, 
+    0x50, 0x42, 0x9d, 0xc1, 0x3c, 0xcb, 0x06, 0x00, 0x00, 0x00, 
+    0x05, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 
+    0x02, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x05, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 
+    0x05, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x03, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x04, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 
+    0x04, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 
+    0x07, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x06, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x04, 0x00, 
+    0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 
+    0x02, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x07, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x06, 0x00, 
+    0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x04, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 
+    0x02, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x03, 0x00, 
+    0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0xed, 0x2a, 0x5a, 0x40, 0x7a, 0x0f, 
+    0xcb, 0x42, 0x80, 0x94, 0xe0, 0x5c, 0x86, 0x3c, 0x5a, 0x0e, 
+    0x52, 0x63, 0x2b, 0x66, 0xc5, 0x7d, 0x67, 0xb9, 0x1d, 0x88, 
+    0x03, 0x2f, 0x46, 0x8b, 0x8d, 0x5f, 0x3b, 0x2a, 0xe8, 0xed, 
+    0x25, 0x8e, 0xf1, 0x6d, 0x7e, 0xa1, 0x26, 0xa1, 0x09, 0x3a, 
+    0xc0, 0xff, 0x73, 0x3f, 0x61, 0x3d, 0x92, 0x64, 0x1e, 0x4a, 
+    0xec, 0x0a, 0x75, 0x9f, 0x0e, 0xe2, 0xd7, 0x39, 0x7b, 0x78, 
+    0xfc, 0xc0, 0x18, 0xf5, 0x57, 0x16, 0x4d, 0x46, 0x6a, 0x3d, 
+    0xa8, 0x04, 0xd8, 0xaa, 0x45, 0xaa, 0x10, 0xc1, 0xd7, 0x58, 
+    0x22, 0x79, 0x30, 0x02, 0xb1, 0x3f, 0x01, 0x07, 0xc0, 0x96, 
+    0x3e, 0x33, 0xd0, 0x27, 0xaf, 0x49, 0x7b, 0xca, 0x3f, 0x1c, 
+    0x26, 0x6a, 0x69, 0x50, 0xd0, 0x3e, 0x27, 0x0f, 0x6f, 0x50, 
+    0xd0, 0x79, 0x83, 0x86, 0xa4, 0xdd, 0xbc, 0x7d, 0x03, 0xac, 
+    0x4e, 0x7d, 0xb6, 0x8e, 0x65, 0x17, 0x43, 0x26, 0x71, 0x53, 
+    0x4e, 0xc7, 0x93, 0xea, 0xd2, 0xe8, 0x53, 0xa1, 0xfe, 0xc7, 
+    0x24, 0x66, 0x50, 0xcc, 0xa4, 0x80, 0x05, 0x2a, 0x50, 0x48, 
+    0xe4, 0xac, 0xf7, 0xa7, 0xc9, 0x0a, 0x55, 0xea, 0x61, 0x1c, 
+    0x9e, 0xce, 0x61, 0x5b, 0x74, 0xf9, 0x31, 0x4a, 0x94, 0xc3, 
+    0x58, 0x55, 0x8a, 0xcc, 0xe6, 0x14, 0x52, 0x45, 0xe2, 0xc9, 
+    0xff, 0x4a, 0x72, 0x7e, 0x93, 0xd8, 0x92, 0xed, 0x67, 0x57, 
+    0xe7, 0x12, 0x79, 0x62, 0x6f, 0x12, 0x5d, 0x07, 0x6e, 0x08, 
+    0x72, 0x1b, 0xb2, 0xc9, 0xfd, 0x66, 0x84, 0x5d, 0x9d, 0xfc, 
+    0xee, 0x14, 0x52, 0x47, 0xe2, 0xcd, 0xef, 0x08, 0x12, 0x7e, 
+    0x93, 0xd8, 0x86, 0xed, 0x67, 0x57, 0xe7, 0x02, 0x7d, 0x6a, 
+    0x6f, 0x12, 0x5d, 0x07, 0x62, 0x0a, 0x72, 0x5b, 0xb3, 0xc9, 
+    0xdd, 0xe6, 0x84, 0x5d, 0x9d, 0xfc, 0x11, 0xeb, 0xad, 0xb8, 
+    0x1d, 0x32, 0x10, 0xf7, 0xed, 0x81, 0x6c, 0x27, 0x79, 0x12, 
+    0x98, 0xa8, 0x18, 0xfd, 0x82, 0x95, 0x90, 0xed, 0xa2, 0xf8, 
+    0x9d, 0xf5, 0x8d, 0xa4, 0x4c, 0x36, 0x22, 0x19, 0x7b, 0xa2, 
+    0x62, 0x03, 0x09, 0x8c, 0xf2, 0xb7, 0x04, 0x41, 0x19, 0xad, 
+    0x82, 0x1f, 0x29, 0x91, 0x1e, 0x0e, 0xe5, 0x78, 0xdd, 0x58, 
+    0x70, 0x01, 0x11, 0x06, 0x6d, 0xb0, 0x5f, 0x14, 0x3a, 0x22, 
+    0xd8, 0xb0, 0x0b, 0xcc, 0x82, 0x4d, 0x2b, 0x54, 0x19, 0xd5, 
+    0xd7, 0x6f, 0xc1, 0x0f, 0x01, 0x16, 0x15, 0x8f, 0x49, 0x5f, 
+    0xe6, 0x26, 0xd4, 0x98, 0xe0, 0x08, 0x48, 0x18, 0x36, 0x53, 
+    0x8d, 0x1a, 0x07, 0xb6, 0xe7, 0x07, 0x71, 0x81, 0x19, 0x82, 
+    0xf3, 0xe7, 0x1e, 0x49, 0xd3, 0x19, 0x1e, 0xac, 0x91, 0x7e, 
+    0x98, 0x24, 0x8e, 0xcc, 0x61, 0x80, 0xa4, 0x9d, 0xd5, 0x0f, 
+    0x1f, 0x53, 0xaa, 0xad, 0xce, 0x22, 0x8a, 0xf6, 0x0d, 0x93, 
+    0xcc, 0x21, 0xf2, 0x61, 0xd8, 0x4a, 0x17, 0x84, 0xcd, 0x19, 
+    0xd1, 0x3a, 0xd0, 0x3a, 0xd0, 0x45, 0x32, 0x43, 0x6b, 0x01, 
+    0x78, 0xf7, 0xa4, 0xe7, 0x05, 0x12, 0xd2, 0x2a, 0xd4, 0x80, 
+    0xc5, 0x77, 0x10, 0xbd, 0x15, 0xe7, 0x0c, 0x68, 0x78, 0xef, 
+    0x64, 0x65, 0x8e, 0x02, 0x02, 0xf0, 0xcb, 0x02, 0xf4, 0x82, 
+    0x77, 0xda, 0x4e, 0x89, 0x23, 0xd2, 0x1d, 0xbd, 0x60, 0xc1, 
+    0x71, 0x94, 0xef, 0xbc, 0x2f, 0x13, 0x2b, 0xb4, 0x22, 0xbd, 
+    0xa6, 0xe8, 0xa3, 0x8c, 0xb9, 0x1c, 0x34, 0x43, 0x8b, 0x11, 
+    0x3e, 0xdd, 0x0d, 0x03, 0xd0, 0x31, 0x76, 0xfe, 0x00, 0x8c, 
+    0xa2, 0x3a, 0x7c, 0xc5, 0x45, 0x6d, 0x84, 0x52, 0xf7, 0x34, 
+    0xaa, 0x22, 0x03, 0x16, 0x23, 0xef, 0x8d, 0xee, 0xc1, 0x8d, 
+    0x20, 0x6c, 0x14, 0x43, 0x9d, 0x5c, 0xb6, 0xed, 0x49, 0x02, 
+    0xd0, 0xb3, 0x76, 0xfe, 0x06, 0x8c, 0xa2, 0x3a, 0x7c, 0xc5, 
+    0x44, 0xed, 0xc4, 0x52, 0xf7, 0x34, 0xaa, 0x02, 0x03, 0x16, 
+    0x23, 0xaf, 0x8d, 0xee, 0xc1, 0x8d, 0x30, 0x6c, 0x14, 0x43, 
+    0x8d, 0x5c, 0xb6, 0xdd, 0xb6, 0xfd, 0x2f, 0x4c, 0x89, 0x01, 
+    0xf9, 0x73, 0x5d, 0xc5, 0x83, 0x3a, 0xbb, 0x12, 0x3b, 0xad, 
+    0x08, 0xcb, 0x55, 0xfd, 0xfc, 0xe9, 0xdc, 0x50, 0x72, 0x11, 
+    0x3e, 0x72, 0xcf, 0x93, 0xeb, 0xbc, 0x72, 0xa3, 0x49, 0x22, 
+    0x02, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x03, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x03, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 
+    0x05, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x02, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x07, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 
+    0x06, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x03, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 
+    0x07, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x07, 0x00, 
+    0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 
+    0x03, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x02, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 
+    0x05, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 
+    0x03, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x02, 0x00, 
+    0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x05, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x05, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x04, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 
+    0x04, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x07, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 
+    0x02, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 
+    0x03, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x13, 0x21, 
+    0xa0, 0x85, 0xd1, 0x3e, 0xe6, 0x79, 0x69, 0x7e, 0x19, 0x19, 
+    0x83, 0xd5, 0x6f, 0x58, 0xdc, 0xa4, 0x1e, 0x7f, 0x8a, 0x1e, 
+    0x02, 0x9e, 0x23, 0x7f, 0xb8, 0x41, 0xdd, 0x32, 0xa9, 0x88, 
+    0x97, 0xd3, 0x7a, 0xec, 0x32, 0x78, 0x6d, 0x9e, 0x17, 0x3a, 
+    0x8e, 0x75, 0x79, 0xe0, 0x1d, 0x86, 0x7d, 0x17, 0x8a, 0xbc, 
+    0xca, 0x8f, 0x2b, 0xcd, 0x8e, 0x6b, 0x44, 0xcb, 0x8f, 0x5b, 
+    0xf6, 0xe5, 0xf2, 0xee, 0xda, 0xc0, 0x24, 0x2c, 0x72, 0xe9, 
+    0xff, 0x6c, 0xb3, 0x82, 0x1b, 0x50, 0x0b, 0xee, 0xa9, 0x50, 
+    0xc4, 0x0e, 0x5f, 0x9b, 0x76, 0xfb, 0x22, 0xe2, 0x55, 0x39, 
+    0x6e, 0xb5, 0xcc, 0xa3, 0x80, 0x96, 0x6b, 0xac, 0x63, 0x96, 
+    0x43, 0x9c, 0xe5, 0x49, 0xa9, 0x2e, 0xf6, 0x58, 0x09, 0x70, 
+    0xed, 0xb2, 0x89, 0x48, 0xc7, 0xed, 0x74, 0xaf, 0x43, 0x37, 
+    0x79, 0x89, 0x73, 0xe1, 0x86, 0x47, 0x4b, 0x7b, 0xed, 0xd7, 
+    0x9e, 0xe9, 0xaa, 0x68, 0xc6, 0xaa, 0xde, 0xbe, 0xcd, 0x33, 
+    0xaf, 0xa3, 0x6f, 0x18, 0x35, 0x3c, 0x8f, 0x0a, 0x1d, 0x14, 
+    0x92, 0x9b, 0xbe, 0x12, 0x8e, 0xc3, 0x13, 0x00, 0x7c, 0xbd, 
+    0x60, 0x5e, 0x45, 0x7d, 0xcd, 0x16, 0x77, 0xa5, 0xfd, 0xe1, 
+    0x78, 0xec, 0xb9, 0x9a, 0x06, 0xc3, 0x27, 0xaa, 0x6b, 0x9a, 
+    0x64, 0x37, 0x49, 0x70, 0x94, 0x50, 0xfd, 0x99, 0x34, 0xf8, 
+    0x8b, 0x72, 0x73, 0xd4, 0x50, 0xd0, 0x28, 0x5d, 0x05, 0xb1, 
+    0xd3, 0x16, 0x97, 0xe5, 0xf9, 0xa1, 0x30, 0x5d, 0xcb, 0x88, 
+    0xe7, 0x40, 0x20, 0xe9, 0x6b, 0x98, 0x24, 0x34, 0x49, 0x72, 
+    0x9d, 0x50, 0xff, 0x99, 0x34, 0x78, 0x8b, 0x72, 0x53, 0xd0, 
+    0x50, 0xd1, 0x28, 0x5c, 0x05, 0x39, 0xc7, 0x16, 0xb7, 0xe5, 
+    0xf9, 0xe1, 0x78, 0x59, 0xcb, 0x88, 0xe7, 0x41, 0x22, 0xe9, 
+    0x94, 0x67, 0xdb, 0xcb, 0xb6, 0x8d, 0x62, 0xaf, 0x00, 0x66, 
+    0xcb, 0x87, 0x74, 0x8d, 0xac, 0x2f, 0xaf, 0x2e, 0xd7, 0xa3, 
+    0xfa, 0xc6, 0x38, 0xe9, 0x48, 0x1a, 0x06, 0x1e, 0x87, 0xa6, 
+    0x34, 0x77, 0x18, 0xbe, 0xdd, 0x16, 0xaa, 0xc0, 0x05, 0xc9, 
+    0x74, 0xd5, 0x2a, 0xb9, 0x7d, 0xea, 0xa1, 0xcc, 0xd2, 0x6e, 
+    0xb0, 0xd3, 0xb4, 0x15, 0x6f, 0x7f, 0x3c, 0xba, 0x30, 0xe4, 
+    0x11, 0x5e, 0x15, 0xc4, 0x0d, 0x3e, 0x65, 0x16, 0x58, 0x71, 
+    0x37, 0xcb, 0x92, 0x21, 0x23, 0xbf, 0xef, 0x18, 0x5b, 0x32, 
+    0x38, 0xeb, 0x27, 0x2c, 0x72, 0xc7, 0xa0, 0x39, 0x3a, 0x91, 
+    0x47, 0x91, 0xf7, 0xb3, 0xd4, 0x3f, 0x2e, 0x54, 0xa1, 0x46, 
+    0xcd, 0x3b, 0x34, 0x6f, 0xd9, 0x7a, 0x64, 0xf0, 0xb9, 0x87, 
+    0x09, 0xd8, 0x3d, 0x01, 0xc9, 0x36, 0x67, 0x2b, 0x49, 0xe2, 
+    0xf5, 0xf9, 0xb0, 0xdc, 0xb8, 0x69, 0xda, 0x93, 0xe0, 0x3c, 
+    0xd0, 0xaa, 0xc2, 0x9b, 0xea, 0x7e, 0xb7, 0x64, 0x46, 0xae, 
+    0x1a, 0x62, 0x31, 0x33, 0xc7, 0x53, 0xbb, 0x0d, 0x8d, 0x89, 
+    0x59, 0xda, 0xb5, 0xba, 0x9b, 0xd9, 0xda, 0x99, 0x48, 0x96, 
+    0x00, 0x1d, 0x53, 0x48, 0x77, 0xe4, 0x58, 0x39, 0xad, 0xa9, 
+    0x43, 0xe6, 0x5f, 0xc7, 0x88, 0xd8, 0x67, 0x34, 0x00, 0x22, 
+    0x8a, 0x55, 0xeb, 0x31, 0x8d, 0xb9, 0x2d, 0x31, 0x5d, 0x6b, 
+    0xc2, 0x7a, 0x38, 0x1d, 0xfb, 0x98, 0x13, 0x67, 0xe2, 0x7d, 
+    0x78, 0xba, 0x16, 0xab, 0xfb, 0x6e, 0x5f, 0xb7, 0xf8, 0xd4, 
+    0x4f, 0x59, 0x1b, 0x16, 0xac, 0xbf, 0xcf, 0x59, 0xb0, 0x3b, 
+    0x3d, 0xd9, 0x43, 0x70, 0xd8, 0x2a, 0x8b, 0xf1, 0x5f, 0x1c, 
+    0xf4, 0xdd, 0x71, 0x09, 0xae, 0x6d, 0x81, 0x1e, 0xe7, 0x48, 
+    0xa9, 0xd2, 0x7d, 0xbe, 0x78, 0xa0, 0x5e, 0x99, 0x8b, 0x17, 
+    0x26, 0x00, 0x8b, 0x59, 0xb0, 0x3b, 0x3d, 0xd9, 0x43, 0x70, 
+    0xdc, 0x2a, 0x8b, 0xf1, 0x5b, 0x1d, 0xf0, 0xdc, 0x31, 0x09, 
+    0xae, 0x4d, 0xa1, 0x3a, 0xf6, 0xe8, 0xa9, 0xd2, 0x7d, 0xbe, 
+    0x78, 0xa4, 0x5e, 0x99, 0x0b, 0x16, 0x24, 0x20, 0x74, 0xa6, 
+    0x4f, 0xc4, 0xc2, 0x26, 0xbc, 0x8f, 0x23, 0xd5, 0x74, 0x0e, 
+    0xa4, 0xe2, 0x0f, 0x23, 0xce, 0xf6, 0x51, 0xb2, 0x5e, 0xc5, 
+    0x09, 0x17, 0x56, 0x2d, 0x82, 0x41, 0x87, 0x5b, 0xa1, 0x66, 
+    0xf4, 0xe9, 0xdb, 0xdf, 0x07, 0x00, 0x00, 0x00, 0x03, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x04, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x03, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x03, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 
+    0x07, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 
+    0x06, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x03, 0x00, 
+    0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 
+    0x03, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x06, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 
+    0x04, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x07, 0x00, 
+    0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 
+    0x05, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x07, 0x00, 
+    0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x07, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x06, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x04, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 
+    0x06, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x03, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 
+    0x03, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x02, 0x00, 
+    0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 
+    0x03, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 
+    0x03, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x02, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 
+    0x06, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x04, 0x00, 
+    0x00, 0x00, 0x5c, 0x2d, 0xcc, 0xff, 0x03, 0xa7, 0x5b, 0x18, 
+    0x1b, 0x43, 0xa3, 0x52, 0x04, 0x96, 0x90, 0x8a, 0x25, 0x56, 
+    0xa2, 0x7d, 0x61, 0x0f, 0xc2, 0x83, 0x7d, 0x32, 0x9b, 0x3f, 
+    0x94, 0xb8, 0xdc, 0x89, 0x7c, 0x5a, 0x75, 0xcd, 0x10, 0xef, 
+    0x9b, 0xb9, 0xe6, 0x46, 0x09, 0x92, 0x7c, 0xa0, 0xe0, 0x5e, 
+    0xf0, 0x61, 0xab, 0xc8, 0xe2, 0x41, 0x04, 0x7e, 0x2b, 0x26, 
+    0x7a, 0x33, 0xa2, 0x33, 0xa8, 0x2d, 0xe5, 0x55, 0x5b, 0x1e, 
+    0xfe, 0x1b, 0xd2, 0xdd, 0xde, 0x3f, 0x98, 0xcd, 0x70, 0x1f, 
+    0x62, 0x03, 0x94, 0xd6, 0xe0, 0x30, 0xd3, 0xbd, 0x5d, 0xdd, 
+    0x25, 0xc1, 0x98, 0x12, 0x39, 0xf6, 0x44, 0x2c, 0xdd, 0x42, 
+    0x89, 0x2f, 0x8d, 0xc9, 0x2b, 0x6f, 0x13, 0xf9, 0x9b, 0x65, 
+    0xdf, 0x3c, 0x9d, 0x50, 0xe5, 0x3f, 0x5e, 0x51, 0xa5, 0x4b, 
+    0x5e, 0x02, 0x35, 0xcc, 0x73, 0xf4, 0x91, 0x8f, 0x71, 0x8c, 
+    0xe0, 0x9c, 0xb6, 0xc4, 0xa9, 0xde, 0x37, 0x3c, 0x8c, 0x56, 
+    0x40, 0x9f, 0xfc, 0x07, 0x2a, 0xb6, 0x81, 0xf3, 0x2c, 0x02, 
+    0x7c, 0x1f, 0x60, 0xfe, 0x96, 0x7f, 0x2d, 0xa4, 0x55, 0xae, 
+    0xc6, 0xf1, 0xfb, 0x47, 0x22, 0x27, 0x0f, 0xf6, 0x40, 0xe5, 
+    0x37, 0x51, 0xaa, 0x27, 0x9a, 0xfd, 0xf4, 0x5b, 0x64, 0x6c, 
+    0x6c, 0x0e, 0x8b, 0x83, 0x9c, 0x43, 0xac, 0x0b, 0x7e, 0x71, 
+    0xb0, 0x79, 0x6a, 0xb3, 0x15, 0xfd, 0x4e, 0xe3, 0x10, 0x42, 
+    0xba, 0xae, 0xc1, 0xf6, 0x30, 0xc9, 0xf1, 0x3f, 0xaa, 0x16, 
+    0xcf, 0xb9, 0x90, 0x0b, 0x71, 0x31, 0x35, 0x06, 0x89, 0x83, 
+    0x9c, 0x43, 0x8c, 0x0f, 0x78, 0x79, 0xb0, 0x7b, 0x6a, 0xb2, 
+    0x15, 0xaf, 0x46, 0xf3, 0xd0, 0x43, 0x3a, 0xae, 0x41, 0xf6, 
+    0x30, 0xc1, 0xf1, 0x77, 0xaa, 0x36, 0xcf, 0xb9, 0x90, 0x1b, 
+    0x71, 0x31, 0x3d, 0x0e, 0x76, 0x7c, 0x63, 0xbc, 0x53, 0xf0, 
+    0x87, 0x86, 0x4f, 0x84, 0x95, 0x4d, 0xea, 0x50, 0xb9, 0x0c, 
+    0x2f, 0xbc, 0xc5, 0x51, 0xbe, 0x09, 0xcf, 0x3e, 0x0e, 0x88, 
+    0x55, 0xc9, 0x30, 0x46, 0x6f, 0xe4, 0x8e, 0xce, 0xc2, 0xf1, 
+    0x21, 0x92, 0xb3, 0x12, 0x7a, 0xad, 0x8c, 0x5f, 0x0c, 0x90, 
+    0x8f, 0x40, 0xb9, 0x15, 0xf8, 0x7a, 0x36, 0x2d, 0x59, 0xb5, 
+    0xfb, 0xee, 0xa9, 0x1c, 0xa1, 0x03, 0xea, 0xf6, 0x9d, 0xa4, 
+    0x79, 0x8a, 0xb3, 0x20, 0x50, 0x1c, 0x68, 0x85, 0x7d, 0x3d, 
+    0x84, 0x94, 0x21, 0x1b, 0x3a, 0x20, 0xf3, 0xcd, 0x71, 0x95, 
+    0x73, 0xa7, 0x78, 0x09, 0x52, 0x9a, 0x7f, 0x22, 0x1f, 0xbf, 
+    0x85, 0x14, 0xe2, 0x28, 0xf2, 0xf7, 0x67, 0x3a, 0x14, 0xc3, 
+    0x4f, 0xd0, 0x85, 0x79, 0x63, 0xdc, 0x80, 0x86, 0x53, 0x8f, 
+    0xf9, 0x2d, 0xc4, 0x85, 0x65, 0xdc, 0xe6, 0x95, 0x88, 0xc8, 
+    0x38, 0x98, 0xcc, 0x28, 0xfd, 0xac, 0x74, 0x14, 0x91, 0x96, 
+    0x30, 0x08, 0x26, 0xe8, 0x7f, 0xe8, 0xc6, 0x97, 0xb7, 0xa7, 
+    0x7e, 0x07, 0xca, 0x63, 0xb1, 0x71, 0x11, 0xb6, 0xe4, 0x27, 
+    0xa5, 0x5e, 0x69, 0x25, 0x56, 0x72, 0xcd, 0x6b, 0xfc, 0xef, 
+    0xa0, 0x7a, 0x13, 0xb6, 0x9e, 0xdd, 0xf4, 0x68, 0xd3, 0x4c, 
+    0xec, 0x73, 0x03, 0x99, 0x3f, 0x05, 0x6c, 0xed, 0x71, 0xe2, 
+    0x8a, 0x69, 0x9f, 0x92, 0xe4, 0xa1, 0xbe, 0x2e, 0x8e, 0xc0, 
+    0xc8, 0x1f, 0xf9, 0xe0, 0x7f, 0x17, 0xe3, 0x15, 0x97, 0x28, 
+    0x53, 0x65, 0x60, 0x98, 0xd0, 0x32, 0xe2, 0x78, 0x78, 0x7b, 
+    0xfe, 0xcd, 0x66, 0xe5, 0xd8, 0x87, 0xc0, 0xf5, 0xd1, 0x52, 
+    0xe0, 0x26, 0x91, 0x1a, 0x21, 0xb5, 0x4a, 0x1a, 0x74, 0xaa, 
+    0x9c, 0x56, 0x03, 0x55, 0x17, 0x1c, 0x9b, 0xc6, 0x70, 0x58, 
+    0xd9, 0xc3, 0xac, 0x50, 0xe2, 0x3a, 0xbf, 0x8d, 0x6e, 0xe5, 
+    0xd8, 0xc7, 0x80, 0xe5, 0xd1, 0x52, 0xe0, 0x27, 0xb0, 0x1a, 
+    0x25, 0xa5, 0x08, 0x1a, 0x7c, 0xe8, 0xdc, 0x56, 0x23, 0x55, 
+    0x17, 0x1c, 0x9b, 0x46, 0x70, 0x58, 0xd9, 0xc3, 0xac, 0x70, 
+    0xf2, 0x3b, 0x40, 0x72, 0x91, 0x1a, 0x27, 0x38, 0x7f, 0x1a, 
+    0x2e, 0xad, 0x1f, 0xd8, 0x4f, 0xe5, 0xda, 0x5a, 0xf7, 0xe5, 
+    0x83, 0x17, 0x23, 0xa9, 0xdc, 0xaa, 0xe8, 0xe3, 0x64, 0xb9, 
+    0x8f, 0xa7, 0x26, 0x3c, 0x53, 0x8f, 0x0d, 0xc4, 0x06, 0x00, 
+    0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x06, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x07, 0x00, 
+    0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x03, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 
+    0x05, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x06, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 
+    0x06, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x04, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x02, 0x00, 
+    0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x02, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x06, 0x00, 
+    0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x05, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x02, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x04, 0x00, 
+    0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x05, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x03, 0x00, 
+    0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x03, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 
+    0x03, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x02, 0x00, 
+    0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 
+    0x06, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x07, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0xac, 0x1e, 0xa9, 0xe7, 
+    0x99, 0xc5, 0x06, 0x78, 0x7f, 0x8d, 0xe2, 0xef, 0xa4, 0x88, 
+    0xdb, 0x2a, 0x8e, 0xcd, 0xc2, 0x31, 0x58, 0xac, 0xa0, 0x96, 
+    0xb7, 0xd2, 0x12, 0x40, 0x46, 0x86, 0xcd, 0xc6, 0x51, 0x6d, 
+    0xd0, 0x52, 0x73, 0x0b, 0xff, 0x9f, 0x07, 0x0a, 0xaa, 0x6b, 
+    0x58, 0xbd, 0xba, 0xb2, 0x29, 0x62, 0x87, 0xaa, 0xa3, 0x5c, 
+    0x79, 0x7d, 0x73, 0x97, 0x50, 0x7c, 0x3a, 0x05, 0x67, 0x50, 
+    0x8d, 0x6a, 0xa1, 0xfc, 0xc0, 0x92, 0x85, 0x19, 0xbb, 0xac, 
+    0x3b, 0x85, 0x31, 0xe1, 0x53, 0x7c, 0x46, 0xdc, 0xca, 0xe3, 
+    0x77, 0x89, 0xcf, 0xa7, 0x31, 0x02, 0x5b, 0x50, 0xbd, 0x68, 
+    0x22, 0xc0, 0x3d, 0x0d, 0xd4, 0x76, 0x55, 0x5c, 0x9a, 0xd0, 
+    0x9d, 0x7a, 0x6b, 0xf7, 0x7f, 0x7e, 0xfd, 0xaf, 0xff, 0xd8, 
+    0xeb, 0x25, 0x26, 0xe8, 0xf6, 0x4b, 0xf4, 0x2c, 0x96, 0xf7, 
+    0xe1, 0xb9, 0xac, 0xe4, 0xaa, 0x17, 0x7a, 0x3f, 0x0e, 0x6a, 
+    0x2c, 0x3b, 0xfe, 0x37, 0xc6, 0xc5, 0xfc, 0x19, 0xe8, 0x76, 
+    0xf1, 0xf6, 0xbc, 0x8c, 0xee, 0x52, 0x7c, 0x09, 0xdb, 0xa4, 
+    0x04, 0x0f, 0xb5, 0x0b, 0x8d, 0x19, 0x0b, 0x0b, 0xd7, 0xa7, 
+    0xf0, 0x2e, 0x02, 0x0e, 0x1e, 0x4a, 0x47, 0x4d, 0x36, 0xc6, 
+    0x29, 0xbc, 0x9f, 0xd7, 0x16, 0x73, 0xd3, 0x34, 0x9c, 0xcd, 
+    0x7d, 0x53, 0xfb, 0x4b, 0x0f, 0x04, 0x0c, 0x07, 0x73, 0x2b, 
+    0x95, 0x95, 0xa3, 0x0f, 0xde, 0x90, 0xf6, 0x3f, 0x32, 0x6e, 
+    0x17, 0x5f, 0x8c, 0xed, 0x38, 0x06, 0x39, 0x0d, 0x8d, 0x7c, 
+    0x85, 0xf1, 0xf3, 0x34, 0xbc, 0xcd, 0x7d, 0x52, 0xf9, 0x4b, 
+    0x0f, 0x84, 0x04, 0x07, 0x73, 0x2b, 0x95, 0x95, 0xa3, 0x0b, 
+    0xde, 0x90, 0xf6, 0x3f, 0x32, 0x6e, 0x1f, 0x4b, 0x8c, 0x6d, 
+    0x3c, 0x06, 0x2d, 0x0d, 0x9d, 0x7e, 0x85, 0x73, 0x0c, 0xcb, 
+    0x43, 0x32, 0x82, 0xad, 0x06, 0xb4, 0xf0, 0x7b, 0xfb, 0xf8, 
+    0x8c, 0xd4, 0x6a, 0x6a, 0x5c, 0xf4, 0x21, 0x6f, 0x09, 0xc0, 
+    0xcd, 0x91, 0xe0, 0xb4, 0x73, 0x92, 0xc3, 0xf9, 0xd2, 0xf2, 
+    0x62, 0x81, 0x7a, 0x8c, 0x6b, 0x8a, 0xee, 0xf0, 0x45, 0x3f, 
+    0xeb, 0xf5, 0x64, 0xb0, 0xd1, 0x54, 0x04, 0xe7, 0x16, 0x02, 
+    0x74, 0xa4, 0x8b, 0xbd, 0x35, 0xa4, 0x33, 0xa1, 0x00, 0x39, 
+    0x16, 0x16, 0x2a, 0x0f, 0xfa, 0xb6, 0x95, 0x11, 0x71, 0xb5, 
+    0x34, 0x1b, 0xbe, 0x28, 0x96, 0xa7, 0xe9, 0x64, 0x0c, 0x04, 
+    0xf2, 0xf4, 0xd6, 0x0e, 0xe7, 0xdd, 0x59, 0xef, 0x79, 0x29, 
+    0x7c, 0x68, 0x49, 0xf4, 0x66, 0xde, 0x44, 0x93, 0x33, 0xab, 
+    0xd1, 0x05, 0x83, 0x6e, 0xb0, 0x84, 0xc3, 0xa6, 0xc3, 0x84, 
+    0x5f, 0xdd, 0xee, 0x1a, 0x65, 0x57, 0xb9, 0xee, 0xfe, 0x82, 
+    0xf4, 0xd8, 0x8e, 0x8b, 0xbc, 0x84, 0x5e, 0xa1, 0x0c, 0x4f, 
+    0x40, 0x81, 0xda, 0x2d, 0xf0, 0xd3, 0x1a, 0x87, 0xff, 0x7e, 
+    0xd0, 0x9f, 0x39, 0xbc, 0x9f, 0x0f, 0x05, 0x06, 0x6e, 0x5b, 
+    0xc7, 0x4a, 0xfa, 0x89, 0xc3, 0xcf, 0x3d, 0xe2, 0x5f, 0x4f, 
+    0x80, 0x15, 0x93, 0xa5, 0x22, 0x12, 0x61, 0xae, 0x57, 0xe2, 
+    0x02, 0x40, 0x07, 0xd4, 0xb8, 0xf9, 0x33, 0x47, 0x1a, 0x42, 
+    0x4e, 0x84, 0xa2, 0xb0, 0x86, 0xf4, 0x07, 0x96, 0xf5, 0x37, 
+    0x84, 0x37, 0x2c, 0xa6, 0xc5, 0x0d, 0xf1, 0xf6, 0x56, 0x59, 
+    0x3a, 0xe6, 0xf6, 0x73, 0xf1, 0x1f, 0x9d, 0xe1, 0x81, 0x1d, 
+    0x36, 0x42, 0x85, 0x59, 0x2a, 0xee, 0x73, 0xc0, 0xa1, 0x8c, 
+    0xd7, 0x50, 0x0e, 0x1e, 0x5a, 0x57, 0x8c, 0x33, 0xbd, 0x0c, 
+    0x1d, 0x0b, 0x03, 0xc4, 0x37, 0xfd, 0xbb, 0x36, 0xea, 0x67, 
+    0xd9, 0x7c, 0x9f, 0x02, 0x11, 0x04, 0x37, 0x47, 0x89, 0x01, 
+    0x2a, 0xe6, 0x5b, 0x80, 0xa1, 0x84, 0xd6, 0xd0, 0x0e, 0x1e, 
+    0xf8, 0x57, 0x8c, 0x37, 0xbd, 0x0c, 0x5d, 0x0f, 0x01, 0xc4, 
+    0x17, 0xf9, 0xbb, 0x36, 0xe2, 0x67, 0xd9, 0x74, 0x9d, 0x43, 
+    0x11, 0x04, 0x37, 0x4e, 0x99, 0x01, 0xd5, 0x19, 0xa4, 0x7f, 
+    0x5e, 0x7b, 0x29, 0x2f, 0xf1, 0xe1, 0x07, 0xa8, 0x73, 0xc8, 
+    0x42, 0xf3, 0xa2, 0xf0, 0xfe, 0x3b, 0xe8, 0x06, 0x44, 0xc9, 
+    0x1d, 0x98, 0x26, 0x8b, 0x62, 0xbc, 0xee, 0xfb, 0xc8, 0xb1, 
+    0x66, 0xfe, 0x04, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 
+    0x07, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x02, 0x00, 
+    0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 
+    0x02, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x04, 0x00, 
+    0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x03, 0x00, 
+    0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x06, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x02, 0x00, 
+    0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x04, 0x00, 
+    0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 
+    0x06, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 
+    0x02, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x06, 0x00, 
+    0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 
+    0x06, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x07, 0x00, 
+    0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 
+    0x03, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 
+    0x07, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x02, 0x00, 
+    0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 
+    0x06, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 
+    0x02, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x03, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x03, 0xad, 0x91, 0xce, 0xbd, 0x79, 0xce, 0x17, 0xea, 0x0b, 
+    0x09, 0xf1, 0xf1, 0xb6, 0xab, 0x71, 0x10, 0xb8, 0xb7, 0xae, 
+    0xf9, 0x51, 0x02, 0x8c, 0xc5, 0x3d, 0xd7, 0x67, 0x02, 0x4f, 
+    0xcd, 0x2d, 0x06, 0xdc, 0x9d, 0x63, 0x59, 0xbf, 0xe1, 0xfe, 
+    0xab, 0x2b, 0x85, 0xf5, 0x90, 0x24, 0x89, 0xe1, 0xcb, 0xc9, 
+    0x91, 0x08, 0x98, 0x58, 0x12, 0xb5, 0xd8, 0xb7, 0x89, 0x14, 
+    0x23, 0x99, 0x5f, 0x4f, 0x46, 0x14, 0x08, 0xc5, 0xa2, 0xea, 
+    0xf8, 0xf2, 0x7d, 0xd5, 0xc3, 0x73, 0x7a, 0x4e, 0xe0, 0x46, 
+    0xaf, 0x98, 0xdd, 0xe2, 0x4a, 0x2d, 0x4f, 0xf5, 0xe6, 0x9b, 
+    0xfe, 0x13, 0xae, 0xaa, 0x71, 0x9e, 0xdb, 0xe6, 0xcd, 0xe6, 
+    0x63, 0x45, 0xc5, 0x88, 0x2f, 0x02, 0x72, 0x95, 0xb3, 0x33, 
+    0xa2, 0xd2, 0x08, 0xa0, 0x92, 0x82, 0x4f, 0x0d, 0x97, 0x21, 
+    0xa2, 0x1f, 0x8d, 0x14, 0x1a, 0x77, 0x30, 0x76, 0x5e, 0x62, 
+    0x36, 0x39, 0x94, 0xd3, 0xa0, 0x8a, 0x4a, 0xe4, 0x31, 0x33, 
+    0xe9, 0xac, 0x5b, 0xb3, 0x2e, 0x9d, 0xce, 0x19, 0xfc, 0xcf, 
+    0x7f, 0xd6, 0xd6, 0x0b, 0x17, 0x3d, 0x90, 0xea, 0x2b, 0x0b, 
+    0x00, 0x4f, 0x1b, 0x2e, 0x8c, 0x2f, 0x20, 0xdd, 0x6f, 0xf1, 
+    0x08, 0x2c, 0xd1, 0x87, 0x8e, 0x5a, 0xab, 0x02, 0xa1, 0x25, 
+    0x31, 0x1e, 0x8d, 0x75, 0xbd, 0x6d, 0x72, 0xd6, 0x5a, 0x0f, 
+    0x51, 0x0d, 0xf0, 0x83, 0xef, 0xb1, 0x3a, 0x83, 0x1a, 0x2f, 
+    0x1e, 0x2a, 0x62, 0x4d, 0x5f, 0xdf, 0x48, 0xb2, 0x34, 0xdf, 
+    0x14, 0x14, 0x68, 0x6a, 0xe2, 0x87, 0x31, 0x9f, 0xcd, 0x71, 
+    0xbd, 0x6d, 0x72, 0xd6, 0x5e, 0x0d, 0x55, 0x05, 0x90, 0x83, 
+    0xaf, 0xb1, 0x2a, 0x0b, 0x1a, 0x2f, 0x1e, 0x2a, 0x22, 0x5d, 
+    0x7f, 0xff, 0x48, 0xb2, 0x74, 0xc7, 0x14, 0x10, 0x68, 0x0a, 
+    0xe2, 0x87, 0xce, 0x60, 0x32, 0x8e, 0x42, 0x92, 0x8d, 0x29, 
+    0xa1, 0xf2, 0xaa, 0xfa, 0x6f, 0x7c, 0x50, 0x4e, 0xd5, 0xf4, 
+    0xe5, 0xd0, 0xe1, 0xd5, 0xdd, 0xa2, 0x80, 0x00, 0xb7, 0x4d, 
+    0x8b, 0x38, 0xeb, 0xef, 0x97, 0xf5, 0x1d, 0x78, 0x6f, 0x53, 
+    0x3a, 0xfe, 0x44, 0x3f, 0x84, 0x8a, 0x82, 0xf2, 0x95, 0x36, 
+    0x4d, 0xa7, 0x56, 0x12, 0x01, 0xa4, 0x57, 0x40, 0xf6, 0x1f, 
+    0x88, 0x33, 0xae, 0x10, 0xec, 0xe2, 0x06, 0x5d, 0x0f, 0xb9, 
+    0x5d, 0xfd, 0x04, 0x19, 0xd4, 0x9b, 0x6f, 0xf8, 0x33, 0x60, 
+    0xec, 0xdb, 0xd9, 0x57, 0xe0, 0x8f, 0xe5, 0xc4, 0x71, 0x50, 
+    0x9f, 0x62, 0x54, 0xd3, 0xda, 0x6c, 0xdd, 0x5f, 0x63, 0x39, 
+    0xff, 0x58, 0xde, 0xa9, 0xbf, 0xfe, 0x0b, 0x23, 0x0e, 0xb9, 
+    0xd3, 0x8c, 0x3d, 0xf9, 0x78, 0xb1, 0xef, 0xee, 0xb9, 0x8f, 
+    0xce, 0x78, 0x29, 0x03, 0xe4, 0x8b, 0x0c, 0x7c, 0x3e, 0x25, 
+    0x96, 0xe5, 0x4e, 0x31, 0xae, 0x18, 0x76, 0xc1, 0xfe, 0x84, 
+    0xe7, 0x70, 0xfb, 0x2b, 0x89, 0x15, 0x6d, 0xef, 0x28, 0xdd, 
+    0xd1, 0x58, 0x02, 0xb0, 0x80, 0x39, 0xb5, 0xe9, 0xa6, 0xd0, 
+    0x7e, 0x27, 0x54, 0xe9, 0xbd, 0xf6, 0xf8, 0x8f, 0x80, 0x29, 
+    0x88, 0x22, 0xda, 0xde, 0xa7, 0xa4, 0xd2, 0x63, 0x46, 0xca, 
+    0xd0, 0x95, 0x8e, 0xf3, 0x67, 0xfc, 0x7c, 0xd3, 0x97, 0x5a, 
+    0x7c, 0x3b, 0xf2, 0xa9, 0x8a, 0x34, 0x64, 0xf1, 0xa0, 0x3c, 
+    0x66, 0x43, 0x1b, 0xba, 0xb0, 0x8a, 0x1c, 0x68, 0x53, 0x43, 
+    0xd0, 0x24, 0x25, 0xcc, 0x42, 0x72, 0x49, 0x54, 0xd0, 0x77, 
+    0xef, 0xfb, 0xc6, 0xd1, 0x93, 0xf0, 0x72, 0x39, 0x1a, 0xab, 
+    0xb8, 0x74, 0x65, 0x4c, 0x8e, 0x6e, 0x62, 0xe7, 0xd2, 0xe6, 
+    0x78, 0x2a, 0x88, 0x31, 0x03, 0x62, 0x66, 0x3f, 0x65, 0x68, 
+    0x42, 0xf3, 0x1a, 0x50, 0xc0, 0x77, 0xef, 0xfb, 0xce, 0xd1, 
+    0x93, 0xf0, 0x70, 0x39, 0x12, 0xa1, 0xb8, 0x74, 0x65, 0x45, 
+    0x86, 0x6e, 0x66, 0xe7, 0xd2, 0xe6, 0x78, 0x2a, 0x88, 0x31, 
+    0x53, 0x62, 0x62, 0x34, 0x65, 0x68, 0x42, 0xf2, 0x1a, 0x54, 
+    0x3f, 0x88, 0x10, 0x04, 0x31, 0x2e, 0x6c, 0x0f, 0x8f, 0xc6, 
+    0xed, 0x56, 0x47, 0x8b, 0x9a, 0xba, 0x79, 0x91, 0x99, 0x18, 
+    0x2d, 0x19, 0x87, 0xd5, 0x77, 0xce, 0xac, 0x9d, 0x9d, 0xcb, 
+    0x9a, 0x97, 0xbd, 0x0d, 0xe5, 0xab, 0x02, 0x00, 0x00, 0x00, 
+    0x05, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x04, 0x00, 
+    0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x03, 0x00, 
+    0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 
+    0x03, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 
+    0x07, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 
+    0x05, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x03, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x02, 0x00, 
+    0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 
+    0x06, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x07, 0x00, 
+    0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x04, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x02, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x03, 0x00, 
+    0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x06, 0x00, 
+    0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 
+    0x02, 0x00, 0x00, 0x00, 0xe5, 0x77, 0xae, 0xd6, 0x41, 0x1e, 
+    0x28, 0x48, 0x6f, 0x6d, 0x57, 0x6e, 0x48, 0x31, 0x03, 0x78, 
+    0xa6, 0x49, 0x91, 0x1a, 0xf2, 0x2b, 0x28, 0x7c, 0x2a, 0xa5, 
+    0x1c, 0xd1, 0x1e, 0x0f, 0xe7, 0x40, 0xcc, 0x52, 0xcc, 0xa3, 
+    0x90, 0x42, 0x99, 0xdc, 0xfe, 0xdd, 0xff, 0xea, 0x35, 0xf8, 
+    0x60, 0xe0, 0x78, 0xf4, 0x7d, 0x92, 0x2d, 0xbc, 0xc2, 0x1a, 
+    0x63, 0x59, 0x06, 0xf6, 0x41, 0xdd, 0x9d, 0xb8, 0xcf, 0xfa, 
+    0xf0, 0xe5, 0xd6, 0x9c, 0x68, 0x52, 0x5d, 0x05, 0x43, 0x36, 
+    0xa1, 0x1c, 0x39, 0x73, 0xa9, 0xca, 0xb1, 0x3c, 0xbf, 0xca, 
+    0x0f, 0x18, 0xfd, 0xe4, 0x44, 0x5c, 0xf1, 0x8b, 0xd2, 0xef, 
+    0xe3, 0x83, 0xb1, 0xc4, 0xb8, 0xf4, 0x21, 0xe8, 0xab, 0x37, 
+    0x6b, 0x8f, 0xa7, 0x58, 0xe4, 0xca, 0x54, 0x8f, 0x8c, 0x4e, 
+    0x28, 0x23, 0x95, 0xf4, 0xa5, 0x5b, 0xb2, 0x29, 0x87, 0x06, 
+    0x30, 0xcd, 0x51, 0xad, 0x2f, 0xd2, 0xf6, 0xe7, 0x0b, 0xa9, 
+    0xc5, 0x57, 0xe8, 0x36, 0x5d, 0x2d, 0xe0, 0x18, 0x52, 0xfa, 
+    0x2e, 0x02, 0xf7, 0xf8, 0xd8, 0x76, 0x0e, 0x6e, 0x0d, 0x15, 
+    0xd4, 0x9c, 0x04, 0xc9, 0x50, 0xbd, 0xed, 0x69, 0x33, 0x41, 
+    0x67, 0xdd, 0xa2, 0xd9, 0x64, 0x20, 0xb1, 0x60, 0xf5, 0xb1, 
+    0xd1, 0xe3, 0x94, 0x86, 0x5b, 0xfb, 0x3f, 0x04, 0xfd, 0x7d, 
+    0x58, 0x72, 0x8f, 0x66, 0x87, 0x07, 0x08, 0x68, 0x60, 0x38, 
+    0x9f, 0xaf, 0xee, 0x6d, 0xf5, 0x11, 0xeb, 0xcb, 0xba, 0x51, 
+    0x31, 0x90, 0x8b, 0xb2, 0xd0, 0xbb, 0x57, 0x53, 0x34, 0xda, 
+    0x53, 0xfb, 0x3f, 0x06, 0xfd, 0x7d, 0x58, 0x72, 0x0f, 0x6e, 
+    0x85, 0x07, 0x08, 0x68, 0x20, 0x38, 0x97, 0xae, 0x6f, 0x6d, 
+    0xf1, 0x01, 0xe3, 0xca, 0xba, 0x51, 0x31, 0x90, 0x8d, 0xb2, 
+    0xd0, 0xbb, 0xd7, 0x53, 0xb4, 0x9a, 0xac, 0x04, 0xc0, 0xf9, 
+    0x02, 0x82, 0xa7, 0x8d, 0xf0, 0x91, 0x7a, 0xf8, 0xf7, 0x97, 
+    0xdf, 0xc7, 0x68, 0x51, 0x90, 0x92, 0x0e, 0xfe, 0x1c, 0x35, 
+    0x45, 0xae, 0xce, 0x6f, 0x72, 0x4d, 0x2f, 0x44, 0x28, 0xac, 
+    0x4b, 0x65, 0x01, 0x35, 0x1e, 0xf1, 0x97, 0xb5, 0x9d, 0xa3, 
+    0x4a, 0xe5, 0x7e, 0x81, 0x1a, 0x37, 0xca, 0x9d, 0xb3, 0x10, 
+    0x96, 0x34, 0x1f, 0xf7, 0x3d, 0x26, 0xb2, 0x7d, 0x02, 0x8c, 
+    0x75, 0xd7, 0xb6, 0x19, 0xaf, 0x81, 0xa9, 0x4d, 0xe9, 0xbc, 
+    0xe6, 0x04, 0x48, 0x74, 0x67, 0x72, 0xe0, 0x93, 0x99, 0x05, 
+    0x55, 0x3f, 0xdb, 0xeb, 0x87, 0x17, 0x04, 0xc1, 0x5b, 0xd6, 
+    0xb8, 0xd7, 0x7b, 0x7f, 0xf6, 0x26, 0xeb, 0xb4, 0x23, 0x40, 
+    0xe4, 0x25, 0x48, 0xfb, 0x44, 0xbd, 0xc0, 0x71, 0x9e, 0xe8, 
+    0x93, 0x50, 0x80, 0x98, 0xcd, 0x52, 0xf3, 0x62, 0x2a, 0x92, 
+    0x1e, 0x67, 0x79, 0x8e, 0x4b, 0xd4, 0x02, 0xab, 0xa0, 0x48, 
+    0xda, 0x69, 0xf2, 0x35, 0x1d, 0xda, 0x02, 0x39, 0xaf, 0x76, 
+    0xe9, 0xda, 0xfa, 0x22, 0x0b, 0xfb, 0x7e, 0x5b, 0xce, 0x7d, 
+    0x46, 0xc0, 0x59, 0x18, 0x71, 0x36, 0x34, 0xea, 0xcb, 0x5f, 
+    0x86, 0x0c, 0xea, 0xff, 0x9b, 0x47, 0xd2, 0xf5, 0x05, 0x21, 
+    0x73, 0xe2, 0xdf, 0x92, 0x58, 0x51, 0x2a, 0x8e, 0x69, 0xd7, 
+    0xa6, 0xac, 0x81, 0xb0, 0xa5, 0xe8, 0x62, 0x32, 0xc0, 0xbe, 
+    0x84, 0x09, 0xac, 0x84, 0x75, 0xbe, 0xbc, 0x9c, 0x67, 0x98, 
+    0x6c, 0xc0, 0xd6, 0xa1, 0x38, 0x74, 0xab, 0xcd, 0x1e, 0xcf, 
+    0x5a, 0x02, 0x60, 0x45, 0x81, 0x40, 0x8e, 0x6c, 0xdb, 0x1c, 
+    0x10, 0xea, 0x27, 0x32, 0xc0, 0x9e, 0x45, 0x13, 0x74, 0x43, 
+    0x0f, 0xbe, 0xa8, 0x9c, 0x62, 0x98, 0x6d, 0x06, 0xda, 0xe6, 
+    0xf9, 0x75, 0x0a, 0x4c, 0x5f, 0x87, 0xce, 0x02, 0x60, 0x0f, 
+    0xc1, 0x42, 0x8e, 0xec, 0xdb, 0x38, 0x94, 0xe8, 0x26, 0x30, 
+    0xc0, 0x9e, 0x45, 0x13, 0x74, 0x43, 0x4f, 0xbe, 0xa8, 0x9c, 
+    0x62, 0x98, 0x6c, 0x06, 0xda, 0xa4, 0xf9, 0x75, 0x0a, 0x4c, 
+    0x5f, 0x87, 0x4e, 0x02, 0x9f, 0xf0, 0x3e, 0xbd, 0x71, 0x13, 
+    0x24, 0xc7, 0x6b, 0x17, 0xd9, 0xcf, 0x3f, 0x61, 0xba, 0xec, 
+    0x8b, 0xbc, 0xb0, 0x41, 0x57, 0x63, 0x9d, 0x67, 0x93, 0xf9, 
+    0x25, 0x5b, 0x06, 0x8a, 0xf5, 0xb3, 0xa0, 0x78, 0xb1, 0xfd, 
+    0x05, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x07, 0x00, 
+    0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x03, 0x00, 
+    0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x03, 0x00, 
+    0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x03, 0x00, 
+    0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 
+    0x03, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 
+    0x02, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x07, 0x00, 
+    0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 
+    0x03, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 
+    0x02, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 
+    0x04, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x02, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x07, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x03, 0x00, 
+    0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 
+    0x04, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x02, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x04, 0x00, 
+    0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 
+    0x05, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x98, 0xf7, 
+    0x0d, 0xd2, 0x0d, 0x3f, 0x4e, 0x78, 0xf9, 0x4b, 0x87, 0xe3, 
+    0x44, 0xb1, 0x82, 0x58, 0x68, 0x8c, 0x3a, 0xc1, 0x10, 0x7a, 
+    0x7f, 0x83, 0xa0, 0x9e, 0x69, 0xf5, 0x5a, 0x88, 0x16, 0x72, 
+    0x68, 0x29, 0x6f, 0x44, 0x72, 0x7b, 0x1e, 0x93, 0x0b, 0x39, 
+    0x43, 0xfa, 0x25, 0xa6, 0x57, 0x55, 0x84, 0x08, 0xd5, 0x31, 
+    0x4a, 0xc5, 0x59, 0xdf, 0x99, 0x59, 0xc8, 0x61, 0x45, 0x5a, 
+    0xc7, 0x7d, 0xea, 0xdb, 0x36, 0x80, 0x54, 0xa7, 0x49, 0x5b, 
+    0xdc, 0x92, 0xca, 0x77, 0x48, 0x1d, 0x27, 0xd5, 0x99, 0xee, 
+    0x9a, 0x0c, 0x7e, 0xb4, 0x60, 0x0a, 0xaf, 0x17, 0x49, 0xa9, 
+    0x7e, 0xce, 0xab, 0x39, 0x39, 0x91, 0x87, 0x78, 0x8d, 0x8c, 
+    0x31, 0x02, 0x18, 0x8b, 0x54, 0xb9, 0x2f, 0x29, 0x3d, 0xcd, 
+    0x35, 0x13, 0xa8, 0x20, 0x88, 0xce, 0xff, 0x2e, 0x9c, 0xc1, 
+    0x34, 0xc0, 0x65, 0x1d, 0x61, 0xc4, 0x2c, 0xa9, 0x23, 0x93, 
+    0xa6, 0xf0, 0x6f, 0xad, 0x5d, 0xeb, 0x74, 0x27, 0xac, 0x64, 
+    0xfd, 0x1d, 0x88, 0x89, 0xf7, 0x86, 0x2a, 0x17, 0x5d, 0xf3, 
+    0x06, 0x1a, 0xa3, 0xa3, 0x28, 0x9b, 0xb9, 0xae, 0x36, 0x46, 
+    0xf1, 0x75, 0xa0, 0x4c, 0xa3, 0x1c, 0x7d, 0x33, 0x69, 0x23, 
+    0xc1, 0xe1, 0xb3, 0x23, 0xdf, 0x0a, 0xad, 0x54, 0x34, 0xcb, 
+    0xb5, 0x91, 0x6b, 0x93, 0xbd, 0x57, 0x84, 0x6b, 0x26, 0xf7, 
+    0xfc, 0x99, 0x78, 0x87, 0xb7, 0x2e, 0x63, 0xe6, 0xf4, 0xcc, 
+    0xa3, 0x10, 0x75, 0x1b, 0x67, 0x43, 0xc6, 0xe1, 0xa7, 0x32, 
+    0xfc, 0x10, 0xab, 0x5c, 0x24, 0x8b, 0xbf, 0x95, 0x6b, 0x93, 
+    0xbd, 0x57, 0x84, 0x6b, 0xa7, 0xa7, 0xfc, 0x99, 0x58, 0x8e, 
+    0xb7, 0x0e, 0x63, 0xe4, 0xf0, 0xcc, 0xa3, 0x10, 0x75, 0x13, 
+    0x67, 0x63, 0xc6, 0xe9, 0xb7, 0x32, 0xfc, 0x02, 0xab, 0x5c, 
+    0xdb, 0x74, 0x40, 0x6a, 0x94, 0x6c, 0x42, 0xa8, 0x7b, 0x94, 
+    0x58, 0x58, 0x03, 0x66, 0xa7, 0x71, 0x48, 0xf1, 0x9c, 0x1b, 
+    0x0f, 0x33, 0x5c, 0xef, 0x8a, 0xec, 0x98, 0x9c, 0x39, 0x16, 
+    0x48, 0xcd, 0x03, 0xfd, 0x54, 0xa3, 0x60, 0x88, 0xe3, 0x0b, 
+    0x99, 0x43, 0xc0, 0x55, 0x98, 0x74, 0xd2, 0x8d, 0xbb, 0x82, 
+    0x99, 0x0b, 0xa9, 0x1e, 0xbf, 0x1c, 0xff, 0x1a, 0x09, 0xaf, 
+    0x5b, 0xbf, 0xc9, 0x4d, 0x10, 0xaf, 0x90, 0x1a, 0x18, 0x41, 
+    0xc9, 0x59, 0xce, 0xb2, 0x76, 0x43, 0x5d, 0x59, 0xfa, 0x1f, 
+    0x4c, 0x22, 0xec, 0x98, 0x1f, 0x68, 0x39, 0xea, 0x11, 0x6b, 
+    0xd1, 0xd0, 0x88, 0x59, 0xa7, 0x16, 0xb7, 0x4e, 0x13, 0x18, 
+    0x1b, 0xc9, 0xc2, 0xb5, 0xb0, 0x1d, 0xde, 0x51, 0xf3, 0xca, 
+    0x83, 0x51, 0x42, 0xc2, 0x90, 0x1d, 0xd3, 0x5b, 0x58, 0x41, 
+    0x92, 0xea, 0x8c, 0x84, 0xc9, 0x9d, 0x2c, 0x15, 0x44, 0x7e, 
+    0x67, 0xb9, 0xc2, 0x9f, 0xac, 0x35, 0xe5, 0x21, 0xbc, 0x7a, 
+    0x97, 0xe3, 0x79, 0x47, 0x54, 0x97, 0x65, 0x06, 0x96, 0xfa, 
+    0x73, 0xd3, 0xca, 0x49, 0x74, 0x28, 0x9e, 0xb0, 0xc3, 0x4f, 
+    0x5a, 0xc1, 0x22, 0x8f, 0x93, 0x1d, 0xd5, 0x9c, 0x52, 0x01, 
+    0xce, 0x8f, 0x8e, 0x40, 0xb4, 0x0b, 0xe2, 0x48, 0x6c, 0x91, 
+    0x3e, 0xd8, 0x76, 0xe3, 0x21, 0x12, 0xc0, 0xfe, 0xd0, 0xe5, 
+    0xc4, 0x16, 0x23, 0x90, 0x89, 0x93, 0xcd, 0xc5, 0xe0, 0xf3, 
+    0x39, 0x44, 0x00, 0xe8, 0xe1, 0x19, 0xc1, 0xc3, 0x4e, 0xfc, 
+    0x22, 0xfe, 0x6f, 0x7c, 0xf6, 0x91, 0x49, 0xd9, 0x37, 0xd3, 
+    0x21, 0x70, 0xfa, 0x0a, 0x50, 0xfc, 0xe4, 0x16, 0x82, 0x90, 
+    0xb5, 0x9f, 0xc8, 0x41, 0x21, 0xbf, 0xbf, 0x04, 0x04, 0x79, 
+    0x61, 0x1c, 0x0f, 0xc3, 0xcd, 0xbd, 0xb2, 0xe5, 0xc2, 0x7a, 
+    0x31, 0x03, 0x49, 0xd1, 0x37, 0xd3, 0x21, 0x72, 0xf2, 0x0a, 
+    0x52, 0xfd, 0xe4, 0x16, 0x82, 0x90, 0x95, 0x9f, 0xc8, 0x41, 
+    0x21, 0xbf, 0xbf, 0x44, 0x04, 0xf9, 0x61, 0x1d, 0x0f, 0xc3, 
+    0xcd, 0xbd, 0xb2, 0xe5, 0xe2, 0x78, 0x35, 0x03, 0xb6, 0x2e, 
+    0xc8, 0x2c, 0xde, 0x8d, 0x0d, 0xf5, 0xad, 0x02, 0x1b, 0xe9, 
+    0x7d, 0x6f, 0x6a, 0x60, 0x37, 0xbe, 0xde, 0x40, 0x40, 0xbb, 
+    0xfb, 0x06, 0x9e, 0xe2, 0xf0, 0x3c, 0x32, 0x42, 0x4d, 0x1a, 
+    0x1d, 0x87, 0xca, 0xfc, 0x04, 0x00, 0x00, 0x00, 0x04, 0x00, 
+    0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x04, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x07, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 
+    0x04, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x07, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x07, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 
+    0x02, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x07, 0x00, 
+    0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 
+    0x06, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 
+    0x03, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x02, 0x00, 
+    0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x06, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x03, 0x00, 
+    0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 
+    0x02, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x03, 0x00, 
+    0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x03, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x06, 0x00, 
+    0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x05, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x07, 0x00, 
+    0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x03, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x04, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x93, 0x95, 0x7e, 0xa4, 0x96, 0xce, 0x6b, 0x84, 
+    0x5f, 0x8f, 0xd6, 0x62, 0x1c, 0x75, 0xd9, 0x0f, 0x4f, 0xe8, 
+    0x67, 0xbe, 0xf4, 0xa7, 0xd9, 0xf9, 0x69, 0x68, 0xff, 0x61, 
+    0x0f, 0x86, 0x93, 0x8c, 0xdf, 0x2e, 0xbd, 0xd4, 0x73, 0x04, 
+    0x68, 0x85, 0x2e, 0xe4, 0xff, 0x6c, 0x95, 0x04, 0xab, 0x8e, 
+    0x6d, 0xf6, 0x08, 0xdf, 0x90, 0xed, 0xfe, 0xf4, 0x44, 0x7d, 
+    0xed, 0x8c, 0xaf, 0x85, 0x02, 0x63, 0xb0, 0xc3, 0xff, 0x97, 
+    0xbe, 0x31, 0x20, 0x0c, 0x48, 0x08, 0x2e, 0x28, 0xa0, 0xd8, 
+    0x5d, 0xb8, 0xea, 0xaa, 0x00, 0x25, 0x7d, 0xf2, 0x06, 0xe1, 
+    0x0f, 0x5e, 0xfe, 0x68, 0x27, 0xcc, 0x9e, 0xb4, 0x65, 0x14, 
+    0x1e, 0xd1, 0x52, 0x11, 0xdd, 0x4e, 0x7b, 0x3f, 0xb1, 0xaf, 
+    0xc6, 0x5d, 0x01, 0x8a, 0x82, 0x5c, 0xd0, 0xb0, 0x55, 0x95, 
+    0xbc, 0x7b, 0x09, 0xa7, 0x63, 0x9d, 0xdd, 0x98, 0x0c, 0x0d, 
+    0xb3, 0x07, 0x63, 0x08, 0x8c, 0xd6, 0xc5, 0x36, 0xe4, 0x05, 
+    0x96, 0x90, 0xaf, 0x2e, 0xde, 0xe3, 0x8e, 0xec, 0xfd, 0x5f, 
+    0x46, 0xfa, 0x2a, 0x92, 0x4f, 0x41, 0x50, 0xcc, 0xcc, 0xa9, 
+    0x34, 0xb3, 0xa1, 0x7c, 0xd4, 0x7f, 0x51, 0xc4, 0x71, 0x63, 
+    0xf1, 0x5e, 0x4c, 0xf7, 0xe1, 0x0d, 0xe6, 0x25, 0x9d, 0x4b, 
+    0x6b, 0xc6, 0x4c, 0x23, 0x6f, 0x09, 0x04, 0x1a, 0x3e, 0x90, 
+    0x47, 0xe7, 0xdd, 0xcc, 0x8c, 0xc7, 0x16, 0xa7, 0x27, 0x18, 
+    0x9c, 0x6f, 0x71, 0xfc, 0x79, 0x69, 0x81, 0xc7, 0xde, 0x2f, 
+    0xaa, 0x44, 0x9d, 0xd8, 0x9c, 0xae, 0x1a, 0x46, 0x4e, 0x21, 
+    0x6d, 0x09, 0x06, 0x5a, 0x7e, 0x90, 0x47, 0xe7, 0xdc, 0xcc, 
+    0x8c, 0xc7, 0x16, 0xa7, 0x27, 0x38, 0x9c, 0x6f, 0x71, 0xfc, 
+    0x59, 0x69, 0x81, 0xc7, 0xce, 0x27, 0xe9, 0x45, 0x95, 0xd8, 
+    0x9c, 0x2e, 0x1b, 0xc6, 0xb1, 0xde, 0x92, 0xf6, 0xf9, 0xa5, 
+    0x81, 0x6f, 0xb8, 0x18, 0x23, 0x33, 0x73, 0x38, 0xe9, 0x58, 
+    0xd8, 0xc7, 0x63, 0x90, 0x8e, 0x03, 0xa6, 0x96, 0x7e, 0x38, 
+    0x31, 0xd8, 0x16, 0xba, 0x6a, 0x27, 0x63, 0xd1, 0xe4, 0x39, 
+    0xa5, 0xd1, 0x26, 0xe4, 0xca, 0x7a, 0x4f, 0x67, 0x90, 0xd2, 
+    0xb7, 0xfb, 0x20, 0xb7, 0x9d, 0xeb, 0xd4, 0xe8, 0x94, 0x13, 
+    0x7a, 0x9a, 0xb6, 0x93, 0x8c, 0x02, 0x39, 0xa1, 0x0b, 0x6b, 
+    0x1f, 0x1d, 0x33, 0xeb, 0x64, 0x00, 0xa6, 0x62, 0x12, 0xa2, 
+    0xd0, 0xd8, 0xeb, 0x9e, 0x7b, 0x50, 0xf8, 0x6f, 0xbc, 0x0e, 
+    0x50, 0xa4, 0xf9, 0x7d, 0x41, 0xf3, 0xbb, 0x2c, 0x5d, 0xd7, 
+    0x72, 0x39, 0x25, 0x12, 0xff, 0x33, 0x0c, 0x9b, 0x9c, 0x8a, 
+    0x68, 0xd5, 0xfe, 0x46, 0x62, 0xf1, 0x32, 0x94, 0x3d, 0xce, 
+    0xc0, 0x33, 0xfb, 0x3c, 0x61, 0xe0, 0xaa, 0x3c, 0xb5, 0x77, 
+    0x3c, 0x96, 0x47, 0x46, 0xa0, 0x6f, 0x9e, 0x68, 0xf2, 0x4a, 
+    0x53, 0xd5, 0x82, 0xf0, 0x4e, 0x2e, 0x4e, 0x55, 0x1b, 0x89, 
+    0xaf, 0x00, 0x06, 0xdd, 0x91, 0x56, 0xf8, 0x3a, 0x4e, 0xbf, 
+    0xa8, 0x08, 0x20, 0x0a, 0xe5, 0x33, 0x9f, 0xe2, 0x13, 0xc3, 
+    0x9e, 0xa4, 0x1d, 0x19, 0x8c, 0x97, 0x6b, 0xae, 0xba, 0x14, 
+    0x9f, 0x0c, 0xf4, 0x41, 0xe7, 0x99, 0x3d, 0xde, 0xdb, 0x56, 
+    0x92, 0x68, 0x5e, 0x8c, 0xde, 0x97, 0xe7, 0x4c, 0x13, 0x25, 
+    0xae, 0xc6, 0x38, 0x9e, 0xb6, 0xa3, 0xe4, 0x8d, 0xa2, 0x77, 
+    0xee, 0xd8, 0xd0, 0x54, 0x20, 0xd4, 0x6a, 0x3d, 0xbd, 0xb9, 
+    0xae, 0x13, 0xb5, 0x4e, 0x93, 0x66, 0xd0, 0xed, 0x7d, 0x1c, 
+    0x5c, 0xb5, 0xe3, 0x8c, 0x26, 0x77, 0x98, 0x97, 0x1c, 0x83, 
+    0x8d, 0x43, 0xbe, 0x35, 0x96, 0x2a, 0xbf, 0xfc, 0x10, 0x94, 
+    0x02, 0xa0, 0x13, 0x10, 0xb5, 0xf5, 0xa7, 0x13, 0xb5, 0x4e, 
+    0x93, 0x72, 0x90, 0xed, 0x7c, 0x1c, 0x5e, 0xb7, 0xa3, 0x4c, 
+    0x26, 0x77, 0x88, 0x87, 0x1c, 0x83, 0x8d, 0x43, 0xae, 0xa5, 
+    0x9e, 0x2a, 0xbf, 0xdc, 0x10, 0x94, 0x32, 0xb0, 0x13, 0x10, 
+    0xb5, 0xf1, 0x58, 0xec, 0x4a, 0xb1, 0x6c, 0x8d, 0x6f, 0x12, 
+    0x83, 0xe3, 0xa1, 0x48, 0x5c, 0xb3, 0xd9, 0x88, 0x77, 0x78, 
+    0xe3, 0x7c, 0x72, 0xbc, 0x51, 0x5a, 0x61, 0xd5, 0x40, 0x23, 
+    0xef, 0x6b, 0xcd, 0x4f, 0xec, 0xef, 0x4a, 0x0e, 0x02, 0x00, 
+    0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x04, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 
+    0x03, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x04, 0x00, 
+    0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 
+    0x02, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x07, 0x00, 
+    0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 
+    0x06, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x02, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x05, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x06, 0x00, 
+    0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x04, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x03, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x05, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 
+    0x02, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x07, 0x00, 
+    0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x04, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x04, 0x00, 
+    0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x03, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x04, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 
+    0x06, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x2a, 0xbf, 0x90, 0xb2, 
+    0xd0, 0x0d, 0x6d, 0x55, 0x3d, 0x93, 0x08, 0x8e, 0xde, 0xf2, 
+    0x8a, 0x38, 0xda, 0x6c, 0x3f, 0x12, 0xfe, 0x62, 0xb9, 0xa9, 
+    0xa4, 0x1e, 0x6a, 0x23, 0xfc, 0x16, 0xf0, 0x37, 0xb0, 0xe4, 
+    0xeb, 0x72, 0xa8, 0xa0, 0xc7, 0x7e, 0x0e, 0xa3, 0xae, 0x25, 
+    0xff, 0xee, 0xdb, 0x54, 0x83, 0x82, 0xd2, 0x9c, 0xbf, 0x2e, 
+    0x78, 0xc4, 0x32, 0x08, 0xf4, 0x60, 0x88, 0xa6, 0xcd, 0x1f, 
+    0x91, 0x50, 0x71, 0x26, 0x9a, 0x9a, 0x98, 0x2a, 0xc6, 0x53, 
+    0x15, 0xc9, 0xcc, 0x47, 0x3b, 0x64, 0xd6, 0x21, 0xe3, 0x9b, 
+    0x13, 0x27, 0x81, 0x43, 0x86, 0xc3, 0xe8, 0xd2, 0x62, 0xfb, 
+    0xae, 0x5b, 0xc5, 0x5b, 0x55, 0x45, 0xa5, 0x64, 0xb2, 0x8c, 
+    0xf8, 0x9c, 0x78, 0x51, 0x7e, 0x95, 0x39, 0x7d, 0x8f, 0x95, 
+    0x10, 0x4b, 0xa3, 0xba, 0x45, 0x0b, 0xc4, 0x1f, 0x4f, 0xf0, 
+    0x89, 0x75, 0x97, 0x1a, 0x16, 0x5a, 0x1d, 0x21, 0xc8, 0xb1, 
+    0x24, 0xd7, 0x9d, 0xe8, 0x1e, 0x99, 0x32, 0xfa, 0xed, 0x16, 
+    0xc3, 0x97, 0x5d, 0x64, 0xbf, 0x16, 0xdb, 0x60, 0x1e, 0x58, 
+    0x62, 0xc3, 0xfb, 0x17, 0x09, 0xb9, 0x68, 0x58, 0xee, 0x75, 
+    0xfd, 0x98, 0x74, 0xea, 0x1e, 0x38, 0x66, 0x5a, 0x1f, 0xd9, 
+    0x19, 0x43, 0x48, 0xdc, 0x98, 0x66, 0x02, 0xd7, 0xfc, 0x47, 
+    0x0f, 0x06, 0xd6, 0xee, 0x26, 0x9a, 0xe1, 0x01, 0x1b, 0x13, 
+    0x47, 0xc3, 0xf0, 0x42, 0xa4, 0x10, 0xbd, 0x7c, 0x74, 0x4e, 
+    0x15, 0xf1, 0x45, 0x5f, 0x8f, 0x89, 0xc5, 0x13, 0x48, 0xd9, 
+    0x53, 0x26, 0x82, 0xd7, 0xfc, 0x65, 0x0f, 0x06, 0xd6, 0x6a, 
+    0x26, 0x98, 0xe1, 0x01, 0x1b, 0x13, 0x4f, 0xd3, 0xf0, 0x42, 
+    0xa4, 0x10, 0xbd, 0x7c, 0x74, 0x4e, 0x14, 0xf1, 0x45, 0x5f, 
+    0x8f, 0x89, 0xc5, 0x13, 0x48, 0xd8, 0x5a, 0x26, 0x7d, 0x28, 
+    0x03, 0x9a, 0xf0, 0xf9, 0x29, 0x95, 0xd9, 0x67, 0x1e, 0xfe, 
+    0xe4, 0xec, 0xb0, 0x2c, 0x0f, 0xbd, 0x5b, 0xef, 0x42, 0x83, 
+    0x8b, 0xb1, 0xeb, 0x0e, 0xba, 0xa0, 0x70, 0x76, 0x3a, 0xec, 
+    0xb7, 0x27, 0xa5, 0xd9, 0x8b, 0x03, 0xb2, 0xa5, 0x06, 0x4e, 
+    0x04, 0x58, 0x73, 0x6b, 0x76, 0x95, 0xe1, 0x26, 0x04, 0xd1, 
+    0x68, 0x1d, 0x52, 0x58, 0x2b, 0x67, 0x02, 0xcb, 0x77, 0xc3, 
+    0x62, 0x17, 0xc7, 0x85, 0x2e, 0xc3, 0x09, 0x60, 0x78, 0x56, 
+    0x3d, 0x31, 0xfd, 0x95, 0xd1, 0x36, 0x78, 0x01, 0x1d, 0x1c, 
+    0xda, 0x0b, 0xa3, 0x4f, 0x7f, 0xe2, 0x0b, 0xaa, 0x96, 0x81, 
+    0xa0, 0x62, 0x0d, 0x7a, 0x09, 0xd2, 0x9e, 0x41, 0xb1, 0x01, 
+    0xa8, 0x71, 0x50, 0x85, 0x4c, 0xf5, 0x75, 0xdf, 0xd0, 0x16, 
+    0x1d, 0x14, 0x3f, 0x0f, 0x76, 0x9c, 0xd3, 0x76, 0xdc, 0x31, 
+    0xb5, 0x32, 0x22, 0xf3, 0x12, 0x7c, 0x2e, 0xbc, 0xaa, 0xdb, 
+    0xbe, 0x45, 0xd6, 0x95, 0x83, 0x7d, 0x54, 0x1f, 0x8c, 0x94, 
+    0x4e, 0xa1, 0x6d, 0x93, 0x58, 0x3e, 0x08, 0x6a, 0xee, 0xfc, 
+    0x58, 0x31, 0x0d, 0x07, 0x13, 0xd5, 0x40, 0xb8, 0x9d, 0x9a, 
+    0x08, 0x92, 0x3c, 0x17, 0x6e, 0x0c, 0x39, 0x94, 0x9d, 0xb3, 
+    0xb7, 0x88, 0x6a, 0x87, 0x3b, 0x90, 0x91, 0xf5, 0x88, 0xd6, 
+    0x19, 0x48, 0xf3, 0xcf, 0x80, 0x05, 0x76, 0x6e, 0x5b, 0x85, 
+    0xc9, 0xeb, 0x4a, 0x97, 0xfe, 0xc3, 0xe8, 0x79, 0xd4, 0x63, 
+    0x0c, 0x55, 0x6c, 0xc3, 0x67, 0x55, 0x3c, 0xbe, 0x7e, 0x83, 
+    0x1b, 0xbb, 0x35, 0x22, 0xe7, 0x12, 0xb8, 0x79, 0xf4, 0xec, 
+    0xa0, 0x45, 0x74, 0x62, 0x73, 0x4e, 0x49, 0x87, 0xc0, 0x32, 
+    0xdd, 0x00, 0xca, 0xff, 0xb4, 0x27, 0x06, 0x4f, 0x7c, 0xd6, 
+    0xe7, 0x15, 0x3d, 0x44, 0x7a, 0x49, 0x3b, 0xe1, 0x86, 0xb3, 
+    0xed, 0x92, 0x19, 0x69, 0xf4, 0xec, 0xa0, 0x45, 0x74, 0x60, 
+    0x73, 0x4e, 0x09, 0xc7, 0xc0, 0x32, 0xdd, 0x02, 0xc8, 0xfd, 
+    0xb4, 0x27, 0x0e, 0x4e, 0x7c, 0xd2, 0xe5, 0x15, 0x3d, 0x24, 
+    0x7a, 0x43, 0x3b, 0xa1, 0x86, 0xb3, 0x12, 0x6d, 0xe6, 0x96, 
+    0x0b, 0x13, 0x5f, 0xba, 0x8b, 0x9f, 0x8c, 0xb1, 0xf6, 0x38, 
+    0x3f, 0xcd, 0x22, 0xfd, 0x37, 0x02, 0x4b, 0xd8, 0xf1, 0xb1, 
+    0x83, 0x2d, 0x1a, 0xea, 0xc2, 0xdb, 0x85, 0xbc, 0xc4, 0x5e, 
+    0x79, 0x4c, 0x04, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x02, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x05, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x02, 0x00, 
+    0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 
+    0x07, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x04, 0x00, 
+    0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x07, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x06, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x04, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 
+    0x03, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x04, 0x00, 
+    0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x07, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x04, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x04, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 
+    0x05, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x06, 0x00, 
+    0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 
+    0x06, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x08, 0xb0, 0x69, 0x3d, 0xf0, 0x30, 0x89, 0x7c, 0x2a, 0x99, 
+    0x41, 0xf5, 0x62, 0xf2, 0x31, 0xc9, 0x5e, 0xe5, 0x5a, 0x83, 
+    0x52, 0x81, 0x5a, 0x27, 0xfc, 0xd7, 0x96, 0x34, 0xa1, 0xe5, 
+    0x6f, 0xa1, 0xec, 0xf3, 0x8d, 0x5d, 0x2d, 0xf9, 0x52, 0x0c, 
+    0x06, 0x21, 0x95, 0xab, 0x7f, 0xc9, 0xd2, 0x18, 0x1a, 0x70, 
+    0x9b, 0x03, 0x52, 0xf4, 0x78, 0x1a, 0x7c, 0x38, 0xaf, 0xe5, 
+    0x95, 0x74, 0xf2, 0x12, 0x9c, 0x75, 0x92, 0x8a, 0x48, 0xec, 
+    0x73, 0x21, 0x61, 0xbc, 0x75, 0x18, 0xc6, 0x46, 0xee, 0xf5, 
+    0xbb, 0x5b, 0xb2, 0x03, 0x39, 0x1f, 0x3c, 0x31, 0xba, 0x49, 
+    0xf0, 0xfb, 0xe4, 0x19, 0x99, 0x17, 0x46, 0x47, 0x4a, 0x46, 
+    0x98, 0x10, 0x46, 0x8a, 0x7a, 0x76, 0xdf, 0x94, 0x83, 0x20, 
+    0x27, 0x27, 0x68, 0xc4, 0x9b, 0x4a, 0x56, 0x2c, 0x2d, 0xb5, 
+    0x14, 0xe0, 0xc6, 0x60, 0x08, 0x08, 0xd2, 0xa8, 0xca, 0xad, 
+    0x65, 0x6f, 0xa4, 0xf5, 0x63, 0xd1, 0x6a, 0x85, 0xa9, 0xb2, 
+    0x87, 0x0c, 0xad, 0xe9, 0x8a, 0xc8, 0x2e, 0x39, 0xaa, 0x20, 
+    0xfa, 0xf7, 0xd7, 0x6d, 0x70, 0x5f, 0xbc, 0x8f, 0x91, 0xaf, 
+    0xff, 0x60, 0xdc, 0x75, 0x0d, 0x72, 0xe7, 0x79, 0xba, 0x59, 
+    0x6d, 0x38, 0x8e, 0xce, 0x9e, 0xa4, 0x64, 0x46, 0xf4, 0x22, 
+    0xfc, 0x28, 0x39, 0x3a, 0xae, 0x3a, 0xba, 0xff, 0x57, 0xe8, 
+    0x99, 0xbf, 0xfc, 0xe4, 0xc9, 0x8e, 0x2f, 0x60, 0xde, 0x70, 
+    0xd7, 0xe5, 0xc3, 0x7d, 0x99, 0x1b, 0x00, 0x05, 0xbc, 0x64, 
+    0xd3, 0xac, 0x04, 0xef, 0xfa, 0x38, 0xfc, 0xa8, 0x3b, 0x3b, 
+    0xae, 0x2a, 0xba, 0xff, 0x57, 0xa8, 0xb1, 0xbf, 0xfc, 0xe5, 
+    0xc9, 0xae, 0x2f, 0x60, 0xd8, 0x71, 0xc7, 0xe5, 0xc3, 0x7d, 
+    0x99, 0x1b, 0x00, 0x05, 0xbc, 0xc4, 0xdb, 0xac, 0x04, 0xee, 
+    0xfa, 0x38, 0x03, 0x57, 0xc4, 0xc4, 0x51, 0xd5, 0x45, 0x00, 
+    0xa8, 0x57, 0x4e, 0x40, 0x03, 0x1a, 0x36, 0x51, 0xd0, 0x9f, 
+    0x27, 0x8e, 0x38, 0x1a, 0x3c, 0x82, 0x66, 0xe4, 0xff, 0xfa, 
+    0x43, 0x3b, 0x24, 0x53, 0xfb, 0x11, 0x05, 0xc7, 0x02, 0xb3, 
+    0x45, 0x5e, 0x4a, 0xbb, 0x70, 0x66, 0xcc, 0x3c, 0xf1, 0xbc, 
+    0xd4, 0x95, 0xc9, 0x36, 0x85, 0x7b, 0xc5, 0xb3, 0xf7, 0xdb, 
+    0x65, 0xbb, 0xd2, 0xed, 0xb2, 0xae, 0x40, 0x6b, 0x45, 0x46, 
+    0xd5, 0x07, 0xf4, 0xa7, 0x20, 0x11, 0xe8, 0xaf, 0x4c, 0x8c, 
+    0x7a, 0xa8, 0xc7, 0x26, 0x74, 0x84, 0x29, 0x59, 0x08, 0x70, 
+    0xcf, 0x46, 0xf5, 0xd2, 0x99, 0x60, 0x0f, 0x19, 0x5b, 0xf5, 
+    0x32, 0xc5, 0x4e, 0xf8, 0x76, 0x13, 0x18, 0x00, 0x56, 0x02, 
+    0x32, 0x2a, 0x78, 0xb5, 0x85, 0xcb, 0x29, 0x91, 0x42, 0xa8, 
+    0xf1, 0xd7, 0xd5, 0x41, 0x43, 0x87, 0xd2, 0x4a, 0xa3, 0x25, 
+    0x40, 0x5d, 0x95, 0xf9, 0xe5, 0x01, 0x9e, 0xe0, 0x49, 0xac, 
+    0x9e, 0x99, 0x85, 0xdf, 0x4b, 0xb3, 0x4f, 0xce, 0x86, 0x49, 
+    0xb3, 0x3c, 0x59, 0x44, 0x7b, 0x87, 0xb3, 0x30, 0x31, 0xeb, 
+    0xe5, 0x1c, 0x72, 0x29, 0x8e, 0x80, 0xe1, 0x16, 0xf9, 0x25, 
+    0xa3, 0xe6, 0xa8, 0x03, 0xfe, 0x25, 0x0f, 0xbf, 0x13, 0x5c, 
+    0x03, 0x1d, 0xc5, 0xf7, 0x17, 0x9e, 0xa9, 0x18, 0x03, 0x35, 
+    0x38, 0x17, 0xa7, 0xe8, 0x54, 0x6b, 0x0b, 0xbd, 0xc9, 0x15, 
+    0xb6, 0x64, 0x2a, 0xe3, 0x08, 0x0a, 0x6a, 0x96, 0x00, 0x5a, 
+    0xdc, 0x8c, 0x08, 0x17, 0xb6, 0x3d, 0x83, 0x44, 0xdd, 0xca, 
+    0x60, 0xba, 0x45, 0x10, 0x71, 0x20, 0xf1, 0xaf, 0xd7, 0xc5, 
+    0x56, 0x6f, 0x8a, 0xa5, 0x80, 0x32, 0xb4, 0x40, 0x2a, 0x67, 
+    0xe6, 0x25, 0x6b, 0xc7, 0x00, 0x40, 0xd2, 0x85, 0x03, 0xb5, 
+    0xd2, 0x1f, 0x81, 0x5f, 0xdd, 0xca, 0x62, 0xba, 0x25, 0x10, 
+    0x79, 0x20, 0xf0, 0xaf, 0xd7, 0xe5, 0x54, 0x6f, 0x8b, 0xa5, 
+    0x80, 0x20, 0xb6, 0x40, 0x2a, 0xe7, 0xe4, 0x25, 0x6b, 0xc6, 
+    0x00, 0x40, 0xda, 0x85, 0x03, 0xb7, 0xd2, 0x1f, 0xc1, 0x5f, 
+    0x22, 0x35, 0x9d, 0x45, 0xda, 0xef, 0x86, 0xdf, 0x0f, 0x50, 
+    0x28, 0x1a, 0xab, 0x90, 0x74, 0x5a, 0x7f, 0xdf, 0x49, 0xbf, 
+    0xd5, 0x18, 0x1b, 0xda, 0x94, 0x39, 0xff, 0xbf, 0x25, 0x7a, 
+    0xfc, 0x48, 0x2d, 0xe0, 0x3e, 0xa0, 0x05, 0x00, 0x00, 0x00, 
+    0x04, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x05, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x02, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 
+    0x05, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x03, 0x00, 
+    0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 
+    0x03, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 
+    0x03, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 
+    0x02, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x06, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x04, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x07, 0x00, 
+    0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x07, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x04, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x02, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 
+    0x07, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x06, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x07, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x07, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x06, 0x00, 
+    0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x05, 0x00, 0x00, 0x00, 0x5e, 0x8c, 0xe5, 0x94, 0x09, 0xbf, 
+    0x89, 0x20, 0xb5, 0xdd, 0x62, 0x7d, 0xab, 0x5e, 0xe2, 0xe9, 
+    0x9e, 0x0e, 0x41, 0x1b, 0x0d, 0xd5, 0xd6, 0x27, 0xe3, 0x4b, 
+    0x6f, 0xd3, 0x1b, 0xa4, 0x66, 0x57, 0xfd, 0x22, 0xad, 0x00, 
+    0x00, 0xa6, 0xe4, 0x7b, 0xe0, 0x76, 0xc0, 0xaa, 0x0c, 0x8e, 
+    0x84, 0xbf, 0x9c, 0xb9, 0x19, 0x24, 0x53, 0x97, 0x41, 0x8f, 
+    0x38, 0x30, 0xcd, 0x4c, 0x35, 0x1d, 0xb0, 0x02, 0x8b, 0x64, 
+    0x0d, 0x6c, 0x1c, 0x60, 0x79, 0x8a, 0x7f, 0x02, 0x13, 0x63, 
+    0xd8, 0xe3, 0x2a, 0x48, 0x5c, 0x0f, 0x11, 0xaf, 0xc9, 0xc8, 
+    0x60, 0x87, 0xef, 0x22, 0xc9, 0x53, 0x5f, 0x3f, 0x5c, 0x21, 
+    0x07, 0x73, 0x3d, 0xd2, 0xc1, 0x55, 0x21, 0x33, 0xd9, 0xa4, 
+    0xda, 0xa6, 0x43, 0xd6, 0xa7, 0x23, 0xb2, 0x84, 0x33, 0xa4, 
+    0x26, 0xf5, 0x10, 0x51, 0x26, 0x1b, 0x75, 0x2f, 0x5d, 0xde, 
+    0x61, 0x26, 0xf1, 0x1d, 0x48, 0x5c, 0x70, 0x37, 0x35, 0x49, 
+    0x8f, 0x9d, 0xe1, 0x11, 0xf2, 0x07, 0xde, 0x6b, 0xcd, 0xf2, 
+    0xfe, 0xf9, 0xc4, 0xae, 0x35, 0x67, 0x57, 0xc3, 0xbf, 0x5d, 
+    0xd1, 0x56, 0x96, 0x1c, 0x44, 0x99, 0x9d, 0x32, 0x76, 0xf4, 
+    0x03, 0x36, 0x46, 0x97, 0xfc, 0x92, 0x93, 0xce, 0xf4, 0xb6, 
+    0xb0, 0x15, 0xd1, 0x1e, 0xce, 0xf2, 0xfd, 0xbb, 0xec, 0xb2, 
+    0x37, 0x27, 0xc3, 0xf3, 0x99, 0xfd, 0xe3, 0x77, 0x18, 0x53, 
+    0xc4, 0x83, 0x01, 0x68, 0x7e, 0x5f, 0x77, 0x70, 0x66, 0x93, 
+    0xae, 0x50, 0x09, 0x17, 0xe0, 0x5a, 0xc0, 0x03, 0xf9, 0xdc, 
+    0xcf, 0xf2, 0xfd, 0xbb, 0xe4, 0xb2, 0x37, 0x27, 0x47, 0xf3, 
+    0x99, 0xdd, 0xe3, 0x77, 0x18, 0x17, 0xc4, 0x83, 0x41, 0x22, 
+    0x7e, 0x5f, 0x65, 0x70, 0x66, 0x93, 0xac, 0x52, 0x0b, 0x17, 
+    0xe0, 0x3a, 0x90, 0x03, 0xf1, 0xde, 0x30, 0x0d, 0x02, 0x44, 
+    0x1b, 0x4d, 0xc8, 0xd8, 0xb8, 0x0c, 0x66, 0x22, 0x1c, 0x88, 
+    0xe7, 0xe8, 0x3b, 0x7c, 0xbe, 0xdd, 0x81, 0xa0, 0x9a, 0x8f, 
+    0x99, 0x6c, 0x53, 0xad, 0xf4, 0xe8, 0x1f, 0xc5, 0x6f, 0xfc, 
+    0x0e, 0x21, 0xc4, 0xe9, 0x80, 0x29, 0x1b, 0x9a, 0x33, 0xcd, 
+    0x51, 0x7f, 0xdf, 0x12, 0x1f, 0x1d, 0xcb, 0xcc, 0x75, 0xe6, 
+    0x78, 0x04, 0x30, 0x24, 0x24, 0x52, 0xbd, 0x34, 0x79, 0x7f, 
+    0x4a, 0x8a, 0xd0, 0x42, 0xa3, 0x91, 0xc6, 0x25, 0xec, 0xce, 
+    0x34, 0xc9, 0x0c, 0x1e, 0x70, 0xa9, 0xd2, 0xcb, 0xcc, 0xe7, 
+    0xd4, 0xfd, 0x83, 0x48, 0xc6, 0x7b, 0x7e, 0x42, 0x19, 0x3b, 
+    0x3c, 0x05, 0xa9, 0x40, 0xa6, 0xaf, 0x74, 0xfa, 0x59, 0x73, 
+    0x5f, 0x09, 0x6c, 0x6f, 0x8c, 0x3e, 0x04, 0xb7, 0xf8, 0xd4, 
+    0xef, 0xda, 0x6a, 0xe2, 0x17, 0x4a, 0x18, 0x38, 0xae, 0x29, 
+    0xf9, 0x3c, 0x57, 0x0b, 0xcd, 0x64, 0x32, 0x53, 0x32, 0xd8, 
+    0xd1, 0x68, 0x32, 0xc6, 0x9c, 0x9d, 0x2d, 0x37, 0x81, 0x42, 
+    0xf8, 0x11, 0xf9, 0x95, 0xe5, 0x21, 0xe1, 0x26, 0x38, 0x18, 
+    0x4e, 0xcc, 0x17, 0xb0, 0x2a, 0x1e, 0x80, 0x10, 0x36, 0xfd, 
+    0x42, 0x1b, 0xf6, 0x4c, 0x8d, 0x9a, 0x55, 0x0a, 0x0f, 0x7e, 
+    0x76, 0xe8, 0xc6, 0x6f, 0x7a, 0x63, 0xd1, 0xe4, 0x58, 0x95, 
+    0x85, 0x12, 0x02, 0x02, 0xc3, 0xb7, 0x4e, 0x96, 0x4c, 0x8b, 
+    0xd8, 0xf6, 0xce, 0xd4, 0x89, 0xee, 0xa4, 0x4b, 0xd7, 0x00, 
+    0x50, 0x98, 0x94, 0x99, 0x26, 0x01, 0xa2, 0x38, 0x0d, 0x80, 
+    0xb7, 0x13, 0xfd, 0xe0, 0x78, 0x37, 0x29, 0x92, 0x39, 0x0a, 
+    0xc3, 0xa7, 0x96, 0x8a, 0x44, 0x0d, 0x5b, 0xde, 0xde, 0x52, 
+    0x8a, 0xae, 0xb0, 0x48, 0x97, 0x10, 0x7c, 0x81, 0x50, 0xbc, 
+    0x06, 0x1a, 0xb9, 0xaa, 0x44, 0x63, 0x2a, 0x61, 0xf9, 0xe0, 
+    0x78, 0xb5, 0x2d, 0x92, 0x38, 0x0a, 0xc3, 0xa7, 0xd6, 0xca, 
+    0x44, 0x19, 0xda, 0xde, 0xde, 0x50, 0x8a, 0xae, 0xb0, 0x48, 
+    0x97, 0x10, 0x7c, 0x81, 0x50, 0xbc, 0x06, 0x1a, 0xbb, 0xa8, 
+    0x44, 0x63, 0x3f, 0x61, 0x06, 0x1f, 0x87, 0x4a, 0xd2, 0x6d, 
+    0xc7, 0xf5, 0x3c, 0x58, 0x29, 0x35, 0xbb, 0xe6, 0x25, 0x21, 
+    0x21, 0xaf, 0x75, 0x51, 0x4f, 0xb7, 0x68, 0xef, 0x83, 0x7e, 
+    0xaf, 0x43, 0xf9, 0xe5, 0x44, 0x57, 0xbb, 0x9c, 0xc0, 0x9e, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x06, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 
+    0x04, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x07, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 
+    0x02, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x04, 0x00, 
+    0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x02, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x06, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 
+    0x07, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x03, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 
+    0x07, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x02, 0x00, 
+    0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x06, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x06, 0x00, 
+    0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x03, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x03, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x02, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x04, 0x00, 
+    0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x06, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x03, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x07, 0x00, 
+    0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x02, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0xa1, 0x4c, 
+    0xe2, 0xf8, 0xca, 0x04, 0xba, 0x55, 0x88, 0x7c, 0xb0, 0x5a, 
+    0x80, 0x66, 0x17, 0x3f, 0x3e, 0x3a, 0xdd, 0x3e, 0xbd, 0x7d, 
+    0x53, 0x7e, 0xf5, 0x2d, 0xfd, 0xd3, 0xc0, 0x78, 0xba, 0xf6, 
+    0x0b, 0xe0, 0xca, 0x1d, 0xcd, 0x1a, 0xe1, 0x6c, 0xc2, 0x50, 
+    0x84, 0xec, 0x3f, 0xfb, 0xe1, 0x09, 0x88, 0x60, 0x0c, 0x6b, 
+    0xa8, 0x58, 0x3b, 0x3b, 0x06, 0x5a, 0x5a, 0xa4, 0x5d, 0x0c, 
+    0xd0, 0xe4, 0x7c, 0x57, 0x06, 0xdc, 0xc4, 0x4c, 0xe6, 0x0d, 
+    0xc4, 0x5e, 0xec, 0x2a, 0x7b, 0xf9, 0x03, 0x5e, 0x21, 0xe2, 
+    0xd9, 0x52, 0x05, 0x7c, 0xb6, 0xa9, 0x8c, 0x47, 0x18, 0x2f, 
+    0x2a, 0x56, 0xb1, 0xcc, 0x6c, 0xf4, 0xfa, 0x3c, 0xd9, 0xe4, 
+    0x40, 0x55, 0x9d, 0x81, 0x45, 0x65, 0xe8, 0x0d, 0x52, 0xca, 
+    0x75, 0x4e, 0x49, 0xb4, 0xf6, 0xe0, 0x24, 0x97, 0x9d, 0xff, 
+    0x10, 0x01, 0x4a, 0xce, 0x58, 0x55, 0x91, 0x2c, 0x19, 0x78, 
+    0xfe, 0x75, 0x21, 0x56, 0x32, 0xe2, 0x4f, 0x00, 0x85, 0x23, 
+    0x7d, 0x12, 0x9e, 0xfa, 0xcc, 0xcc, 0xab, 0x02, 0x71, 0xe0, 
+    0x7f, 0x82, 0x78, 0x0b, 0x47, 0x22, 0x70, 0xa2, 0xd8, 0xa7, 
+    0x54, 0xec, 0x0b, 0x8f, 0x8a, 0x1b, 0xd8, 0x74, 0xdb, 0xc4, 
+    0x68, 0xdb, 0xa4, 0xb1, 0xf6, 0x07, 0x48, 0x6d, 0xd8, 0xa2, 
+    0xec, 0xcc, 0x16, 0xda, 0xaf, 0xc4, 0x35, 0xf2, 0x58, 0x83, 
+    0x43, 0xf6, 0x30, 0xa8, 0xdb, 0x25, 0x76, 0x3d, 0xa3, 0xba, 
+    0x2d, 0xd9, 0xf8, 0x58, 0x5d, 0x94, 0xaa, 0x53, 0xe4, 0xb0, 
+    0x36, 0x07, 0x7d, 0xa4, 0xd8, 0xe2, 0xec, 0xcc, 0x16, 0xda, 
+    0xa7, 0xe4, 0x37, 0xe2, 0x58, 0x83, 0x47, 0xf6, 0x10, 0xa0, 
+    0xdb, 0xa5, 0x56, 0x3d, 0x83, 0xbe, 0xad, 0xd9, 0xf8, 0x5c, 
+    0xdd, 0xc4, 0xaa, 0x53, 0xe4, 0xb0, 0x36, 0x07, 0x6d, 0x24, 
+    0x27, 0x1d, 0x13, 0x33, 0xe9, 0x25, 0x58, 0x1b, 0xc8, 0x1d, 
+    0xa7, 0x7c, 0xb8, 0x09, 0xef, 0x5f, 0x24, 0x5a, 0xa9, 0xc2, 
+    0x7c, 0x41, 0x52, 0x26, 0x07, 0xa3, 0x22, 0x3b, 0x55, 0xac, 
+    0x1b, 0x4f, 0xc9, 0xf8, 0x92, 0xdb, 0xf5, 0xc8, 0x2c, 0x56, 
+    0x56, 0x94, 0x49, 0x68, 0xc1, 0xf7, 0x03, 0x1d, 0x14, 0x4b, 
+    0x24, 0xd7, 0xb0, 0x02, 0x28, 0x24, 0x23, 0x70, 0x83, 0x59, 
+    0x14, 0x49, 0x2f, 0xcf, 0xb7, 0x13, 0xf0, 0xac, 0x7b, 0x4f, 
+    0xcb, 0x65, 0xbb, 0x3a, 0xf8, 0x25, 0x4c, 0x22, 0x6c, 0xa1, 
+    0x46, 0xd3, 0x61, 0xdb, 0xf6, 0x42, 0xcd, 0x5e, 0x96, 0x6d, 
+    0x2a, 0xb1, 0xdd, 0x01, 0x43, 0xce, 0xf7, 0x2c, 0x4e, 0x37, 
+    0x50, 0xcb, 0xe9, 0x5d, 0x80, 0x7a, 0x4f, 0x16, 0x12, 0x7f, 
+    0xe6, 0x9c, 0x31, 0xe7, 0x1e, 0xff, 0xbc, 0x52, 0x6d, 0x11, 
+    0xf2, 0xae, 0x59, 0x96, 0x55, 0x8a, 0x2e, 0xa2, 0x98, 0x4a, 
+    0xe4, 0x24, 0x89, 0x55, 0xb0, 0x67, 0x21, 0x73, 0x12, 0x47, 
+    0x42, 0x66, 0x01, 0xca, 0x66, 0xa2, 0x2c, 0x07, 0x2b, 0xc5, 
+    0x5e, 0x5c, 0x8a, 0xca, 0xdf, 0x2e, 0x02, 0xaf, 0x43, 0xcd, 
+    0xb0, 0x83, 0x8b, 0xdb, 0x51, 0x75, 0xf6, 0xca, 0x43, 0x63, 
+    0x3a, 0x93, 0x40, 0x02, 0x52, 0x60, 0x7c, 0x50, 0x37, 0x79, 
+    0x0a, 0xaa, 0x0c, 0x20, 0x4e, 0xde, 0xd0, 0x6a, 0xda, 0xfe, 
+    0xef, 0xb6, 0x88, 0x87, 0x01, 0xb5, 0x4c, 0xda, 0x94, 0x6c, 
+    0xaa, 0xc6, 0x58, 0x8b, 0x16, 0xc7, 0x7d, 0x6e, 0x6d, 0x0a, 
+    0x79, 0x76, 0x01, 0x07, 0xb9, 0x85, 0x0c, 0xaf, 0xa4, 0xc2, 
+    0x4a, 0xeb, 0xbd, 0x78, 0xc1, 0xc3, 0xae, 0x37, 0x08, 0x0f, 
+    0x61, 0x93, 0x4e, 0xd1, 0x1e, 0x79, 0xd7, 0x82, 0xf1, 0x82, 
+    0xd7, 0x86, 0xfc, 0x12, 0xc1, 0x02, 0x79, 0x47, 0xd1, 0x87, 
+    0x3d, 0xdf, 0x4c, 0xae, 0xa4, 0xc2, 0x4e, 0xef, 0xfc, 0x78, 
+    0xc3, 0xeb, 0xaf, 0x37, 0x00, 0x0f, 0x61, 0x93, 0x4e, 0xd1, 
+    0x1e, 0x79, 0xb3, 0xc2, 0xf1, 0x82, 0xd3, 0x86, 0xfc, 0x12, 
+    0xc1, 0x02, 0x79, 0x66, 0x51, 0x87, 0x3d, 0x9f, 0xb3, 0x51, 
+    0x5b, 0x3d, 0xb1, 0x10, 0x03, 0x87, 0x3c, 0x14, 0x50, 0xc8, 
+    0xff, 0xf0, 0x9e, 0x6c, 0xb1, 0x2e, 0xe1, 0x86, 0x4c, 0x3d, 
+    0x0e, 0x7d, 0x2c, 0x79, 0x03, 0xed, 0x3e, 0xfd, 0x86, 0x99, 
+    0xae, 0x78, 0xc2, 0x60, 0x03, 0x00, 0x00, 0x00, 0x07, 0x00, 
+    0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x06, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 
+    0x04, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x02, 0x00, 
+    0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 
+    0x02, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x06, 0x00, 
+    0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x07, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x07, 0x00, 
+    0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x03, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 
+    0x04, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x03, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x04, 0x00, 
+    0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 
+    0x04, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 
+    0x04, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x04, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 
+    0x03, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x39, 0x04, 0xdf, 0x14, 0x5d, 0x57, 0x3d, 0xb9, 
+    0x1c, 0xc3, 0x1b, 0x0a, 0x74, 0xb8, 0x81, 0x09, 0x55, 0x55, 
+    0xb3, 0x9d, 0x0d, 0x7a, 0x83, 0xa8, 0x48, 0x47, 0x7a, 0x3d, 
+    0xff, 0xd1, 0xaf, 0xe1, 0xec, 0x2b, 0xe7, 0x67, 0xed, 0xe9, 
+    0xef, 0x2b, 0xdf, 0x10, 0x57, 0x78, 0x04, 0x1e, 0x3e, 0xcf, 
+    0x50, 0xe1, 0x5b, 0x41, 0xe2, 0x52, 0x56, 0x24, 0xac, 0x0f, 
+    0x43, 0x6d, 0x6b, 0x5e, 0x23, 0x1e, 0x96, 0xb2, 0x9e, 0x1a, 
+    0xd6, 0xa2, 0xfb, 0x90, 0x79, 0x04, 0x6d, 0xc4, 0x80, 0xc3, 
+    0x6d, 0x1f, 0x16, 0xee, 0x1c, 0x59, 0xf2, 0x6f, 0x7c, 0x15, 
+    0xbb, 0x4d, 0x5d, 0x24, 0x92, 0xfd, 0x4f, 0x3f, 0x05, 0xc2, 
+    0x9e, 0x5b, 0x78, 0x08, 0x84, 0x9a, 0x46, 0x82, 0xf4, 0x41, 
+    0x9a, 0x79, 0x68, 0x08, 0x61, 0x1e, 0x95, 0x67, 0x8d, 0x77, 
+    0xc9, 0x33, 0xbf, 0x7b, 0xa7, 0x00, 0x34, 0xf8, 0x26, 0x0b, 
+    0x66, 0x59, 0xf7, 0x75, 0xbd, 0x65, 0x79, 0xf1, 0x8b, 0xf1, 
+    0x18, 0x86, 0x04, 0x91, 0x54, 0xa0, 0xfa, 0x30, 0xaf, 0x3a, 
+    0xce, 0xad, 0x0c, 0xc3, 0x3d, 0x11, 0x01, 0x12, 0xa6, 0x9e, 
+    0x5b, 0xa8, 0xf0, 0x6e, 0x7f, 0x49, 0x25, 0x3c, 0x70, 0x87, 
+    0xb8, 0xae, 0x22, 0xff, 0xc3, 0xbf, 0xab, 0xf0, 0x3b, 0xb6, 
+    0x52, 0xc3, 0x8a, 0x62, 0xeb, 0x8a, 0x45, 0x88, 0xc6, 0xde, 
+    0x5f, 0x02, 0xc8, 0xe8, 0x97, 0x7c, 0x82, 0x38, 0x73, 0x60, 
+    0x4b, 0x40, 0x42, 0x38, 0xf0, 0x15, 0x3a, 0x26, 0x7a, 0xf5, 
+    0xad, 0xaf, 0xcb, 0x42, 0xf3, 0xb4, 0xd4, 0xc3, 0x9a, 0x62, 
+    0xeb, 0x8a, 0x45, 0x8c, 0x84, 0xde, 0x5f, 0x02, 0x48, 0xf8, 
+    0x97, 0x3e, 0x92, 0x38, 0x71, 0x60, 0x4b, 0x40, 0x06, 0x38, 
+    0xf0, 0x15, 0xb8, 0x26, 0x7a, 0xf5, 0xaf, 0xaf, 0xcb, 0x42, 
+    0xf3, 0xb4, 0x54, 0xc1, 0x65, 0x9d, 0x14, 0x75, 0xba, 0x73, 
+    0x7b, 0x21, 0xa0, 0xfd, 0xb7, 0x07, 0x68, 0xc1, 0x6d, 0xc7, 
+    0x8e, 0x9f, 0xb4, 0xbf, 0xf9, 0xc7, 0x0f, 0xea, 0x47, 0xd9, 
+    0x85, 0x0a, 0x50, 0x50, 0x34, 0xbd, 0x0c, 0x4b, 0xab, 0x3e, 
+    0xce, 0x72, 0x02, 0x24, 0x61, 0x4e, 0x7e, 0xfc, 0xd1, 0x99, 
+    0x90, 0x8e, 0xb5, 0x9d, 0x51, 0x96, 0xbb, 0x0b, 0x61, 0x9e, 
+    0xf3, 0xfa, 0x25, 0xb1, 0x9c, 0xdd, 0xb1, 0xb3, 0x09, 0xf4, 
+    0xe1, 0x35, 0xce, 0xa4, 0xa8, 0xf5, 0xd3, 0x3f, 0xb6, 0xad, 
+    0x39, 0x9b, 0x18, 0x35, 0x41, 0x6f, 0xb5, 0x3a, 0x72, 0x1b, 
+    0xac, 0x28, 0x6c, 0xeb, 0x2c, 0xc4, 0xdd, 0x1f, 0x90, 0xa6, 
+    0xbb, 0xef, 0xa3, 0x50, 0xbd, 0x26, 0xf0, 0xb5, 0x69, 0x45, 
+    0xd3, 0x1f, 0xff, 0x94, 0xb6, 0x12, 0x13, 0xfc, 0x50, 0x7f, 
+    0xff, 0x03, 0x41, 0x2b, 0x3b, 0x9f, 0x0a, 0x29, 0x9d, 0x82, 
+    0x7d, 0xc7, 0x92, 0x2f, 0x42, 0xd1, 0x7a, 0xbd, 0x32, 0x94, 
+    0x41, 0x64, 0xd1, 0xf3, 0x07, 0x9c, 0xd5, 0xb9, 0x33, 0xda, 
+    0x2f, 0x3e, 0xe0, 0x74, 0xb6, 0x7f, 0x82, 0x05, 0x02, 0x8d, 
+    0x57, 0x1e, 0x80, 0xd3, 0x02, 0xa3, 0x40, 0x87, 0x4d, 0x62, 
+    0xa6, 0xc6, 0x09, 0x7b, 0x10, 0xcd, 0x67, 0xec, 0x50, 0x17, 
+    0x21, 0xc8, 0xeb, 0x05, 0xc7, 0xb0, 0xc6, 0x60, 0x4a, 0xa2, 
+    0xaa, 0xbf, 0x4e, 0xfe, 0x9b, 0x2f, 0x83, 0x7c, 0xcc, 0x67, 
+    0xb3, 0xfa, 0x6b, 0x81, 0x5d, 0x92, 0x08, 0x2c, 0x0c, 0x5c, 
+    0x45, 0x91, 0x11, 0x91, 0x33, 0x1e, 0x2e, 0xf8, 0xf5, 0x2c, 
+    0xb7, 0x98, 0x8e, 0xc0, 0x87, 0x97, 0x8c, 0x36, 0x43, 0x8e, 
+    0xaf, 0x2f, 0x13, 0x90, 0xed, 0x6c, 0x87, 0x71, 0x76, 0xa7, 
+    0x49, 0xc0, 0x22, 0x1a, 0x0d, 0x39, 0x5c, 0x81, 0xb3, 0x81, 
+    0x39, 0xb6, 0x2a, 0xb2, 0xa3, 0xa5, 0xb7, 0x98, 0x8e, 0xc0, 
+    0x83, 0x9e, 0xac, 0x37, 0x43, 0x8e, 0xaf, 0x2f, 0x12, 0x98, 
+    0xec, 0x6e, 0x87, 0x71, 0x76, 0xa7, 0x49, 0xc0, 0x22, 0x1a, 
+    0x0d, 0x39, 0x5c, 0x89, 0xb3, 0x91, 0x31, 0x96, 0x2a, 0xb0, 
+    0xa3, 0xa5, 0x48, 0x67, 0x71, 0x3f, 0x7c, 0x61, 0x53, 0xc8, 
+    0xbc, 0x71, 0x50, 0xd0, 0xec, 0x67, 0x13, 0x91, 0x78, 0x8e, 
+    0x89, 0x58, 0xb6, 0x3f, 0xdd, 0xe5, 0xf2, 0xc6, 0xa3, 0x76, 
+    0x4c, 0x6e, 0xce, 0x69, 0xd5, 0x4f, 0x5c, 0x5a, 0x05, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x04, 0x00, 
+    0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x03, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x07, 0x00, 
+    0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 
+    0x07, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x04, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x07, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x06, 0x00, 
+    0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 
+    0x05, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x03, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 
+    0x06, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x06, 0x00, 
+    0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x02, 0x00, 
+    0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 
+    0x04, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 
+    0x02, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x02, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x16, 0xdf, 0x9b, 0x40, 
+    0x89, 0x8e, 0xcc, 0x66, 0x93, 0xa9, 0xbd, 0xc9, 0x51, 0xde, 
+    0x26, 0x23, 0x0e, 0x6b, 0x72, 0x28, 0x09, 0x13, 0xf0, 0x3a, 
+    0x26, 0x29, 0x3a, 0xc5, 0xf1, 0xd7, 0xf1, 0x1e, 0x68, 0x03, 
+    0xb1, 0xb1, 0x8e, 0xe3, 0xbf, 0x49, 0xad, 0xaf, 0xac, 0xe0, 
+    0x23, 0x27, 0x87, 0xe9, 0x5e, 0x0d, 0x21, 0xc6, 0xa6, 0xa0, 
+    0x07, 0xcd, 0x90, 0x2e, 0x90, 0x5b, 0x29, 0x0b, 0xcb, 0x15, 
+    0x0c, 0xf9, 0xab, 0xc8, 0xe0, 0x54, 0x48, 0x54, 0x34, 0x8b, 
+    0xa7, 0x9e, 0x52, 0xde, 0x11, 0x4b, 0x30, 0x43, 0xf3, 0x5e, 
+    0xdd, 0x08, 0x16, 0x48, 0x36, 0x71, 0xb8, 0x05, 0x3a, 0x14, 
+    0x2e, 0x5b, 0xcc, 0x2a, 0xd4, 0x88, 0x4e, 0x0f, 0x7e, 0xbc, 
+    0x0c, 0x48, 0x8a, 0x49, 0xbb, 0xbf, 0x92, 0xa8, 0xa7, 0x5c, 
+    0x3b, 0x19, 0x81, 0x9b, 0xac, 0xe8, 0x62, 0x5c, 0xa3, 0x4a, 
+    0x93, 0x0a, 0x74, 0xb9, 0x28, 0x91, 0x2b, 0xae, 0xf6, 0xb7, 
+    0x70, 0x96, 0x6e, 0xd1, 0xa2, 0x39, 0xd1, 0x5d, 0x8a, 0x42, 
+    0x1a, 0xe9, 0x55, 0x84, 0x90, 0xf4, 0xbd, 0x85, 0xcd, 0xcb, 
+    0xa7, 0xfc, 0x36, 0xbf, 0x80, 0xaa, 0x3a, 0x64, 0xd5, 0x6f, 
+    0xeb, 0x2c, 0x31, 0x07, 0xe4, 0xbb, 0x93, 0xcf, 0xc9, 0x89, 
+    0xd9, 0xda, 0xb3, 0x54, 0xc1, 0x97, 0x3a, 0x33, 0xdd, 0x2c, 
+    0x25, 0xd4, 0xfd, 0xf9, 0x91, 0xda, 0xee, 0xea, 0x2a, 0xde, 
+    0x54, 0x48, 0x31, 0xc4, 0xd9, 0xbb, 0x79, 0x2d, 0x36, 0xef, 
+    0x04, 0xa3, 0x02, 0x96, 0x48, 0x49, 0x88, 0xe8, 0x53, 0x65, 
+    0xe3, 0xe2, 0x3b, 0x23, 0xdd, 0x0c, 0x25, 0xd4, 0xfd, 0xb9, 
+    0x81, 0xca, 0xee, 0xea, 0x2a, 0xfe, 0x44, 0x48, 0x31, 0x44, 
+    0xd9, 0xab, 0x69, 0x2d, 0x36, 0xef, 0x04, 0xa3, 0x02, 0x96, 
+    0x48, 0x49, 0x98, 0xf8, 0x53, 0x75, 0xc3, 0xc2, 0xc4, 0xdc, 
+    0x22, 0xf3, 0xda, 0x2b, 0x02, 0x46, 0x7e, 0x35, 0x11, 0x15, 
+    0xd5, 0x01, 0xbb, 0xb7, 0xce, 0xbb, 0x26, 0x54, 0x96, 0xd2, 
+    0xc9, 0x10, 0xfb, 0x5c, 0xfd, 0x69, 0xb7, 0xb6, 0x67, 0x07, 
+    0xac, 0x8a, 0x3c, 0x3d, 0xb8, 0x6d, 0x43, 0x90, 0x58, 0xb3, 
+    0x49, 0xe2, 0x24, 0x01, 0x45, 0x8b, 0x83, 0x6a, 0x73, 0x6b, 
+    0xff, 0x8c, 0xec, 0x11, 0x8b, 0x61, 0xb8, 0x95, 0x7c, 0xe9, 
+    0x20, 0x71, 0xa6, 0x3f, 0x20, 0xd0, 0xb8, 0xa9, 0xd3, 0xf3, 
+    0x97, 0xf1, 0x5d, 0x0f, 0xcd, 0xc2, 0x4b, 0x15, 0x11, 0x7e, 
+    0xca, 0x46, 0xf0, 0xea, 0x90, 0x7d, 0x93, 0x28, 0xb2, 0x0f, 
+    0x6f, 0x63, 0x79, 0x74, 0x3a, 0x3b, 0x07, 0x73, 0x6a, 0x77, 
+    0xba, 0x1c, 0x16, 0xca, 0x3f, 0xde, 0xd9, 0x97, 0x8d, 0x7c, 
+    0x1b, 0xbf, 0xb6, 0x77, 0x21, 0x17, 0x62, 0xc5, 0x8b, 0xce, 
+    0xbc, 0xc6, 0x09, 0xd1, 0xae, 0x1e, 0xd1, 0x5b, 0x2d, 0x7d, 
+    0xa1, 0xad, 0xb5, 0x14, 0x18, 0xef, 0x86, 0xb2, 0xbe, 0x99, 
+    0x46, 0x5f, 0x39, 0x95, 0xfa, 0x67, 0xe6, 0x08, 0xcd, 0xf4, 
+    0xa3, 0x5b, 0x58, 0xa4, 0x3c, 0x71, 0x63, 0xf1, 0x36, 0xcc, 
+    0xaa, 0x5f, 0xda, 0xe4, 0x53, 0x4f, 0x34, 0x15, 0x78, 0xd0, 
+    0xc4, 0xb4, 0x60, 0x91, 0xbb, 0x29, 0x03, 0xb9, 0xf3, 0xf7, 
+    0xc1, 0x96, 0x60, 0xbb, 0xa0, 0x4a, 0xb3, 0xb8, 0x7f, 0x07, 
+    0x30, 0xa8, 0xfb, 0x7f, 0x1d, 0xcb, 0x70, 0x61, 0x52, 0x75, 
+    0x46, 0xb1, 0x2d, 0x47, 0x3d, 0x2f, 0x8a, 0x03, 0x96, 0x1e, 
+    0x2e, 0x13, 0x7b, 0x03, 0xf7, 0xd5, 0xc9, 0xb0, 0x6c, 0xaa, 
+    0x20, 0xc6, 0x86, 0xe2, 0xfe, 0x50, 0xf2, 0x2b, 0x6e, 0xaf, 
+    0x9b, 0xfb, 0xda, 0x41, 0x17, 0xd3, 0x5a, 0x37, 0x30, 0x46, 
+    0xac, 0x70, 0x9a, 0x7b, 0x00, 0x1e, 0x3b, 0x93, 0x77, 0x0e, 
+    0xf3, 0xd5, 0xc9, 0xb0, 0x6c, 0xba, 0x20, 0x46, 0x83, 0xaa, 
+    0xf6, 0x00, 0xf2, 0x2b, 0x6a, 0xaf, 0x9f, 0xcb, 0xda, 0x41, 
+    0x17, 0xd3, 0x5a, 0x37, 0x34, 0x46, 0xac, 0x70, 0x9a, 0x73, 
+    0x00, 0x1e, 0x3b, 0x93, 0x77, 0x0b, 0x0c, 0x2a, 0x36, 0x4f, 
+    0x93, 0x45, 0xdf, 0xb9, 0x7c, 0x55, 0x09, 0xff, 0x0d, 0xd4, 
+    0x95, 0x50, 0x60, 0x34, 0x25, 0xbe, 0xe8, 0x2c, 0xa5, 0xc8, 
+    0xcb, 0xb9, 0x53, 0x8f, 0x65, 0x8c, 0xff, 0xe1, 0xc4, 0x6c, 
+    0x88, 0xf4, 0x05, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x06, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x02, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 
+    0x07, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x02, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 
+    0x04, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x06, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 
+    0x05, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 
+    0x06, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x02, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 
+    0x04, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x02, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 
+    0x06, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x04, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x04, 0x00, 
+    0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x03, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x07, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x02, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 
+    0x04, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x07, 0x00, 
+    0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x03, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 
+    0x0a, 0xa0, 0x1d, 0x4e, 0xe6, 0xac, 0xf3, 0xf3, 0xb4, 0xe8, 
+    0x06, 0x13, 0x46, 0x9b, 0x76, 0xee, 0x9a, 0x6c, 0x83, 0x07, 
+    0x09, 0xd8, 0xd2, 0x20, 0x34, 0xcd, 0xb8, 0x6b, 0x07, 0x94, 
+    0x8b, 0x1c, 0x02, 0x95, 0xb5, 0xb7, 0x20, 0x48, 0x98, 0xfe, 
+    0x48, 0xbf, 0x08, 0xcb, 0x1e, 0xbf, 0x1a, 0x00, 0x72, 0x8e, 
+    0xd9, 0x6f, 0x47, 0xe5, 0xa9, 0x37, 0x64, 0xc4, 0xbf, 0xa4, 
+    0x52, 0x17, 0x27, 0x45, 0x40, 0x43, 0x21, 0xb6, 0x3a, 0xe0, 
+    0xfa, 0x06, 0x9b, 0x43, 0x09, 0x6c, 0x22, 0x64, 0x42, 0xde, 
+    0x3f, 0xdc, 0x6c, 0xd0, 0xcd, 0xac, 0x5d, 0xe7, 0x01, 0x69, 
+    0x01, 0x8f, 0x69, 0x02, 0xdb, 0xa9, 0xc9, 0x93, 0xd7, 0x71, 
+    0x8a, 0x5c, 0xf1, 0x25, 0x3d, 0x34, 0x80, 0x59, 0xeb, 0x54, 
+    0x01, 0x59, 0x52, 0x54, 0x3f, 0x70, 0xa6, 0xd5, 0x56, 0x06, 
+    0x96, 0x0e, 0x1b, 0xb1, 0xc1, 0xe6, 0x95, 0xc5, 0x6e, 0x51, 
+    0xe1, 0x08, 0x77, 0x72, 0x5e, 0xbf, 0xaf, 0xdc, 0x84, 0x51, 
+    0xe1, 0x4c, 0x16, 0xac, 0xcd, 0x34, 0x63, 0x94, 0xc3, 0x56, 
+    0x6b, 0x69, 0x6b, 0x92, 0x7e, 0xa2, 0x96, 0x28, 0xea, 0xa6, 
+    0x21, 0x85, 0x14, 0xee, 0xe9, 0xa3, 0xc5, 0x94, 0x39, 0x60, 
+    0x42, 0xa9, 0xfd, 0xb8, 0x88, 0x45, 0x57, 0x37, 0xad, 0x60, 
+    0xcc, 0x7e, 0x03, 0xa4, 0x61, 0x16, 0x43, 0x2e, 0xf9, 0x99, 
+    0x7e, 0xaa, 0xd3, 0xad, 0xd8, 0xa5, 0x44, 0x83, 0x15, 0xd7, 
+    0xad, 0x85, 0xf7, 0x02, 0x3a, 0x83, 0x96, 0xa3, 0x8c, 0xde, 
+    0x98, 0x35, 0x6d, 0x2c, 0x0d, 0x25, 0xcc, 0x7e, 0x03, 0xa4, 
+    0x61, 0x16, 0x4b, 0x2a, 0xfb, 0x99, 0x7e, 0xa2, 0xd7, 0xad, 
+    0xd8, 0xa5, 0x41, 0x83, 0x15, 0xc7, 0xad, 0x85, 0xf7, 0x02, 
+    0x3a, 0x83, 0x96, 0xab, 0x8c, 0xde, 0x98, 0x75, 0x6d, 0x2c, 
+    0x0d, 0x25, 0x33, 0x81, 0xfc, 0x5b, 0x9e, 0xe9, 0xb4, 0xd5, 
+    0x04, 0x66, 0x81, 0x5d, 0x28, 0x52, 0x27, 0x5a, 0xbe, 0x7c, 
+    0xea, 0x38, 0x52, 0x7a, 0x08, 0xfd, 0xc5, 0x7c, 0x69, 0x54, 
+    0x73, 0x21, 0x67, 0x8a, 0x92, 0xd3, 0xf2, 0xda, 0xbd, 0x2e, 
+    0x77, 0x1e, 0x9e, 0x27, 0x73, 0xb7, 0x0d, 0xb4, 0xcb, 0x16, 
+    0x65, 0x21, 0x94, 0xed, 0xa8, 0x68, 0x2f, 0xa9, 0x0d, 0x6e, 
+    0x15, 0x96, 0x9d, 0xdd, 0x42, 0xba, 0xae, 0x1a, 0x87, 0x6a, 
+    0x1d, 0x5e, 0x58, 0xb9, 0x6d, 0x59, 0xc0, 0x0e, 0xb6, 0x5c, 
+    0x6b, 0x5e, 0xab, 0xc3, 0x99, 0xf5, 0xd2, 0x13, 0xa2, 0xf7, 
+    0x71, 0x1b, 0x54, 0x51, 0x88, 0xe7, 0x2a, 0xab, 0xa6, 0x87, 
+    0x82, 0xf8, 0x80, 0xe6, 0xc7, 0x0d, 0x10, 0x9f, 0xc4, 0x94, 
+    0xfe, 0xd6, 0x92, 0x91, 0x1e, 0x22, 0x92, 0x2c, 0x80, 0xf2, 
+    0xf7, 0x22, 0x55, 0x89, 0x62, 0xd5, 0xcc, 0x31, 0x68, 0xbc, 
+    0x4a, 0x37, 0xc2, 0xd4, 0x90, 0xe1, 0xb4, 0x33, 0x8e, 0xd8, 
+    0x1e, 0x02, 0x42, 0x8a, 0xa3, 0x1d, 0x08, 0x0b, 0x46, 0x78, 
+    0xa9, 0x6c, 0x6d, 0x67, 0x4e, 0x18, 0x36, 0x67, 0x54, 0xf4, 
+    0x3b, 0x2a, 0xf6, 0xf4, 0xd7, 0x6f, 0x49, 0xbb, 0xfa, 0xfa, 
+    0x22, 0xb3, 0x06, 0x05, 0x50, 0xe8, 0x62, 0xc6, 0x4e, 0x50, 
+    0x33, 0x97, 0x38, 0x8d, 0xaa, 0x40, 0xad, 0x7d, 0x70, 0xe0, 
+    0x24, 0x37, 0x34, 0xfa, 0x28, 0x2c, 0xa7, 0x0b, 0xfb, 0x34, 
+    0x42, 0xb2, 0x89, 0x6a, 0xb7, 0x82, 0x87, 0x5d, 0x7b, 0x93, 
+    0x8f, 0xbd, 0xf5, 0x04, 0x15, 0xa9, 0xfb, 0x70, 0x82, 0x85, 
+    0x9b, 0xdc, 0xbd, 0x75, 0xff, 0xf2, 0x03, 0xf3, 0x76, 0xf6, 
+    0x69, 0x2c, 0x10, 0x4a, 0xc6, 0xf4, 0x64, 0x3c, 0x0b, 0x33, 
+    0x37, 0x42, 0xd7, 0x39, 0xfe, 0x92, 0x1d, 0xbf, 0x66, 0x8c, 
+    0xd5, 0x8b, 0x03, 0x76, 0x02, 0x85, 0x8a, 0x58, 0xbd, 0x75, 
+    0xfd, 0xe2, 0x07, 0xf3, 0x36, 0xf6, 0x69, 0x2c, 0x13, 0x4a, 
+    0xc6, 0xb4, 0x66, 0x3c, 0x0b, 0x33, 0x37, 0xc2, 0xd7, 0x39, 
+    0x7a, 0x92, 0x1f, 0xbf, 0x66, 0x8c, 0xd5, 0x8b, 0x23, 0x76, 
+    0xfd, 0x7a, 0x75, 0xa7, 0x42, 0x8a, 0x02, 0x1d, 0xf8, 0x0c, 
+    0xc9, 0x09, 0x96, 0xd3, 0xec, 0xb5, 0x39, 0x4b, 0x99, 0xc3, 
+    0xf4, 0xcc, 0xc8, 0x3d, 0x28, 0xc6, 0x85, 0x6d, 0xe0, 0x40, 
+    0x99, 0x73, 0x2a, 0x74, 0xdc, 0x89, 0x07, 0x00, 0x00, 0x00, 
+    0x07, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x02, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x06, 0x00, 
+    0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 
+    0x05, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x04, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x07, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x06, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x03, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x07, 0x00, 
+    0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 
+    0x03, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x07, 0x00, 
+    0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x06, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x03, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x02, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 
+    0x06, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x03, 0x00, 
+    0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 
+    0x07, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x02, 0x00, 
+    0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x03, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x03, 0x00, 
+    0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x03, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 
+    0x02, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 
+    0x04, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x07, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 
+    0x04, 0x00, 0x00, 0x00, 0xe6, 0x0d, 0x2b, 0x9a, 0x9b, 0xc4, 
+    0xaf, 0xd4, 0x6b, 0x51, 0xfb, 0x3e, 0x14, 0x46, 0xe3, 0xa2, 
+    0xc3, 0xc8, 0x07, 0xb3, 0xd5, 0x54, 0xc4, 0x0a, 0xe0, 0xf5, 
+    0x9c, 0x3c, 0xec, 0xd0, 0x4d, 0x38, 0x79, 0x6a, 0xdf, 0xd3, 
+    0x83, 0xc6, 0xbe, 0x9c, 0xb5, 0xc0, 0xe0, 0x79, 0xc1, 0x85, 
+    0xae, 0x15, 0xe7, 0x14, 0x57, 0x7f, 0x43, 0x67, 0xc8, 0x40, 
+    0x7f, 0xb5, 0xed, 0xa4, 0xd3, 0x44, 0x59, 0x84, 0xef, 0xfc, 
+    0xf1, 0x40, 0x39, 0x7b, 0xd9, 0x17, 0x6e, 0x12, 0xbf, 0x77, 
+    0x61, 0xd1, 0x23, 0x22, 0x36, 0x29, 0x59, 0x45, 0x98, 0x18, 
+    0xc1, 0x0b, 0xad, 0x22, 0x80, 0xf6, 0x46, 0x54, 0xf2, 0x58, 
+    0x98, 0x8b, 0xe5, 0xe3, 0x9c, 0x65, 0x23, 0x6c, 0x27, 0x89, 
+    0xcb, 0xa5, 0x87, 0x94, 0x78, 0x8d, 0x1b, 0xc3, 0x2a, 0x9c, 
+    0xa3, 0xd3, 0x78, 0x32, 0xe1, 0xd1, 0x6d, 0x66, 0x1a, 0x5b, 
+    0xc6, 0xbb, 0x22, 0xa0, 0xbd, 0x40, 0xb4, 0x36, 0x93, 0x61, 
+    0x86, 0x11, 0x7c, 0x11, 0xac, 0xd1, 0x60, 0xec, 0x8d, 0x06, 
+    0xb6, 0x95, 0x7e, 0x78, 0x98, 0x95, 0x14, 0xbe, 0x77, 0x16, 
+    0xf9, 0x64, 0xcd, 0xd8, 0x1d, 0x8a, 0x75, 0x94, 0x69, 0x2e, 
+    0x75, 0x49, 0xa6, 0xe9, 0xb9, 0x8b, 0x84, 0xa1, 0xb4, 0x1a, 
+    0xd7, 0x68, 0xa0, 0xf4, 0x9f, 0x75, 0x30, 0x95, 0xb9, 0x4d, 
+    0x9a, 0xd7, 0x36, 0x92, 0x76, 0x76, 0xb6, 0x50, 0xe8, 0x4e, 
+    0x17, 0x1a, 0x40, 0xfd, 0x69, 0x7a, 0xef, 0x6a, 0xba, 0x66, 
+    0xf5, 0xc9, 0x14, 0xd5, 0xfc, 0x52, 0x0f, 0x29, 0xc5, 0xf4, 
+    0x8f, 0x55, 0x30, 0x95, 0xb9, 0x4d, 0x9a, 0xd5, 0x36, 0x92, 
+    0x76, 0x76, 0xb6, 0x50, 0xc9, 0x4e, 0x17, 0x1a, 0x44, 0xbd, 
+    0x69, 0x78, 0xef, 0x6b, 0xba, 0x66, 0xf5, 0xc9, 0x84, 0xd5, 
+    0xfc, 0x52, 0x07, 0x29, 0xc5, 0xf4, 0x70, 0xaa, 0xcf, 0x6a, 
+    0x46, 0xb2, 0x65, 0x2a, 0xc9, 0x6d, 0x89, 0x89, 0x49, 0xaf, 
+    0x36, 0xb1, 0xe8, 0xe5, 0xbb, 0x42, 0x96, 0x87, 0x10, 0x94, 
+    0x45, 0x99, 0x0a, 0x36, 0x7b, 0x2a, 0x03, 0xad, 0xf8, 0xd6, 
+    0x3a, 0x0b, 0x0f, 0x22, 0x47, 0x5f, 0x03, 0x9b, 0x47, 0xe0, 
+    0x43, 0x0a, 0x79, 0x2a, 0x0a, 0xf2, 0xa5, 0x57, 0xcf, 0x22, 
+    0x3f, 0x66, 0x7d, 0x8e, 0x5c, 0x5a, 0xd2, 0x27, 0x86, 0x41, 
+    0x21, 0x57, 0x9d, 0xed, 0xdc, 0xe4, 0x1a, 0xfd, 0x53, 0x00, 
+    0x25, 0x92, 0x34, 0x4c, 0x02, 0xd9, 0x43, 0x12, 0x39, 0x07, 
+    0x3f, 0x8c, 0xd6, 0x65, 0xe0, 0xed, 0xf3, 0xae, 0x6b, 0x7c, 
+    0x7e, 0xb3, 0x17, 0x8d, 0xce, 0xc7, 0xc8, 0x69, 0x3f, 0x94, 
+    0x20, 0x21, 0xdf, 0xee, 0x37, 0x57, 0x27, 0x50, 0x30, 0x8a, 
+    0x21, 0x06, 0x49, 0xbd, 0xbc, 0x9e, 0x0b, 0xb5, 0x0e, 0x37, 
+    0xda, 0x4d, 0x41, 0xa0, 0xfd, 0xbe, 0xf2, 0x48, 0x70, 0x55, 
+    0x49, 0x31, 0xb8, 0x74, 0xda, 0x09, 0x67, 0xbb, 0x1c, 0xd8, 
+    0x51, 0xfb, 0xf8, 0xdd, 0xef, 0x02, 0xb2, 0x84, 0x3e, 0xcb, 
+    0x6c, 0x51, 0x03, 0xd2, 0xfd, 0xde, 0x48, 0x37, 0x39, 0x2c, 
+    0xff, 0x1a, 0x94, 0x15, 0x66, 0x3d, 0x99, 0xcb, 0x7b, 0xcd, 
+    0xfb, 0x77, 0x15, 0xcc, 0x5d, 0x81, 0xc8, 0x28, 0x65, 0x78, 
+    0x6d, 0xc0, 0x35, 0xa0, 0xd8, 0xaa, 0xea, 0xa0, 0xf4, 0xa6, 
+    0x85, 0xed, 0x44, 0x8b, 0x5a, 0x08, 0xf6, 0x13, 0x8c, 0x47, 
+    0xbc, 0x9f, 0xdb, 0x24, 0x25, 0x25, 0xe7, 0x67, 0xda, 0x22, 
+    0x0a, 0xb7, 0xc2, 0x07, 0xed, 0x49, 0x69, 0x00, 0x25, 0x21, 
+    0xeb, 0x8e, 0x64, 0x30, 0x6d, 0xc6, 0xb5, 0xf7, 0x75, 0x83, 
+    0x96, 0xa9, 0xd3, 0x6b, 0x96, 0x37, 0x5c, 0x3f, 0x7b, 0x2a, 
+    0x68, 0x64, 0x07, 0xec, 0xdd, 0x22, 0x1d, 0xe5, 0x42, 0x2b, 
+    0xed, 0x49, 0x69, 0x00, 0x35, 0x20, 0xeb, 0x8e, 0x6c, 0x20, 
+    0x6d, 0xc6, 0xb5, 0xe7, 0x75, 0x03, 0x12, 0xa8, 0xf7, 0x6b, 
+    0x86, 0x07, 0x5c, 0x3f, 0x7b, 0x2a, 0x68, 0x64, 0x27, 0xee, 
+    0xdd, 0x22, 0x1d, 0xe5, 0xbd, 0xd4, 0x12, 0xb6, 0x96, 0xff, 
+    0xca, 0xdf, 0x14, 0x71, 0x93, 0xdf, 0x92, 0x39, 0x4a, 0x18, 
+    0x8a, 0xfc, 0xed, 0x57, 0x08, 0x94, 0x79, 0xf8, 0xa3, 0xc0, 
+    0x84, 0xd5, 0x97, 0x9b, 0xd8, 0x11, 0x22, 0xdd, 0xe2, 0x1a, 
+    0x06, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x02, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 
+    0x05, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x06, 0x00, 
+    0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 
+    0x03, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 
+    0x07, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x02, 0x00, 
+    0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x02, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x04, 0x00, 
+    0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 
+    0x02, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x04, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 
+    0x05, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 
+    0x02, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x02, 0x00, 
+    0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 
+    0x03, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x07, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x04, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x04, 0x00, 
+    0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 
+    0x05, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 
+    0x05, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x02, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x04, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x37, 0x30, 
+    0x87, 0xb8, 0xd1, 0x6b, 0x60, 0x87, 0x74, 0x62, 0x3a, 0xcd, 
+    0x13, 0x86, 0xf1, 0x2d, 0x3e, 0xb6, 0x4c, 0x04, 0x70, 0xd9, 
+    0xbc, 0xc0, 0xd7, 0x77, 0x23, 0xca, 0x07, 0x13, 0x84, 0xe6, 
+    0xd1, 0xe1, 0x7a, 0x7a, 0xc3, 0xa7, 0x72, 0xc1, 0x9a, 0x72, 
+    0x48, 0xb3, 0x4d, 0x66, 0xfd, 0xdd, 0x81, 0x56, 0xc8, 0x6f, 
+    0x47, 0xf5, 0x76, 0x9c, 0xd0, 0xe5, 0x60, 0x12, 0x5d, 0x90, 
+    0xab, 0xbf, 0xb8, 0x32, 0x2e, 0xc0, 0x97, 0xa7, 0x1f, 0x94, 
+    0x95, 0xe2, 0x05, 0x00, 0x80, 0x35, 0x03, 0xf8, 0x2e, 0x5f, 
+    0xb1, 0xde, 0xa7, 0xbf, 0x81, 0xa9, 0xe6, 0xe4, 0x23, 0xe7, 
+    0x88, 0xac, 0x87, 0x05, 0x33, 0x8c, 0x9b, 0xaa, 0x29, 0x5a, 
+    0x92, 0xb4, 0x59, 0x8c, 0x94, 0xbe, 0xe7, 0x13, 0x82, 0xef, 
+    0x8f, 0x87, 0x0d, 0x92, 0xbb, 0xa8, 0xdb, 0x22, 0x0f, 0x92, 
+    0x19, 0x50, 0xaa, 0xfd, 0x02, 0x04, 0x0b, 0x56, 0x15, 0xd5, 
+    0x7d, 0x22, 0xa6, 0x70, 0x30, 0xf8, 0x83, 0x31, 0x36, 0x37, 
+    0xaa, 0x13, 0xb0, 0xf7, 0xbb, 0xad, 0xa8, 0x84, 0x18, 0x40, 
+    0xad, 0xe7, 0x83, 0x34, 0xec, 0x97, 0x5d, 0x51, 0x49, 0x85, 
+    0xc9, 0x75, 0xc1, 0xda, 0x49, 0x48, 0xd6, 0x84, 0xb6, 0x7c, 
+    0x9b, 0x28, 0xae, 0xe4, 0x51, 0xdf, 0x98, 0xec, 0x84, 0xf3, 
+    0x9a, 0xc4, 0x8a, 0xc0, 0xfa, 0xd8, 0x89, 0xf7, 0x83, 0x26, 
+    0xca, 0xa7, 0xdf, 0xd5, 0x4e, 0xa6, 0x92, 0x35, 0xc5, 0x72, 
+    0x25, 0x6e, 0xc2, 0x04, 0xf6, 0x6c, 0x3e, 0x59, 0x89, 0x8c, 
+    0x5b, 0x5c, 0xb0, 0x97, 0x84, 0xf7, 0xba, 0xc5, 0x8a, 0x80, 
+    0x3a, 0xda, 0x89, 0xf3, 0x83, 0x26, 0xca, 0x87, 0xdd, 0xd5, 
+    0x4e, 0xa4, 0x92, 0x75, 0xc5, 0x72, 0x25, 0x4e, 0xc2, 0x04, 
+    0xf6, 0x6c, 0x3a, 0x59, 0x89, 0xa4, 0x5b, 0x5e, 0xb0, 0x97, 
+    0x7b, 0x08, 0x45, 0x3a, 0x75, 0x7f, 0xc5, 0x25, 0x76, 0x0c, 
+    0x7c, 0xd9, 0x35, 0x78, 0x22, 0x2a, 0xb1, 0x5b, 0x6d, 0x8a, 
+    0x3a, 0x8d, 0xda, 0xb1, 0x3d, 0xfb, 0x09, 0x93, 0xc5, 0xa6, 
+    0x76, 0x5b, 0xa4, 0xa1, 0x4f, 0x68, 0x9f, 0x12, 0x29, 0xa2, 
+    0xd0, 0x4a, 0x7a, 0xc4, 0xdb, 0x8f, 0xbc, 0x33, 0x80, 0x1b, 
+    0x55, 0xb1, 0x91, 0xf3, 0x20, 0xbd, 0x25, 0x9e, 0x26, 0x00, 
+    0x84, 0xf7, 0x17, 0x6a, 0xbc, 0xda, 0x4a, 0xc8, 0x9b, 0x14, 
+    0xf6, 0xc5, 0xe2, 0xff, 0xd2, 0x57, 0x2d, 0xd8, 0xd0, 0x45, 
+    0x87, 0xee, 0xfc, 0x05, 0xa7, 0x7b, 0x64, 0x52, 0x7c, 0xde, 
+    0xfb, 0x86, 0x11, 0x25, 0xb2, 0x09, 0x4c, 0xa7, 0x35, 0x2d, 
+    0x67, 0xa4, 0xe6, 0xd0, 0xad, 0x6c, 0xeb, 0xd0, 0x8e, 0x86, 
+    0x1b, 0xf4, 0x5f, 0x02, 0xd5, 0x8c, 0xc1, 0xf5, 0x48, 0x63, 
+    0x07, 0x70, 0x68, 0x78, 0x16, 0xca, 0x2c, 0x51, 0xf6, 0x95, 
+    0xc4, 0x20, 0x0f, 0xa6, 0x13, 0x7b, 0xf9, 0x81, 0x10, 0xd1, 
+    0x86, 0x66, 0x92, 0xd9, 0x95, 0x64, 0xfc, 0x9e, 0xfa, 0x2e, 
+    0x71, 0x1c, 0x00, 0xd4, 0x86, 0x91, 0xfd, 0xf7, 0x4b, 0x24, 
+    0x97, 0x8f, 0x6c, 0x0d, 0x5a, 0x48, 0xc7, 0xc0, 0xa8, 0x4e, 
+    0xaf, 0xf9, 0x50, 0xa4, 0xa3, 0x48, 0x90, 0xb6, 0x71, 0x41, 
+    0xb5, 0xea, 0x77, 0x50, 0x72, 0x66, 0x7c, 0xfb, 0xcb, 0x82, 
+    0x71, 0x8a, 0x01, 0xcb, 0xcf, 0x6a, 0x4b, 0x16, 0x33, 0x50, 
+    0x27, 0xf8, 0xde, 0xbc, 0xdc, 0xb1, 0x2a, 0xe9, 0x3a, 0xdb, 
+    0x7a, 0xe9, 0xac, 0x90, 0x24, 0x18, 0x18, 0xb7, 0x37, 0x14, 
+    0xe6, 0x5c, 0xec, 0xec, 0x9a, 0x5e, 0xa1, 0xa3, 0x01, 0x9b, 
+    0x4f, 0x7a, 0x5f, 0xdc, 0x2b, 0x78, 0xe0, 0xd8, 0x0e, 0xb1, 
+    0xd8, 0xa7, 0xea, 0xe9, 0xab, 0x92, 0x22, 0xeb, 0x95, 0xdc, 
+    0x26, 0x19, 0x9c, 0xb7, 0x37, 0x54, 0xf6, 0x5c, 0xec, 0xec, 
+    0x8a, 0x16, 0xe1, 0x83, 0x01, 0xdb, 0x4f, 0x7a, 0x5f, 0x9e, 
+    0x2b, 0x58, 0x60, 0xd8, 0x8e, 0xb1, 0xd8, 0xa7, 0x6a, 0xe9, 
+    0x2b, 0x92, 0x22, 0xeb, 0x94, 0xdc, 0x26, 0x08, 0x63, 0x48, 
+    0xc8, 0xab, 0x09, 0xa3, 0x13, 0x13, 0x75, 0xe9, 0x1e, 0x7c, 
+    0xfe, 0x24, 0xb0, 0x85, 0xa0, 0x61, 0xd4, 0xa7, 0x9f, 0x27, 
+    0x71, 0x4e, 0x27, 0x58, 0x95, 0x16, 0xd4, 0x6d, 0xdd, 0x14, 
+    0x6b, 0x23, 0xd9, 0xf7, 0x03, 0x00, 0x00, 0x00, 0x06, 0x00, 
+    0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 
+    0x04, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x03, 0x00, 
+    0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 
+    0x02, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x03, 0x00, 
+    0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x02, 0x00, 
+    0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x07, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x07, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x02, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 
+    0x05, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x04, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x02, 0x00, 
+    0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 
+    0x06, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 
+    0x05, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x03, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0xac, 0xab, 0x1b, 0x4d, 0x4b, 0x09, 0x00, 0x7c, 
+    0x35, 0x9c, 0x96, 0x3f, 0x7b, 0x0a, 0x7b, 0x5b, 0xbe, 0x6b, 
+    0x2b, 0x18, 0x85, 0x78, 0x28, 0x62, 0x3d, 0x42, 0x2c, 0x48, 
+    0x22, 0x69, 0x4b, 0xaf, 0xa9, 0xa4, 0x1d, 0xb5, 0x60, 0xeb, 
+    0x29, 0x4f, 0xc6, 0x73, 0x44, 0x78, 0x46, 0x62, 0xeb, 0x40, 
+    0xfe, 0xf7, 0x05, 0x44, 0xe6, 0xc3, 0xf6, 0x68, 0x70, 0x6d, 
+    0x7a, 0x41, 0xc8, 0x18, 0x6f, 0xc1, 0x5e, 0x4b, 0x46, 0xad, 
+    0x55, 0x4e, 0xd1, 0x40, 0x1e, 0x80, 0x5f, 0x0d, 0xbf, 0xe5, 
+    0x4e, 0x79, 0x10, 0x68, 0x39, 0xfb, 0x27, 0x1b, 0x25, 0x25, 
+    0x99, 0xd3, 0x26, 0xed, 0x83, 0xfd, 0x75, 0x94, 0xf9, 0xe3, 
+    0xca, 0x09, 0xa7, 0xd9, 0xa6, 0xc6, 0x9f, 0x26, 0xc9, 0x0a, 
+    0xe5, 0x42, 0xea, 0x88, 0x85, 0xf8, 0xbf, 0xc6, 0xcb, 0xb5, 
+    0x05, 0x1b, 0xfb, 0x89, 0xef, 0x1c, 0xd6, 0x91, 0x62, 0xef, 
+    0x26, 0x52, 0x13, 0x3c, 0x94, 0x8b, 0xd9, 0x93, 0x37, 0x16, 
+    0xb0, 0xa9, 0x17, 0x89, 0x95, 0xa4, 0xc9, 0xa6, 0x44, 0x9f, 
+    0x23, 0x54, 0xa7, 0x66, 0x21, 0x79, 0x66, 0x39, 0x2a, 0x95, 
+    0x25, 0xb1, 0x6d, 0x8c, 0x3f, 0x1e, 0x90, 0x94, 0x2f, 0xe1, 
+    0x5a, 0x40, 0xb5, 0x18, 0x67, 0x51, 0x4a, 0xec, 0xf0, 0xae, 
+    0xc3, 0xb6, 0xcc, 0xcc, 0xc8, 0x9e, 0xe1, 0x56, 0x8f, 0xf6, 
+    0x73, 0xbb, 0x70, 0x6b, 0x68, 0x18, 0xad, 0x16, 0x3c, 0x84, 
+    0x6e, 0xc5, 0x98, 0x50, 0x2f, 0x3b, 0x9f, 0x0a, 0x98, 0x19, 
+    0x77, 0x79, 0x4a, 0xac, 0xd3, 0xef, 0xc4, 0xe1, 0xcd, 0xcc, 
+    0xcc, 0x9f, 0xe1, 0x54, 0x8f, 0xf6, 0x73, 0xb9, 0x70, 0x6b, 
+    0x68, 0x18, 0xad, 0x14, 0x3c, 0x84, 0x7e, 0xcd, 0x90, 0x50, 
+    0x2f, 0x72, 0x9e, 0x0b, 0x98, 0x19, 0x67, 0x79, 0x4a, 0xac, 
+    0xd3, 0xee, 0xc5, 0xa0, 0x32, 0x33, 0x33, 0x60, 0x1e, 0xab, 
+    0x70, 0x09, 0x8c, 0x46, 0x8f, 0x94, 0x97, 0xe7, 0x52, 0xeb, 
+    0xc3, 0x7b, 0x81, 0x32, 0x6f, 0xaf, 0xd0, 0x8d, 0x61, 0xf4, 
+    0x67, 0xe6, 0x98, 0x86, 0xb5, 0x53, 0x2c, 0x11, 0x3a, 0x5f, 
+    0x7d, 0xdf, 0x3e, 0xa5, 0x71, 0x07, 0xbd, 0x83, 0x2b, 0x76, 
+    0x2d, 0xd9, 0x10, 0x4c, 0x3f, 0xda, 0x5c, 0x20, 0xab, 0x77, 
+    0x67, 0x93, 0x85, 0xd5, 0x10, 0x6d, 0x4b, 0xaf, 0x32, 0x3a, 
+    0x4d, 0x79, 0x2b, 0xbc, 0x6c, 0xd7, 0x9d, 0x18, 0x35, 0x0a, 
+    0xe3, 0x8b, 0x5d, 0xa2, 0x47, 0x43, 0xd2, 0xd7, 0xe6, 0xb3, 
+    0xe6, 0x47, 0xe4, 0x74, 0xae, 0x7d, 0x74, 0xd1, 0x74, 0x64, 
+    0x0d, 0xe7, 0xc6, 0xc2, 0xcf, 0x2e, 0x78, 0xa6, 0xd5, 0xa2, 
+    0x0b, 0x32, 0x57, 0x13, 0xbf, 0xc1, 0x6b, 0x4e, 0x63, 0x24, 
+    0xd1, 0x8e, 0x53, 0x09, 0x48, 0x85, 0x2d, 0x3f, 0x0e, 0x15, 
+    0xd8, 0x20, 0x69, 0xb2, 0x06, 0xe4, 0xca, 0x11, 0x40, 0xdb, 
+    0x48, 0xef, 0x01, 0x5f, 0xbe, 0x8e, 0x31, 0xf6, 0x93, 0x1d, 
+    0xa7, 0x0f, 0x7a, 0x41, 0x7e, 0x00, 0x4c, 0x65, 0xb0, 0x2b, 
+    0x4a, 0xed, 0x1b, 0xe2, 0x80, 0x4a, 0x1c, 0x8b, 0x92, 0x66, 
+    0x20, 0x36, 0x23, 0xfe, 0x28, 0x14, 0xb6, 0x00, 0x6a, 0x7d, 
+    0xfd, 0x61, 0x18, 0x44, 0x0f, 0xcb, 0x28, 0xc6, 0x9f, 0x27, 
+    0x5d, 0xa2, 0x1c, 0xf2, 0x6c, 0x45, 0x28, 0x67, 0xdd, 0xbe, 
+    0x7f, 0xf4, 0x9e, 0x2d, 0x25, 0xd3, 0xa7, 0xad, 0x60, 0xce, 
+    0x4f, 0x4e, 0xa8, 0x3b, 0x6d, 0x37, 0xb6, 0x36, 0x94, 0x0c, 
+    0x8e, 0x39, 0xf2, 0xa2, 0xdf, 0x39, 0x57, 0xb0, 0x6d, 0x7e, 
+    0x8b, 0x41, 0x69, 0xd3, 0xdd, 0xee, 0x14, 0x70, 0x3a, 0x7d, 
+    0x21, 0xd0, 0x83, 0x85, 0x20, 0xec, 0x02, 0x4c, 0xae, 0x36, 
+    0x69, 0xbe, 0x16, 0x3f, 0x90, 0x17, 0x8e, 0x19, 0xf2, 0x22, 
+    0xdf, 0x39, 0x57, 0xb0, 0x4d, 0x7e, 0x8a, 0x41, 0x6b, 0xd3, 
+    0xdd, 0xee, 0x14, 0x70, 0x3a, 0x3d, 0x21, 0xd1, 0x8b, 0xa4, 
+    0x20, 0xcc, 0x02, 0x4c, 0xac, 0x36, 0x69, 0xbe, 0x16, 0x3f, 
+    0x90, 0x17, 0x71, 0xe6, 0x0d, 0xdd, 0x20, 0xc6, 0xa8, 0x4f, 
+    0xb2, 0x81, 0x75, 0xbe, 0x94, 0x2c, 0x22, 0x11, 0xeb, 0x8f, 
+    0xc5, 0xc2, 0xde, 0x2e, 0x7c, 0x5b, 0xdf, 0x33, 0xfd, 0xb3, 
+    0x53, 0xc9, 0x96, 0x41, 0xe9, 0xc0, 0x6f, 0xe8, 0x02, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x04, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 
+    0x07, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x03, 0x00, 
+    0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 
+    0x06, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x05, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x06, 0x00, 
+    0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 
+    0x04, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x06, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 
+    0x05, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x06, 0x00, 
+    0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x07, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 
+    0x05, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 
+    0x02, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x06, 0x00, 
+    0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x03, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x03, 0x00, 
+    0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x05, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x06, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 
+    0x02, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x02, 0x00, 
+    0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x07, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x1b, 0xd1, 0x2c, 0xe5, 
+    0xf5, 0xbc, 0x58, 0x46, 0x30, 0xa8, 0xaa, 0x6f, 0xfe, 0x37, 
+    0x2c, 0x5c, 0xd3, 0xb0, 0x2d, 0x72, 0x3f, 0x29, 0xbd, 0xe5, 
+    0x53, 0x02, 0x3e, 0x51, 0xec, 0x64, 0x98, 0xe2, 0x6f, 0xfa, 
+    0x73, 0x17, 0xf2, 0x84, 0x08, 0x4b, 0x51, 0xb1, 0xbf, 0xf1, 
+    0xc5, 0x81, 0x01, 0x8a, 0xb4, 0x78, 0x42, 0x06, 0x62, 0x09, 
+    0x9b, 0x93, 0x7f, 0xaa, 0xc6, 0xb1, 0x49, 0x66, 0x44, 0x00, 
+    0xa0, 0xd5, 0x04, 0x34, 0x1d, 0x06, 0x4f, 0xc9, 0xea, 0x5c, 
+    0x7d, 0x7b, 0x22, 0x96, 0xc8, 0x26, 0x93, 0xe7, 0x5a, 0xc0, 
+    0xd4, 0x67, 0x1c, 0x39, 0xf1, 0x7f, 0x86, 0x4e, 0xd3, 0x76, 
+    0xd9, 0x62, 0xcc, 0x7c, 0x6f, 0xc3, 0x9a, 0x61, 0xf7, 0xb3, 
+    0x23, 0x67, 0x25, 0x46, 0xea, 0xcf, 0x62, 0x04, 0x8e, 0x9d, 
+    0x16, 0xac, 0x93, 0xec, 0x38, 0x41, 0xa5, 0xd1, 0xfc, 0x77, 
+    0x51, 0x11, 0x9f, 0x55, 0x2a, 0x4e, 0x2d, 0xd9, 0x4d, 0xb7, 
+    0xed, 0x50, 0x69, 0x9e, 0x94, 0x94, 0x1c, 0x50, 0x1e, 0xbe, 
+    0x28, 0x51, 0x32, 0x72, 0x8a, 0xa2, 0xfb, 0xe1, 0x27, 0x05, 
+    0xc6, 0x56, 0x2a, 0x22, 0xfd, 0xac, 0x4f, 0x78, 0x13, 0xca, 
+    0xda, 0x10, 0x07, 0x92, 0x50, 0xf4, 0xd9, 0x5a, 0x02, 0x03, 
+    0x9d, 0x8e, 0x94, 0x56, 0x2a, 0xc0, 0xa0, 0x54, 0x31, 0x52, 
+    0x8a, 0xa3, 0xde, 0x68, 0x97, 0x1f, 0x6c, 0x66, 0xad, 0x83, 
+    0x9c, 0xfc, 0x1c, 0xb9, 0x61, 0x21, 0xca, 0xdc, 0x05, 0xda, 
+    0x49, 0x50, 0xf0, 0x5c, 0x1f, 0xf0, 0x1c, 0x9d, 0x80, 0x1a, 
+    0x80, 0x2b, 0xa0, 0x50, 0x30, 0x52, 0x8a, 0xa3, 0xde, 0xe8, 
+    0x97, 0x1f, 0x6c, 0x66, 0xad, 0x83, 0x9c, 0xfc, 0x5d, 0xb9, 
+    0x61, 0x21, 0xca, 0xdc, 0x05, 0xda, 0x41, 0x50, 0xf9, 0x5c, 
+    0x1f, 0xf0, 0x9c, 0x9c, 0x14, 0x12, 0x08, 0x2b, 0x5f, 0xaf, 
+    0xcf, 0xad, 0x75, 0x5c, 0x21, 0x17, 0x68, 0xe0, 0x93, 0x99, 
+    0x52, 0x7c, 0x63, 0x03, 0xa2, 0x46, 0x9e, 0xde, 0x35, 0x23, 
+    0xfa, 0x25, 0xbe, 0xaf, 0x06, 0xa3, 0xe0, 0x0f, 0x63, 0x63, 
+    0xeb, 0xed, 0xf7, 0xd4, 0x69, 0x1a, 0x0a, 0x90, 0x22, 0x49, 
+    0x3c, 0x6b, 0xb2, 0x1b, 0xe8, 0x9e, 0x45, 0x99, 0xb1, 0x2c, 
+    0x17, 0x5c, 0x01, 0xb4, 0x58, 0xa9, 0x7b, 0x49, 0x2d, 0xab, 
+    0xd7, 0x25, 0x7b, 0x99, 0xa0, 0x97, 0x8a, 0xc3, 0x7b, 0x7b, 
+    0x18, 0xcc, 0x39, 0x37, 0xc0, 0x2e, 0x2d, 0xbc, 0xe9, 0x24, 
+    0xd8, 0xc8, 0x2b, 0x01, 0xb9, 0xe2, 0x06, 0x27, 0x71, 0x7c, 
+    0xef, 0x08, 0x9a, 0x2d, 0x42, 0xb7, 0x6f, 0x4b, 0x86, 0x93, 
+    0xfa, 0xeb, 0xf7, 0x9e, 0x75, 0x3f, 0xea, 0x7b, 0xbc, 0xbe, 
+    0xb2, 0x5e, 0xe9, 0x69, 0xa8, 0x60, 0x46, 0x05, 0xb5, 0xaa, 
+    0x2d, 0xd3, 0xfa, 0x47, 0x6e, 0x48, 0xed, 0x9f, 0xd8, 0x33, 
+    0xe1, 0xf8, 0xda, 0x57, 0x6a, 0xf4, 0x8e, 0x18, 0x2c, 0xf0, 
+    0x1a, 0xe8, 0x67, 0xc9, 0x03, 0x2c, 0x23, 0x46, 0xd1, 0x5b, 
+    0xac, 0x4c, 0xd5, 0x4a, 0x2c, 0xfb, 0x9c, 0x2e, 0x27, 0xd3, 
+    0x51, 0x94, 0x14, 0x25, 0x4a, 0xaf, 0x7e, 0xb8, 0xbe, 0x0b, 
+    0x3f, 0x3d, 0x33, 0x71, 0x15, 0x2b, 0x13, 0xf9, 0x42, 0x37, 
+    0xb6, 0xe9, 0x3c, 0x1b, 0x26, 0xe2, 0x32, 0xd8, 0xaa, 0x05, 
+    0xe5, 0x27, 0xc3, 0x9f, 0x80, 0x03, 0xf8, 0x55, 0x50, 0x1c, 
+    0x31, 0x74, 0x85, 0x29, 0x98, 0x71, 0xcd, 0x75, 0xf8, 0xf3, 
+    0xf5, 0xd6, 0x65, 0xb4, 0x4a, 0x1c, 0x2f, 0xed, 0xa5, 0xfb, 
+    0x60, 0xe1, 0x70, 0x91, 0xb2, 0x07, 0x3f, 0x63, 0x89, 0xdd, 
+    0xce, 0xc9, 0x79, 0x5e, 0x74, 0x01, 0x10, 0x75, 0x03, 0x3c, 
+    0xbc, 0x25, 0x06, 0x0d, 0x2d, 0xf1, 0x15, 0x82, 0xe9, 0x36, 
+    0x4a, 0x1c, 0x2f, 0xed, 0xac, 0x7b, 0x60, 0xe1, 0x70, 0xd0, 
+    0xb2, 0x07, 0x37, 0x63, 0x89, 0xdf, 0xc6, 0xcb, 0x78, 0x5e, 
+    0x74, 0x01, 0x10, 0x75, 0x03, 0x38, 0xbc, 0x27, 0x07, 0x0d, 
+    0x2d, 0xf1, 0x15, 0x82, 0x69, 0x36, 0xb5, 0xe3, 0xd0, 0x12, 
+    0x53, 0x84, 0x9f, 0x1e, 0x8f, 0x2f, 0x4d, 0xf8, 0xc8, 0x9c, 
+    0x76, 0x20, 0x39, 0x34, 0x87, 0xa1, 0x8b, 0xfe, 0xef, 0x8a, 
+    0xfc, 0xc7, 0x43, 0xd8, 0xf8, 0xf2, 0xd2, 0x0e, 0xea, 0x7d, 
+    0x96, 0xc9, 0x04, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x07, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x06, 0x00, 
+    0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x04, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x04, 0x00, 
+    0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 
+    0x02, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x02, 0x00, 
+    0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x07, 0x00, 
+    0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 
+    0x06, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x03, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 
+    0x05, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x03, 0x00, 
+    0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 
+    0x07, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x02, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 
+    0x03, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x03, 0x00, 
+    0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 
+    0x06, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00
+};
+
diff --git a/neureka/depthwise/Makefile b/neureka/depthwise/Makefile
new file mode 100644
index 0000000..9c8efb5
--- /dev/null
+++ b/neureka/depthwise/Makefile
@@ -0,0 +1,3 @@
+include ../app/Makefile
+
+STIM_DIR := ../depthwise/
\ No newline at end of file
diff --git a/neureka/depthwise/inc/bias.h b/neureka/depthwise/inc/bias.h
new file mode 100644
index 0000000..7466de0
--- /dev/null
+++ b/neureka/depthwise/inc/bias.h
@@ -0,0 +1,9 @@
+#ifndef __BIAS_H__
+#define __BIAS_H__
+
+#include <pmsis.h>
+
+#define BIAS_SIZE (67)
+extern int32_t bias[BIAS_SIZE];
+
+#endif  // __BIAS_H__
diff --git a/neureka/depthwise/inc/input.h b/neureka/depthwise/inc/input.h
new file mode 100644
index 0000000..ddb3845
--- /dev/null
+++ b/neureka/depthwise/inc/input.h
@@ -0,0 +1,9 @@
+#ifndef __INPUT_H__
+#define __INPUT_H__
+
+#include <pmsis.h>
+
+#define INPUT_SIZE (1407)
+extern uint8_t input[INPUT_SIZE];
+
+#endif  // __INPUT_H__
diff --git a/neureka/depthwise/inc/layer_conf.h b/neureka/depthwise/inc/layer_conf.h
new file mode 100644
index 0000000..a1e9326
--- /dev/null
+++ b/neureka/depthwise/inc/layer_conf.h
@@ -0,0 +1,42 @@
+#ifndef __LAYER_CONF_H__
+#define __LAYER_CONF_H__
+
+#define TEST_NAME "test"
+#define INPUT_HEIGHT (7)
+#define INPUT_WIDTH (3)
+#define INPUT_CHANNEL (67)
+#define INPUT_SIGNED (0)
+#define INPUT_BITS (8)
+
+#define OUTPUT_HEIGHT (5)
+#define OUTPUT_WIDTH (1)
+#define OUTPUT_CHANNEL (67)
+#define OUTPUT_BITS (8)
+
+#define WEIGHT_HEIGHT (3)
+#define WEIGHT_WIDTH (3)
+#define WEIGHT_CHANNEL_IN (1)
+#define WEIGHT_CHANNEL_OUT (67)
+#define WEIGHT_BITS (8)
+#define WEIGHT_OFFSET (-128)
+
+#define SCALE_BITS (8)
+
+#define BIAS_BITS (32)
+
+#define PADDING_TOP (0)
+#define PADDING_BOTTOM (0)
+#define PADDING_LEFT (0)
+#define PADDING_RIGHT (0)
+#define PADDING_VALUE (0)
+
+#define STRIDE_HEIGHT (1)
+#define STRIDE_WIDTH (1)
+
+#define GROUPS (67)
+#define OUTSHIFT (10)
+#define HAS_NORM_QUANT (1)
+#define HAS_BIAS (1)
+#define HAS_RELU (1)
+
+#endif  // __LAYER_CONF_H__
diff --git a/neureka/depthwise/inc/output.h b/neureka/depthwise/inc/output.h
new file mode 100644
index 0000000..43b9b83
--- /dev/null
+++ b/neureka/depthwise/inc/output.h
@@ -0,0 +1,14 @@
+#ifndef __OUTPUT_H__
+#define __OUTPUT_H__
+
+#include <pmsis.h>
+
+#define OUTPUT_SIZE (335)
+extern uint8_t output[OUTPUT_SIZE];
+
+#define GOLDEN_OUTPUT_SIZE (335)
+extern uint8_t golden_output[GOLDEN_OUTPUT_SIZE];
+
+int check_output();
+
+#endif  // __OUTPUT_H__
diff --git a/neureka/depthwise/inc/scale.h b/neureka/depthwise/inc/scale.h
new file mode 100644
index 0000000..cced0eb
--- /dev/null
+++ b/neureka/depthwise/inc/scale.h
@@ -0,0 +1,9 @@
+#ifndef __SCALE_H__
+#define __SCALE_H__
+
+#include <pmsis.h>
+
+#define SCALE_SIZE (67)
+extern uint8_t scale[SCALE_SIZE];
+
+#endif  // __SCALE_H__
diff --git a/neureka/depthwise/inc/weight.h b/neureka/depthwise/inc/weight.h
new file mode 100644
index 0000000..82efe1e
--- /dev/null
+++ b/neureka/depthwise/inc/weight.h
@@ -0,0 +1,9 @@
+#ifndef __WEIGHT_H__
+#define __WEIGHT_H__
+
+#include <pmsis.h>
+
+#define WEIGHT_SIZE (864)
+extern uint8_t weight[WEIGHT_SIZE];
+
+#endif  // __WEIGHT_H__
diff --git a/neureka/depthwise/src/bias.c b/neureka/depthwise/src/bias.c
new file mode 100644
index 0000000..32abcea
--- /dev/null
+++ b/neureka/depthwise/src/bias.c
@@ -0,0 +1,13 @@
+#include "bias.h"
+
+#define BIAS_SIZE (67)
+PI_L1 int32_t bias[BIAS_SIZE] = {
+    0x13874, 0x26f67, -0x101fd, 0x2f1b9, -0x1b8ff, 0x1ff5c, -0x109c8, -0x378c4, 0xa964, 0x14edd, 
+    -0x18c81, 0x349c8, 0x1fcf, -0x34b55, -0x11ff6, 0x343b6, -0x38662, 0x304d8, 0x103ff, 0x16f6f, 
+    -0x1cc56, 0x1fad4, 0x2ca7e, -0x4851, -0x8866, 0x840c, -0x7ef2, 0x3906f, -0x22218, -0x298f, 
+    0x3705d, 0x06, -0xf641, -0x3d441, -0x38f1f, 0x37eb9, 0x1b5ce, -0x16e86, 0x32a32, -0xb606, 
+    -0x129fc, 0x3812, 0x2ee55, 0x3b457, -0x16ad9, -0xb971, -0x4571, -0x784b, -0x15e15, -0x2bdda, 
+    -0x22b2b, -0x3a365, -0x27563, 0x3021a, -0x173c7, 0x2c311, -0x2c268, -0x2dee9, -0x165f6, -0x3e422, 
+    -0x15221, 0x3ed37, 0x164d9, -0x167b, -0x34c83, -0xcb89, 0x262d5
+};
+
diff --git a/neureka/depthwise/src/input.c b/neureka/depthwise/src/input.c
new file mode 100644
index 0000000..7d83843
--- /dev/null
+++ b/neureka/depthwise/src/input.c
@@ -0,0 +1,147 @@
+#include "input.h"
+
+#define INPUT_SIZE (1407)
+PI_L1 uint8_t input[INPUT_SIZE] = {
+    0xc2, 0x2c, 0xe2, 0x81, 0x81, 0x2d, 0xa9, 0xb3, 0x4c, 0x8a, 
+    0x32, 0x75, 0x40, 0xc1, 0xed, 0x7a, 0x9a, 0x7d, 0x5e, 0xdf, 
+    0xc8, 0x44, 0x83, 0xe8, 0x1e, 0x1b, 0xd2, 0x84, 0x39, 0xc5, 
+    0x2c, 0x27, 0x9b, 0x78, 0xb3, 0x70, 0x5a, 0x5b, 0xb5, 0x95, 
+    0x3c, 0x8f, 0xe4, 0x32, 0xfa, 0x12, 0xb7, 0xc7, 0x48, 0x91, 
+    0x02, 0x33, 0xd0, 0xbe, 0x57, 0xae, 0x61, 0xd0, 0x8a, 0x5d, 
+    0xd1, 0x72, 0x25, 0x0a, 0x81, 0x35, 0xd7, 0x36, 0xec, 0x04, 
+    0x37, 0xda, 0x99, 0xf4, 0x28, 0x46, 0xc8, 0xdd, 0x25, 0xae, 
+    0x37, 0x78, 0xf2, 0x15, 0x48, 0x57, 0xc2, 0x5b, 0xe1, 0x63, 
+    0x07, 0xab, 0xac, 0x7c, 0x2a, 0xe0, 0xbc, 0x75, 0xa8, 0xf0, 
+    0x48, 0x74, 0xe1, 0xad, 0xce, 0x17, 0x21, 0xf1, 0x72, 0x3d, 
+    0x08, 0x00, 0x44, 0x04, 0x46, 0xc7, 0x35, 0x39, 0x5f, 0xf9, 
+    0xdc, 0xad, 0x25, 0x23, 0xd5, 0x0f, 0xf0, 0x5c, 0x23, 0x08, 
+    0x25, 0x82, 0xd9, 0x76, 0xe9, 0xd1, 0x10, 0x9b, 0x13, 0x38, 
+    0x67, 0x8b, 0x96, 0x80, 0x7f, 0xe8, 0x58, 0xbf, 0x9e, 0xf3, 
+    0x34, 0x20, 0x06, 0xb6, 0xcd, 0x80, 0x3f, 0xf5, 0x2b, 0x95, 
+    0x90, 0x66, 0xa7, 0xc3, 0xf0, 0x01, 0x13, 0x11, 0x09, 0xbd, 
+    0x94, 0xfe, 0x52, 0x98, 0xda, 0x19, 0x19, 0x1a, 0x47, 0xca, 
+    0x5f, 0x7b, 0x64, 0xd8, 0x15, 0xb1, 0xcd, 0x3a, 0xe0, 0x29, 
+    0x61, 0x93, 0xa6, 0xb9, 0x99, 0x3b, 0x66, 0x46, 0x37, 0x4f, 
+    0x64, 0x87, 0xb3, 0xe3, 0x1c, 0xdf, 0x20, 0xc4, 0x48, 0xca, 
+    0x9c, 0xa8, 0x48, 0xd0, 0x52, 0x06, 0x3d, 0x87, 0x29, 0x23, 
+    0x99, 0x68, 0xed, 0x0a, 0x5c, 0xee, 0x5a, 0xec, 0xd6, 0x6e, 
+    0x65, 0x75, 0xa7, 0x0f, 0x17, 0x4b, 0xa0, 0xcf, 0xc4, 0xdc, 
+    0x28, 0xc2, 0x69, 0x4a, 0x61, 0x24, 0x55, 0xa2, 0x0c, 0xcd, 
+    0x6c, 0x79, 0xf8, 0xa1, 0xc9, 0x39, 0xc0, 0xe2, 0xc8, 0x00, 
+    0xe5, 0xbe, 0xf0, 0x81, 0x2c, 0xc4, 0xa9, 0x0c, 0xdf, 0xce, 
+    0xee, 0xe8, 0xb8, 0x62, 0x49, 0x07, 0xaa, 0x3c, 0xed, 0x30, 
+    0xd9, 0x77, 0x10, 0xd7, 0x32, 0x5e, 0x3d, 0xef, 0x27, 0x13, 
+    0x50, 0xb5, 0xb5, 0x21, 0x76, 0x14, 0x8f, 0x77, 0x7d, 0xb1, 
+    0xf8, 0xc3, 0x45, 0xfd, 0x35, 0x59, 0xa4, 0xca, 0xca, 0xbd, 
+    0x8d, 0xea, 0x83, 0x78, 0xa2, 0x6b, 0x11, 0xef, 0xae, 0x9d, 
+    0x1f, 0x9d, 0xcc, 0x98, 0x89, 0x22, 0x8b, 0x7d, 0xe8, 0xa7, 
+    0x61, 0x9e, 0x0c, 0xe3, 0x7e, 0xcc, 0x6d, 0x19, 0xbc, 0x71, 
+    0x8c, 0x01, 0x2a, 0x03, 0xa7, 0x81, 0x99, 0x12, 0x93, 0xc7, 
+    0x18, 0x3d, 0x66, 0x37, 0x98, 0x08, 0xf2, 0xb6, 0x0b, 0xa2, 
+    0x89, 0x65, 0x34, 0x07, 0x6e, 0x09, 0x84, 0xfe, 0x73, 0xf8, 
+    0x96, 0xbd, 0x09, 0x5c, 0x47, 0x1e, 0x0e, 0xa9, 0x58, 0xe7, 
+    0x5d, 0xc1, 0xdb, 0xe8, 0x67, 0x40, 0x21, 0x2b, 0x6a, 0x00, 
+    0x49, 0x57, 0xd1, 0x67, 0x18, 0xfa, 0x79, 0x87, 0xd1, 0x45, 
+    0x5a, 0xbb, 0x43, 0x3b, 0x2f, 0xd9, 0xbe, 0x8b, 0x61, 0x1f, 
+    0xc9, 0xa9, 0xe9, 0x10, 0xcb, 0x5b, 0x24, 0x82, 0x30, 0x5a, 
+    0x77, 0xe4, 0x2f, 0x40, 0x67, 0x55, 0xd1, 0x84, 0x29, 0x91, 
+    0x6f, 0x4b, 0x21, 0x94, 0xaa, 0x1f, 0x50, 0x1c, 0xc0, 0xb9, 
+    0x19, 0x0a, 0xd2, 0xe3, 0x7f, 0x91, 0x4d, 0x26, 0x93, 0x3b, 
+    0x01, 0xfd, 0x69, 0xba, 0x5d, 0xd0, 0x2b, 0x53, 0x6e, 0xd6, 
+    0x0f, 0x95, 0xde, 0x25, 0xbb, 0x3a, 0x6b, 0x36, 0x50, 0xa3, 
+    0xf9, 0x3d, 0x4d, 0xb2, 0x59, 0x49, 0xd1, 0xf3, 0x5c, 0x4e, 
+    0xd4, 0xb9, 0x6d, 0x1a, 0x72, 0x5e, 0x42, 0x92, 0x5b, 0xc8, 
+    0xdc, 0x89, 0x28, 0xe2, 0xae, 0xee, 0x61, 0xe5, 0x79, 0x3e, 
+    0x10, 0x7c, 0x62, 0x2b, 0x9b, 0xf6, 0x0c, 0x6e, 0x61, 0xc1, 
+    0xf1, 0x34, 0x77, 0x52, 0x0f, 0xa2, 0xce, 0x21, 0x27, 0x62, 
+    0xc6, 0xb3, 0xc1, 0x74, 0x43, 0xeb, 0xcb, 0x74, 0x82, 0x91, 
+    0xf9, 0x38, 0x81, 0x76, 0x91, 0x50, 0x8e, 0x96, 0x73, 0x14, 
+    0x61, 0x62, 0x95, 0x6d, 0x01, 0x16, 0x24, 0xb2, 0x66, 0x30, 
+    0x0e, 0x33, 0x47, 0x37, 0xa4, 0xc8, 0x0f, 0xc7, 0x1d, 0xed, 
+    0xc4, 0x01, 0x54, 0xa2, 0xbb, 0xfe, 0x15, 0x0c, 0x0d, 0x64, 
+    0xd5, 0x09, 0x59, 0xf4, 0x3d, 0x46, 0xfb, 0x0c, 0xc6, 0x6d, 
+    0x46, 0x2e, 0x84, 0x28, 0x4a, 0xed, 0x33, 0xad, 0xfd, 0x9b, 
+    0xfd, 0x99, 0xd8, 0x17, 0x89, 0xe0, 0x42, 0x71, 0x39, 0xba, 
+    0xa1, 0xfe, 0x23, 0xf3, 0x88, 0x95, 0xc3, 0x3a, 0xb8, 0x5b, 
+    0x0b, 0x5c, 0x0c, 0x23, 0x64, 0x4b, 0xe9, 0x0d, 0x0f, 0x33, 
+    0xd2, 0x20, 0x0d, 0x1f, 0x7a, 0x39, 0x2d, 0x8f, 0xc5, 0xd1, 
+    0x48, 0x85, 0xba, 0x06, 0x4c, 0xbe, 0xd4, 0xfb, 0x8d, 0x6a, 
+    0xd3, 0xe5, 0x3b, 0xeb, 0xa0, 0xfa, 0x1b, 0x30, 0xd3, 0x08, 
+    0x26, 0x2e, 0x0b, 0x11, 0x43, 0x46, 0xba, 0x8d, 0x4d, 0xf2, 
+    0x6e, 0xc0, 0x99, 0x63, 0xc8, 0x54, 0x0f, 0xbe, 0xcd, 0xd0, 
+    0x65, 0xc0, 0xad, 0x19, 0xcd, 0xdb, 0x9b, 0x29, 0x3a, 0xef, 
+    0xc6, 0x0a, 0x0d, 0xd1, 0xe7, 0x0a, 0xbe, 0x6d, 0xb9, 0x46, 
+    0x76, 0xdc, 0x51, 0x63, 0xbf, 0x08, 0x41, 0x47, 0xbc, 0x30, 
+    0x50, 0x00, 0xf8, 0xc6, 0x92, 0x30, 0xec, 0xf0, 0x31, 0xa2, 
+    0x4e, 0xac, 0xc8, 0x63, 0x14, 0x05, 0x62, 0x22, 0x5e, 0x29, 
+    0x6b, 0x5e, 0x73, 0xdb, 0x7e, 0x5a, 0xf8, 0x9f, 0xd6, 0xfc, 
+    0x05, 0x50, 0x14, 0xa9, 0xf3, 0x66, 0xe5, 0x92, 0xad, 0x1f, 
+    0xe4, 0x65, 0x8b, 0x3b, 0xf5, 0x46, 0x13, 0x71, 0x19, 0x2a, 
+    0xc2, 0xb4, 0x51, 0xb8, 0xd7, 0x99, 0x23, 0xe5, 0x2b, 0xb2, 
+    0xfc, 0xb9, 0x1e, 0xd9, 0x5a, 0x3c, 0x7d, 0x4d, 0x33, 0x61, 
+    0x2a, 0xdf, 0x1a, 0xb5, 0x1f, 0x6a, 0x82, 0x9b, 0xcc, 0xe0, 
+    0x90, 0x6a, 0x6d, 0x41, 0x11, 0xe7, 0x5c, 0xf1, 0x09, 0x3d, 
+    0x04, 0xe0, 0x61, 0x64, 0xba, 0x01, 0x89, 0xa4, 0xee, 0x49, 
+    0x6f, 0x16, 0x8d, 0x72, 0x51, 0x87, 0x1a, 0x56, 0x88, 0x5a, 
+    0x3c, 0x8f, 0xc6, 0xe9, 0xc8, 0xb2, 0xf4, 0x3b, 0x56, 0x19, 
+    0x8c, 0x44, 0x04, 0xe4, 0xa6, 0xb2, 0x05, 0xdd, 0x2b, 0xc0, 
+    0x26, 0x92, 0x6a, 0xe0, 0xe9, 0x43, 0xb4, 0xd6, 0x31, 0x03, 
+    0xfc, 0xd8, 0x60, 0xd8, 0xbe, 0x5d, 0x09, 0x5b, 0xea, 0x2e, 
+    0x3d, 0x19, 0x6a, 0x0b, 0xdb, 0xfc, 0x1a, 0x4c, 0xe7, 0xc4, 
+    0xe0, 0x3a, 0xed, 0xa3, 0xf3, 0x05, 0x29, 0x59, 0xf7, 0x49, 
+    0x45, 0x09, 0x82, 0xb9, 0x54, 0x64, 0x61, 0x5e, 0x73, 0x03, 
+    0xa5, 0xe1, 0x5a, 0x7f, 0x27, 0x3f, 0x9c, 0xf6, 0x11, 0x38, 
+    0xa3, 0x09, 0xc1, 0x2e, 0x2b, 0xb4, 0xed, 0x5a, 0xc9, 0x4b, 
+    0x9b, 0x4b, 0x9a, 0xfa, 0x73, 0xe5, 0xa1, 0xc2, 0x9c, 0xfb, 
+    0x7f, 0x15, 0x54, 0xc0, 0xfe, 0x13, 0xa6, 0x1d, 0x0a, 0xed, 
+    0xb2, 0x79, 0xcb, 0xcc, 0x04, 0xb8, 0x05, 0x99, 0x62, 0x98, 
+    0x43, 0xe8, 0x7d, 0xd0, 0xd1, 0xb1, 0xc1, 0x81, 0xb2, 0xe1, 
+    0x17, 0x2e, 0xd4, 0x54, 0x90, 0x65, 0xf9, 0x5b, 0xc4, 0xc8, 
+    0xee, 0x9a, 0x14, 0x0a, 0x72, 0x65, 0xa5, 0x3c, 0xc5, 0x2d, 
+    0xd5, 0xc1, 0x7d, 0x3f, 0x48, 0xd5, 0x41, 0xe6, 0x1e, 0x55, 
+    0x53, 0xa6, 0x7e, 0x3d, 0xd0, 0xe5, 0xf9, 0xb2, 0x55, 0x2b, 
+    0x21, 0xad, 0xac, 0x18, 0xdf, 0x7a, 0x3a, 0xc4, 0x5e, 0x62, 
+    0x94, 0x0b, 0x0d, 0xfd, 0x33, 0x0b, 0x59, 0x53, 0xaa, 0xbd, 
+    0xc6, 0x26, 0x6f, 0xae, 0xb8, 0x29, 0x43, 0xc4, 0x6a, 0x61, 
+    0x2d, 0x7f, 0x3e, 0x39, 0x81, 0xd0, 0xad, 0x34, 0xa9, 0x50, 
+    0x69, 0x46, 0x07, 0x69, 0xf3, 0xd3, 0x74, 0x74, 0x65, 0xb1, 
+    0x5d, 0x90, 0x8c, 0xb6, 0x39, 0xd5, 0xd1, 0x1b, 0x73, 0xd7, 
+    0xd4, 0x5d, 0xbb, 0x81, 0x37, 0x86, 0x29, 0x6b, 0x81, 0x1b, 
+    0x56, 0xb2, 0x90, 0x35, 0xc1, 0xc1, 0x8c, 0x64, 0x59, 0x3a, 
+    0x6b, 0xa2, 0x35, 0x33, 0x77, 0xf4, 0x14, 0xda, 0xbd, 0x92, 
+    0x3b, 0xa0, 0x9c, 0x68, 0xce, 0xb9, 0x6e, 0x58, 0x5a, 0x56, 
+    0x93, 0x26, 0x41, 0x0b, 0x23, 0x18, 0x35, 0x56, 0x6e, 0x1e, 
+    0x87, 0x16, 0x81, 0xca, 0x03, 0x5e, 0x43, 0x94, 0xec, 0xc5, 
+    0xdd, 0xb9, 0x81, 0xcd, 0x43, 0x54, 0x91, 0x1a, 0x1f, 0x99, 
+    0x83, 0x21, 0xb2, 0xf2, 0x13, 0x3e, 0x23, 0x1b, 0x43, 0xd8, 
+    0x51, 0x51, 0xe2, 0x06, 0x10, 0xc8, 0xb9, 0x46, 0x8e, 0x20, 
+    0x50, 0xa2, 0x6a, 0x5e, 0x2d, 0xa7, 0xf1, 0xa5, 0x47, 0x6c, 
+    0x65, 0x2f, 0xd4, 0xc9, 0x76, 0xe1, 0xef, 0xd7, 0x53, 0x21, 
+    0x52, 0x28, 0xc8, 0x3d, 0x61, 0x77, 0x11, 0x01, 0x71, 0x3b, 
+    0xdc, 0xac, 0x7f, 0xf8, 0xb0, 0xc0, 0x69, 0xda, 0xb1, 0x61, 
+    0x94, 0x61, 0x37, 0x96, 0xc7, 0xc7, 0x5b, 0xdd, 0xfb, 0x7f, 
+    0x00, 0xe0, 0x1f, 0xb4, 0x80, 0x74, 0xf9, 0x30, 0x97, 0xc9, 
+    0x12, 0xc5, 0x55, 0x82, 0x9b, 0xbf, 0xf0, 0x1d, 0xc7, 0x93, 
+    0x5a, 0x32, 0xc6, 0xf5, 0x7c, 0xbd, 0xfe, 0x67, 0x8b, 0x68, 
+    0x3d, 0x70, 0xd8, 0x24, 0x71, 0xc2, 0x01, 0x2a, 0x44, 0x93, 
+    0x84, 0x9d, 0x91, 0x69, 0xb4, 0x6d, 0xa7, 0x7a, 0xab, 0x12, 
+    0xcf, 0xab, 0xaf, 0x7c, 0x7b, 0x75, 0x53, 0xc9, 0x14, 0xb0, 
+    0x04, 0x9f, 0x40, 0x72, 0x96, 0xf8, 0x95, 0xb4, 0xda, 0x3e, 
+    0x1b, 0xf7, 0x3c, 0x91, 0x8d, 0x18, 0x52, 0x62, 0x2a, 0x4f, 
+    0xf4, 0x55, 0x7b, 0x4d, 0x15, 0x59, 0x71, 0xed, 0xf4, 0x25, 
+    0xf8, 0x26, 0x3f, 0xc6, 0x8f, 0x5b, 0x1d, 0x96, 0x3a, 0x13, 
+    0x24, 0x08, 0x5f, 0x5b, 0x69, 0x27, 0x5a, 0xbb, 0xb0, 0x86, 
+    0xdc, 0x84, 0x9d, 0x4a, 0x6b, 0x41, 0xd3, 0x87, 0x57, 0x08, 
+    0x1f, 0x22, 0x32, 0xc1, 0x7f, 0x2f, 0x84, 0xcb, 0xba, 0x86, 
+    0x27, 0x87, 0xa6, 0xa4, 0x1a, 0x99, 0xb5, 0xa5, 0xa1, 0xc3, 
+    0xa9, 0xa4, 0x03, 0x45, 0xc8, 0x2a, 0x94, 0xdf, 0x5c, 0x3b, 
+    0xc0, 0x72, 0x85, 0xf3, 0xe0, 0x1f, 0xad, 0x44, 0x66, 0x85, 
+    0xdd, 0x7b, 0xf3, 0x3a, 0x59, 0xd7, 0x81, 0x27, 0x65, 0xcf, 
+    0xba, 0xf0, 0x74, 0xfd, 0xf5, 0xf5, 0x12, 0xce, 0x31, 0xdf, 
+    0xf1, 0xdf, 0x22, 0x93, 0x6d, 0x23, 0x7f, 0xa3, 0xa7, 0x54, 
+    0xa0, 0x05, 0x10, 0xe5, 0x2d, 0x49, 0x14, 0x8f, 0x99, 0xe1, 
+    0xf5, 0x8e, 0x51, 0xe0, 0xe3, 0xcf, 0x56, 0x2f, 0x6e, 0xae, 
+    0xa5, 0x91, 0x4e, 0xe2, 0x73, 0x4c, 0xea, 0xc2, 0x25, 0x6a, 
+    0x8c, 0x6a, 0xb6, 0x26, 0x88, 0x28, 0xb5
+};
+
diff --git a/neureka/depthwise/src/output.c b/neureka/depthwise/src/output.c
new file mode 100644
index 0000000..196105c
--- /dev/null
+++ b/neureka/depthwise/src/output.c
@@ -0,0 +1,62 @@
+#include "output.h"
+
+#define OUTPUT_SIZE (335)
+PI_L1 uint8_t output[OUTPUT_SIZE];
+
+#define GOLDEN_OUTPUT_SIZE (335)
+PI_L2 uint8_t golden_output[GOLDEN_OUTPUT_SIZE] = {
+    0x26, 0xff, 0x00, 0x8c, 0x00, 0x92, 0x00, 0x54, 0x00, 0x5e, 
+    0x00, 0xcd, 0x04, 0x00, 0x00, 0xf8, 0x00, 0x7d, 0x7b, 0x00, 
+    0x00, 0x00, 0xb2, 0xc4, 0x00, 0x15, 0x00, 0xdb, 0x00, 0x9d, 
+    0xfe, 0x00, 0x00, 0x00, 0x00, 0xd9, 0x44, 0x00, 0xaf, 0x00, 
+    0x00, 0xb6, 0x4b, 0xfb, 0x00, 0x00, 0x00, 0x09, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0xa9, 0x00, 0x00, 0x00, 0x00, 
+    0xff, 0x68, 0x69, 0x32, 0x00, 0x00, 0x90, 0x58, 0xff, 0x00, 
+    0xff, 0x00, 0x8e, 0x00, 0x00, 0x00, 0x06, 0x00, 0xca, 0x00, 
+    0x00, 0x00, 0xd3, 0x00, 0xad, 0x0b, 0x00, 0x00, 0x0d, 0xb0, 
+    0x43, 0x00, 0x87, 0x00, 0xd7, 0x00, 0x3c, 0xff, 0x00, 0xff, 
+    0x00, 0x00, 0xea, 0x2f, 0x00, 0xc9, 0x10, 0x00, 0xd4, 0x08, 
+    0xff, 0x00, 0x00, 0x00, 0x00, 0x00, 0x02, 0x3b, 0x00, 0x00, 
+    0x00, 0x00, 0xff, 0x00, 0x00, 0x00, 0x00, 0x99, 0xff, 0x77, 
+    0x35, 0x00, 0x00, 0xff, 0x4a, 0xff, 0x00, 0xa9, 0x00, 0x7f, 
+    0x00, 0x00, 0x00, 0x81, 0x00, 0xdc, 0x00, 0x00, 0x00, 0xd1, 
+    0x00, 0xa0, 0x7a, 0x00, 0x00, 0x18, 0xb0, 0x7d, 0x00, 0x3d, 
+    0x00, 0xd5, 0x00, 0x4c, 0xcc, 0x00, 0x51, 0x00, 0x00, 0xd9, 
+    0x42, 0x00, 0xce, 0x00, 0x00, 0xac, 0x86, 0xff, 0x00, 0x00, 
+    0x08, 0x00, 0x00, 0x00, 0xd5, 0x00, 0x00, 0x18, 0x00, 0xff, 
+    0x00, 0x00, 0x00, 0x00, 0x72, 0xff, 0x9d, 0x56, 0x00, 0x00, 
+    0xff, 0x2d, 0xff, 0x00, 0xff, 0x00, 0x89, 0x00, 0xef, 0x00, 
+    0xb6, 0x00, 0xc9, 0x00, 0x00, 0x00, 0xff, 0x00, 0x8a, 0x2d, 
+    0x13, 0x00, 0x38, 0xaf, 0xaf, 0x00, 0x46, 0x00, 0xd1, 0x00, 
+    0x88, 0xcc, 0x00, 0x32, 0x1e, 0x00, 0xe5, 0x37, 0x00, 0xf2, 
+    0x00, 0x00, 0x8c, 0x92, 0xfb, 0x70, 0x00, 0x08, 0x00, 0x00, 
+    0x00, 0x62, 0x00, 0x00, 0x68, 0x00, 0xff, 0x00, 0x00, 0x00, 
+    0x00, 0xe7, 0xfd, 0x78, 0x40, 0x00, 0x00, 0xe0, 0x56, 0xff, 
+    0x00, 0x7c, 0x00, 0x8e, 0x00, 0x00, 0x00, 0x45, 0x00, 0xcb, 
+    0x05, 0x00, 0x00, 0xff, 0x00, 0xa3, 0x00, 0x00, 0x00, 0x29, 
+    0xb4, 0xb2, 0x00, 0x5d, 0x00, 0xcc, 0x00, 0xc2, 0xff, 0x00, 
+    0x00, 0x00, 0x00, 0xc2, 0x8f, 0x01, 0xab, 0x00, 0x00, 0x5e, 
+    0x3d, 0xff, 0x00, 0x00, 0x09, 0x0b, 0x00, 0x00, 0x71, 0x00, 
+    0x00, 0x30, 0x00, 0xff, 0x00, 0x00, 0x00, 0x00, 0xeb, 0xff, 
+    0x77, 0x70, 0x00, 0x00, 0xff
+};
+
+int check_output() {
+        printf("Checking the output vector:\n");
+
+        int n_err = 0;
+        for (int i = 0; i < OUTPUT_SIZE; i++) {
+            if (output[i] != golden_output[i]) {
+                printf("ERROR: wrong value of output @ %d: %d vs. golden: %d\n", i, output[i], golden_output[i]);
+                n_err++;
+            }
+        }
+
+        if (n_err == 0)
+            printf("> Success! No errors found.\n");
+        else
+            printf("> Failure! Found %d/%d errors.\n", n_err, OUTPUT_SIZE);
+        return n_err;
+    }
+
+    
\ No newline at end of file
diff --git a/neureka/depthwise/src/scale.c b/neureka/depthwise/src/scale.c
new file mode 100644
index 0000000..8ade6da
--- /dev/null
+++ b/neureka/depthwise/src/scale.c
@@ -0,0 +1,13 @@
+#include "scale.h"
+
+#define SCALE_SIZE (67)
+PI_L1 uint8_t scale[SCALE_SIZE] = {
+    0x05, 0x1c, 0x06, 0x0c, 0x02, 0x04, 0x0b, 0x1b, 0x0e, 0x1b, 
+    0x15, 0x02, 0x04, 0x08, 0x12, 0x07, 0x03, 0x0e, 0x19, 0x1b, 
+    0x04, 0x07, 0x01, 0x1a, 0x03, 0x0f, 0x01, 0x02, 0x14, 0x0f, 
+    0x16, 0x1d, 0x18, 0x15, 0x1e, 0x05, 0x09, 0x12, 0x0f, 0x19, 
+    0x06, 0x0a, 0x0b, 0x05, 0x1b, 0x02, 0x06, 0x1d, 0x0d, 0x19, 
+    0x1b, 0x10, 0x15, 0x19, 0x02, 0x1c, 0x04, 0x12, 0x05, 0x09, 
+    0x1e, 0x1e, 0x09, 0x17, 0x09, 0x08, 0x1a
+};
+
diff --git a/neureka/depthwise/src/weight.c b/neureka/depthwise/src/weight.c
new file mode 100644
index 0000000..b473481
--- /dev/null
+++ b/neureka/depthwise/src/weight.c
@@ -0,0 +1,93 @@
+#include "weight.h"
+
+#define WEIGHT_SIZE (864)
+PI_L1 uint8_t weight[WEIGHT_SIZE] = {
+    0xfa, 0x5e, 0x7d, 0x0e, 0x0d, 0x0d, 0x03, 0xc1, 0x56, 0x6a, 
+    0x95, 0x9e, 0x1e, 0xdf, 0x88, 0x4d, 0xbe, 0x88, 0xc4, 0x67, 
+    0x6e, 0x2e, 0x9e, 0x60, 0xcc, 0xb4, 0x25, 0x04, 0xe5, 0xfc, 
+    0x82, 0x6b, 0x1e, 0x34, 0x8a, 0x08, 0x47, 0xda, 0xe8, 0xeb, 
+    0x39, 0x54, 0x10, 0x81, 0x36, 0x76, 0x53, 0xa3, 0x87, 0x45, 
+    0x86, 0xd4, 0x9b, 0xb1, 0x58, 0xb8, 0xad, 0xa8, 0x47, 0x0a, 
+    0x0c, 0xde, 0x20, 0xc5, 0xae, 0xf5, 0xaa, 0xaf, 0x29, 0x92, 
+    0x48, 0x40, 0xd2, 0x79, 0x9d, 0x5c, 0xa5, 0x8e, 0x6d, 0x1a, 
+    0x0b, 0x12, 0xf4, 0x03, 0xa3, 0x1f, 0x45, 0x3c, 0x16, 0xd5, 
+    0xb6, 0xe1, 0x25, 0xce, 0x4a, 0x1d, 0xe4, 0xc8, 0xd6, 0xda, 
+    0x54, 0xfa, 0x6c, 0x63, 0x35, 0x4b, 0x0b, 0x3a, 0x84, 0xe5, 
+    0x8c, 0x89, 0x18, 0xb9, 0x94, 0x6d, 0x01, 0xf2, 0x5d, 0x39, 
+    0xc0, 0x2d, 0x4d, 0x97, 0x42, 0xaa, 0x49, 0x6b, 0x2b, 0xa0, 
+    0xd4, 0x4b, 0x06, 0xeb, 0x48, 0xab, 0xc5, 0x11, 0x2b, 0xf8, 
+    0xb9, 0xc8, 0xa7, 0xb0, 0xe5, 0xe8, 0x69, 0x5b, 0xe8, 0xaf, 
+    0xdf, 0xa2, 0x32, 0x3a, 0xc6, 0x20, 0x86, 0xc9, 0x2d, 0x81, 
+    0xa3, 0xd2, 0x07, 0xd2, 0x43, 0x24, 0xd8, 0x02, 0x0c, 0x37, 
+    0x48, 0x6f, 0x5f, 0x83, 0x47, 0x76, 0xd5, 0xf6, 0xe6, 0xb1, 
+    0xef, 0xcb, 0x5a, 0x39, 0xc0, 0x85, 0x0e, 0xa9, 0x30, 0x39, 
+    0x75, 0x01, 0x06, 0xc7, 0x29, 0x35, 0xa3, 0x98, 0x56, 0xd9, 
+    0x40, 0x54, 0xe8, 0x0d, 0x4f, 0xb7, 0x68, 0xfb, 0x57, 0x82, 
+    0x2f, 0x56, 0x91, 0x70, 0xe2, 0xab, 0xe7, 0xcb, 0x5a, 0x19, 
+    0xc0, 0xa5, 0x0f, 0xa9, 0xb0, 0x39, 0x75, 0x01, 0x06, 0xcf, 
+    0x2d, 0xb5, 0xa3, 0x98, 0x56, 0xd9, 0x49, 0x54, 0xe8, 0x0b, 
+    0x4c, 0x37, 0x68, 0xfb, 0x57, 0x83, 0x2f, 0x56, 0x91, 0x70, 
+    0xe2, 0xbb, 0x18, 0x34, 0xa5, 0xe6, 0x3f, 0x5a, 0xf0, 0x56, 
+    0xcf, 0xc6, 0x8a, 0xfe, 0xf9, 0x30, 0xd2, 0x4a, 0x5c, 0x67, 
+    0xa9, 0x26, 0xb6, 0xab, 0x17, 0xf4, 0xb3, 0xc8, 0x97, 0x04, 
+    0xa8, 0x7c, 0xd0, 0xa9, 0x6e, 0x8f, 0x1d, 0x44, 0xe0, 0x3e, 
+    0x07, 0xbc, 0x00, 0x64, 0x58, 0x14, 0x26, 0xc1, 0x52, 0x73, 
+    0x88, 0xf8, 0x22, 0xa9, 0x3c, 0x42, 0x0d, 0x4a, 0xec, 0xf8, 
+    0xc2, 0xd7, 0x29, 0x6d, 0x02, 0x04, 0xc4, 0x15, 0x40, 0x12, 
+    0xe1, 0xc4, 0x5c, 0x34, 0x8f, 0x46, 0x86, 0xbc, 0x93, 0xf4, 
+    0xeb, 0x2d, 0xcb, 0x67, 0xb6, 0x56, 0x32, 0x8c, 0xe0, 0x87, 
+    0x3e, 0x17, 0x06, 0xe6, 0x57, 0xf9, 0xfc, 0x7f, 0x6d, 0x3d, 
+    0x08, 0x83, 0x1c, 0x36, 0x47, 0x2b, 0x55, 0xf5, 0x89, 0x66, 
+    0xcf, 0xf5, 0x5a, 0x34, 0x95, 0x7b, 0x6f, 0xfa, 0x1a, 0x5b, 
+    0x1b, 0x05, 0x62, 0x8d, 0x0a, 0x21, 0x83, 0x83, 0x22, 0x3c, 
+    0x39, 0x47, 0x24, 0x17, 0x8f, 0xea, 0xce, 0x2d, 0xdf, 0x47, 
+    0x86, 0x49, 0x4f, 0xba, 0x0e, 0x2f, 0xa5, 0xa5, 0x1b, 0x3d, 
+    0xc6, 0x20, 0x4f, 0xf0, 0x86, 0x62, 0x47, 0xed, 0x22, 0x19, 
+    0x71, 0xad, 0x27, 0xaf, 0xe9, 0x56, 0x93, 0xf5, 0x5c, 0x47, 
+    0xff, 0xce, 0xbf, 0x2a, 0xa8, 0x95, 0x9a, 0xb5, 0x17, 0xc1, 
+    0xa6, 0x3b, 0x6e, 0xd9, 0x6a, 0xde, 0xf3, 0x7d, 0xd8, 0x33, 
+    0xca, 0x0d, 0xa9, 0xd2, 0x48, 0x38, 0x3b, 0x84, 0x48, 0x60, 
+    0xb7, 0x2c, 0x8a, 0x85, 0xb8, 0xcb, 0x01, 0xe5, 0xe5, 0x94, 
+    0xbf, 0xb4, 0x5b, 0x60, 0x98, 0xfb, 0xe6, 0xae, 0x45, 0xf1, 
+    0x6b, 0xbe, 0x65, 0x49, 0x54, 0x33, 0xd7, 0x32, 0x2f, 0xc0, 
+    0x5c, 0x2a, 0x31, 0xa7, 0x0c, 0xe6, 0xaf, 0x2a, 0x9f, 0x64, 
+    0x7c, 0x57, 0xab, 0xbb, 0xa7, 0x15, 0xbc, 0x92, 0x84, 0x2f, 
+    0x81, 0x87, 0xe3, 0x8f, 0x45, 0xf1, 0x6b, 0xbe, 0x65, 0x69, 
+    0x5c, 0x33, 0xc6, 0x32, 0x2f, 0xc0, 0x5c, 0x28, 0x39, 0xa5, 
+    0x0c, 0xe4, 0xaf, 0x2e, 0x9b, 0x64, 0x78, 0x47, 0xab, 0xbf, 
+    0xe7, 0x15, 0xbc, 0x94, 0xc0, 0x2f, 0x90, 0x87, 0xe3, 0x8f, 
+    0xba, 0x0e, 0x94, 0x41, 0x9a, 0x96, 0xa3, 0xcc, 0x39, 0xcd, 
+    0xd0, 0x3f, 0xa3, 0xd7, 0xc6, 0x5a, 0xf3, 0x1b, 0x50, 0xd1, 
+    0x64, 0x9b, 0x87, 0xb8, 0x54, 0x40, 0x18, 0xea, 0x43, 0x6b, 
+    0x3f, 0xd0, 0x6f, 0x78, 0x1c, 0x70, 0x03, 0x00, 0x00, 0x00, 
+    0x03, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x03, 0x00, 
+    0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x04, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x02, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x02, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x06, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x04, 0x00, 
+    0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 
+    0x04, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x07, 0x00, 
+    0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 
+    0x06, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x03, 0x00, 
+    0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x05, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 
+    0x04, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x07, 0x00, 
+    0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x06, 0x00, 
+    0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 
+    0x07, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 
+    0x07, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x02, 0x00, 
+    0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00
+};
+
diff --git a/neureka/idma_pointwise/Makefile b/neureka/idma_pointwise/Makefile
new file mode 100644
index 0000000..41c2bc7
--- /dev/null
+++ b/neureka/idma_pointwise/Makefile
@@ -0,0 +1,5 @@
+APP_SRCS += main.c
+OVERRIDE_MAIN = 1
+include ../app/Makefile
+
+STIM_DIR := ../pointwise/
\ No newline at end of file
diff --git a/neureka/idma_pointwise/inc/bias.h b/neureka/idma_pointwise/inc/bias.h
new file mode 100644
index 0000000..f28d0f5
--- /dev/null
+++ b/neureka/idma_pointwise/inc/bias.h
@@ -0,0 +1,9 @@
+#ifndef __BIAS_H__
+#define __BIAS_H__
+
+#include <pmsis.h>
+
+#define BIAS_SIZE (39)
+extern int32_t bias[BIAS_SIZE];
+
+#endif  // __BIAS_H__
diff --git a/neureka/idma_pointwise/inc/input.h b/neureka/idma_pointwise/inc/input.h
new file mode 100644
index 0000000..985f9a9
--- /dev/null
+++ b/neureka/idma_pointwise/inc/input.h
@@ -0,0 +1,9 @@
+#ifndef __INPUT_H__
+#define __INPUT_H__
+
+#include <pmsis.h>
+
+#define INPUT_SIZE (8995)
+extern uint8_t input[INPUT_SIZE];
+
+#endif  // __INPUT_H__
diff --git a/neureka/idma_pointwise/inc/layer_conf.h b/neureka/idma_pointwise/inc/layer_conf.h
new file mode 100644
index 0000000..e5e4812
--- /dev/null
+++ b/neureka/idma_pointwise/inc/layer_conf.h
@@ -0,0 +1,42 @@
+#ifndef __LAYER_CONF_H__
+#define __LAYER_CONF_H__
+
+#define TEST_NAME "test"
+#define INPUT_HEIGHT (7)
+#define INPUT_WIDTH (5)
+#define INPUT_CHANNEL (257)
+#define INPUT_SIGNED (0)
+#define INPUT_BITS (8)
+
+#define OUTPUT_HEIGHT (7)
+#define OUTPUT_WIDTH (5)
+#define OUTPUT_CHANNEL (39)
+#define OUTPUT_BITS (8)
+
+#define WEIGHT_HEIGHT (1)
+#define WEIGHT_WIDTH (1)
+#define WEIGHT_CHANNEL_IN (257)
+#define WEIGHT_CHANNEL_OUT (39)
+#define WEIGHT_BITS (8)
+#define WEIGHT_OFFSET (-128)
+
+#define SCALE_BITS (8)
+
+#define BIAS_BITS (32)
+
+#define PADDING_TOP (0)
+#define PADDING_BOTTOM (0)
+#define PADDING_LEFT (0)
+#define PADDING_RIGHT (0)
+#define PADDING_VALUE (0)
+
+#define STRIDE_HEIGHT (1)
+#define STRIDE_WIDTH (1)
+
+#define GROUPS (1)
+#define OUTSHIFT (12)
+#define HAS_NORM_QUANT (1)
+#define HAS_BIAS (1)
+#define HAS_RELU (1)
+
+#endif  // __LAYER_CONF_H__
diff --git a/neureka/idma_pointwise/inc/output.h b/neureka/idma_pointwise/inc/output.h
new file mode 100644
index 0000000..68c6c11
--- /dev/null
+++ b/neureka/idma_pointwise/inc/output.h
@@ -0,0 +1,14 @@
+#ifndef __OUTPUT_H__
+#define __OUTPUT_H__
+
+#include <pmsis.h>
+
+#define OUTPUT_SIZE (1365)
+extern uint8_t output[OUTPUT_SIZE];
+
+#define GOLDEN_OUTPUT_SIZE (1365)
+extern uint8_t golden_output[GOLDEN_OUTPUT_SIZE];
+
+int check_output();
+
+#endif  // __OUTPUT_H__
diff --git a/neureka/idma_pointwise/inc/scale.h b/neureka/idma_pointwise/inc/scale.h
new file mode 100644
index 0000000..67dea26
--- /dev/null
+++ b/neureka/idma_pointwise/inc/scale.h
@@ -0,0 +1,9 @@
+#ifndef __SCALE_H__
+#define __SCALE_H__
+
+#include <pmsis.h>
+
+#define SCALE_SIZE (39)
+extern uint8_t scale[SCALE_SIZE];
+
+#endif  // __SCALE_H__
diff --git a/neureka/idma_pointwise/inc/weight.h b/neureka/idma_pointwise/inc/weight.h
new file mode 100644
index 0000000..ba39fae
--- /dev/null
+++ b/neureka/idma_pointwise/inc/weight.h
@@ -0,0 +1,9 @@
+#ifndef __WEIGHT_H__
+#define __WEIGHT_H__
+
+#include <pmsis.h>
+
+#define WEIGHT_SIZE (11232)
+extern uint8_t weight[WEIGHT_SIZE];
+
+#endif  // __WEIGHT_H__
diff --git a/neureka/idma_pointwise/main.c b/neureka/idma_pointwise/main.c
new file mode 100644
index 0000000..dee3678
--- /dev/null
+++ b/neureka/idma_pointwise/main.c
@@ -0,0 +1,360 @@
+/*
+ * Luka Macan <luka.macan@unibo.it>
+ *
+ * Copyright 2023 ETH Zurich and University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ *
+ * SPDX-License-Identifier: Apache-2.0
+ */
+
+#include "nnx_layer.h"
+#include <pmsis.h>
+
+#include "neureka.h"
+#include "neureka_gvsoc.h"
+#include "neureka_testbench_bsp.h"
+#include "neureka_task.h"
+#include "pulp_nnx_neureka.h"
+
+#define NULL 0
+
+typedef neureka_norm_mode_e nnx_norm_mode_e;
+typedef neureka_quant_t nnx_quant_t;
+typedef neureka_quant_function_e nnx_quant_function_e;
+typedef neureka_norm_t nnx_norm_t;
+typedef neureka_task_t nnx_task_t;
+typedef neureka_dev_t nnx_dev_t;
+typedef neureka_testbench_conf_t nnx_bsp_conf_t;
+typedef neureka_task_flag_e nnx_task_flag_e;
+
+#define nnxTaskFlagTrue neurekaTaskFlagTrue
+#define nnxTaskFlagFalse neurekaTaskFlagFalse
+
+#define nnx_task_init neureka_task_init
+#define nnx_task_set_op_to_conv neureka_task_set_op_to_conv
+#define nnx_task_set_bits neureka_task_set_bits
+#define nnx_task_set_norm_quant neureka_task_set_norm_quant
+#define nnx_task_set_weight_offset neureka_task_set_weight_offset
+#define nnx_task_set_weight_source neureka_task_set_weight_source
+#define nnx_task_set_activation_prefetch neureka_task_set_activation_prefetch
+#define nnx_task_set_dims neureka_task_set_dims
+#define nnx_task_set_ptrs_conv neureka_task_set_ptrs_conv
+#define nnx_task_set_ptrs_norm_quant neureka_task_set_ptrs_norm_quant
+
+#define nnx_bsp_get_dev neureka_testbench_get_dev
+
+#define nnx_init neureka_nnx_init
+#define nnx_dispatch_wait neureka_nnx_dispatch_wait
+#define nnx_dispatch neureka_nnx_dispatch
+#define nnx_resolve_wait neureka_nnx_resolve_wait
+#define nnx_term neureka_nnx_term
+
+// Generated headers
+#include "bias.h"
+#include "input.h"
+#include "layer_conf.h"
+#include "output.h"
+#include "scale.h"
+#include "weight.h"
+
+static void task_prepare(nnx_task_t *task) {
+  nnx_task_init(task);
+  nnx_task_set_op_to_conv(task, WEIGHT_HEIGHT, GROUPS > 1);
+  nnx_task_set_bits(task, INPUT_BITS, OUTPUT_BITS, WEIGHT_BITS);
+
+  nnx_task_set_weight_offset(task, weightOffsetModeLayerWise, WEIGHT_OFFSET);
+
+#ifdef NEUREKA_WEIGHT_SOURCE_WMEM
+  nnx_task_set_weight_source(task, neurekaWeightSourceWmem);
+  nnx_task_set_activation_prefetch(task, activationPrefetchOn);
+#else
+  neureka_task_set_weight_source(task, neurekaWeightSourceTcdm);
+  nnx_task_set_activation_prefetch(task, activationPrefetchOff);
+#endif
+#if INPUT_SIGNED == 1
+  neureka_task_set_input_signed(task);
+#else
+  neureka_task_set_input_unsigned(task);
+#endif
+
+  const uint32_t w_in_stride = INPUT_CHANNEL * INPUT_BITS / 8;
+  const uint32_t h_in_stride = INPUT_WIDTH * w_in_stride;
+  const uint32_t w_out_stride = OUTPUT_CHANNEL * OUTPUT_BITS / 8;
+  const uint32_t h_out_stride = OUTPUT_WIDTH * w_out_stride;
+
+#if STRIDE_HEIGHT == 2 && STRIDE_WIDTH == 2
+  nnx_task_set_dims_stride2x2(
+      task, INPUT_HEIGHT, INPUT_WIDTH, INPUT_CHANNEL, h_in_stride, w_in_stride,
+      OUTPUT_HEIGHT, OUTPUT_WIDTH, OUTPUT_CHANNEL, h_out_stride, w_out_stride,
+      WEIGHT_HEIGHT, WEIGHT_WIDTH, PADDING_TOP, PADDING_BOTTOM, PADDING_LEFT,
+      PADDING_RIGHT);
+#else
+  nnx_task_set_dims(task, INPUT_WIDTH, INPUT_CHANNEL, h_in_stride, w_in_stride,
+                    OUTPUT_HEIGHT, OUTPUT_WIDTH, OUTPUT_CHANNEL, h_out_stride,
+                    w_out_stride, PADDING_TOP, PADDING_BOTTOM, PADDING_LEFT,
+                    PADDING_RIGHT);
+#endif
+
+  nnx_task_set_ptrs_conv(task, (uint32_t)input, INPUT_WIDTH, w_in_stride,
+                         PADDING_TOP, PADDING_LEFT, (uint32_t)output,
+                         (uint32_t)weight);
+#if HAS_NORM_QUANT == 1
+#if SCALE_BITS == 8
+  const nnx_norm_mode_e normMode = normMode8Bit;
+#elif SCALE_BITS == 32
+  const nnx_norm_mode_e normMode = normMode32Bit;
+#endif
+
+  const nnx_task_flag_e flag_bias =
+      HAS_BIAS ? nnxTaskFlagTrue : nnxTaskFlagFalse;
+  const uint32_t bias_ptr = (uint32_t)(HAS_BIAS ? bias : NULL);
+
+  nnx_quant_function_e quant_function =
+      HAS_RELU ? quantFunctionRelu : quantFunctionIdentity;
+
+  nnx_task_set_norm_quant(task,
+                          (nnx_quant_t){.shift_amount = OUTSHIFT,
+                                        .function = quant_function,
+                                        .flag_rounding = nnxTaskFlagFalse},
+                          (nnx_norm_t){.mode = normMode,
+                                       .flag_bias = flag_bias,
+                                       .flag_shift = nnxTaskFlagFalse});
+
+  nnx_task_set_ptrs_norm_quant(task, (uint32_t)scale, NULL, bias_ptr);
+#endif // HAS_NORM_QUANT
+}
+
+static void task_execute(nnx_task_t *task) {
+  nnx_dev_t *dev = nnx_bsp_get_dev();
+
+  nnx_bsp_conf_t conf = {.max_stall = 8};
+  nnx_init(dev, &conf);
+
+  nnx_dispatch_wait(dev);
+
+  // printf("CFG:\n");
+  // for (int i=0; i<sizeof(neureka_task_data_t)/4; i++) {
+  //   printf("%08x\n", ((uint32_t *) &task->data)[i]);
+  // }
+#if STRIDE_HEIGHT == 2 && STRIDE_WIDTH == 2
+  nnx_dispatch_stride2x2(dev, task, INPUT_WIDTH, INPUT_CHANNEL, OUTPUT_HEIGHT,
+                         OUTPUT_WIDTH, OUTPUT_CHANNEL, WEIGHT_HEIGHT,
+                         WEIGHT_WIDTH);
+#else
+  nnx_dispatch(dev, task);
+#endif
+
+  nnx_resolve_wait(dev, task);
+
+  nnx_term(dev);
+
+}
+#include <stdint.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <pulp.h>
+
+#include "layer_util.h"
+#include "nnx_layer.h"
+#include "output.h"
+#include "input.h"
+#include "weight.h"
+
+
+#include <math.h>
+#include "pulp.h"
+
+#define VERBOSE_INIT_MEM
+// #define VERBOSE_EXEC_IDMA
+#define VERBOSE_TEST_MEM
+#define VERBOSE_TEST_NEUREKA
+
+#define MAX_BUFFER_SIZE 0x2000
+
+L2_DATA static unsigned char ext[MAX_BUFFER_SIZE];
+L1_DATA static unsigned char loc[MAX_BUFFER_SIZE];
+
+#define EXT_DATA_ADDR  ((unsigned int) ext)
+#define TCDM_DATA_ADDR ((unsigned int) loc)
+typedef enum {RX, TX} test_type_t;
+
+int test_idma(unsigned int len, test_type_t type, unsigned int ext_addr, unsigned int tcdm_addr);
+void initialize_mem(unsigned int len, test_type_t type, unsigned int ext_addr, unsigned int tcdm_addr);
+int test_mem(unsigned int len, test_type_t type, unsigned int ext_addr, unsigned int tcdm_addr);
+int execute_idma(unsigned int len, test_type_t type, unsigned int ext_addr, unsigned int tcdm_addr);
+
+L1_DATA int global_error = 0;
+
+#define TX_SIZE (MAX_BUFFER_SIZE - 0x100) 
+
+int main() {
+  if (rt_cluster_id() != 0)
+    return bench_cluster_forward(0);
+
+	int error_count = 0;
+	int id = 0;
+
+  nnx_task_t task;
+
+
+	if(get_core_id()==0)
+  {
+		initialize_mem(TX_SIZE, TX, ext, loc);
+  }
+
+  if(get_core_id()==1)
+    task_prepare(&task);
+
+	synch_barrier();
+	
+	if(get_core_id()==0)
+  {
+    // manually delay to force overlapped execution
+    for(volatile cnt=0; cnt<500; cnt++);
+		id = execute_idma(TX_SIZE, TX, ext, loc);
+  }
+
+	if(get_core_id()==1)
+    task_execute(&task);
+	
+	synch_barrier();
+
+	if(get_core_id()==0)
+		error_count = test_mem(TX_SIZE, TX, ext, loc);
+	if(get_core_id()==1)
+		error_count = check_output();
+
+	synch_barrier();
+
+	if(get_core_id()==0)
+		global_error += error_count;
+	
+	synch_barrier();
+	
+	if(get_core_id()==1)
+		global_error += error_count;
+	
+	synch_barrier();
+
+	return global_error;
+}
+
+int test_idma_loop(){
+  int   error_count = 0;
+
+	for ( int i = 5; i < 8045; i=5*i) {
+		error_count += test_idma(i, TX, ext, loc);
+	}
+	for ( int i = 5; i < 8045; i=5*i ) {
+		error_count += test_idma(i, RX, ext, loc);
+	}
+	return error_count;
+}
+
+int test_idma(unsigned int len, test_type_t type, unsigned int ext_addr, unsigned int tcdm_addr){
+	
+	int error_count = 0; 
+
+	initialize_mem(len, TX, ext, loc);
+	execute_idma(len, TX, ext, loc);
+	error_count = test_mem(len, TX, ext, loc);
+	return error_count;
+}
+
+void initialize_mem(unsigned int len, test_type_t type, unsigned int ext_addr, unsigned int tcdm_addr){ 
+	volatile unsigned int i;
+	if (type == RX){
+
+#ifdef VERBOSE_INIT_MEM
+    printf ("INITIALIZING MEMORY FOR RX %d OPERATION: \n", len);
+#endif
+
+    for (i=0; i<len; i++){
+      *(unsigned char*)(ext_addr + i) = i & 0xFF;
+    }
+
+    for (i=0; i<len+16; i++){
+      *(unsigned char*)(tcdm_addr + i) = 0;
+    }
+  } else {
+
+#ifdef VERBOSE_INIT_MEM
+    printf ("INITIALIZING MEMORY FOR TX %d OPERATION: \n", len);
+#endif
+
+    for (i=0; i<len; i++){
+      *(unsigned char*)(tcdm_addr + i) = i & 0xFF;
+    }
+
+    for (i=0; i<len+16; i++){
+      *(unsigned char*)(ext_addr+ i) = 0;
+    }
+  }
+}
+
+int test_mem(unsigned int len, test_type_t type, unsigned int ext_addr, unsigned int tcdm_addr){
+	
+	volatile unsigned int i;
+	volatile unsigned int test,read,error=0;
+
+  if (type == RX){
+    for (i=0; i<len; i++){
+      test = i & 0xFF;
+      read = *(unsigned char*)(tcdm_addr + i);
+      if ( test != read ){
+#ifdef VERBOSE_TEST_MEM
+  			printf("[IDMA TEST] Error!!! Read: %x, Test:%x, Index: %d \n ",read,test,i);
+#endif
+  			error++;
+      }
+    }
+	}
+	else{
+    for (i=0; i<len; i++){
+      test = i & 0xFF;
+      read = *(unsigned char*)(ext_addr + i);
+      if ( test != read ){
+#ifdef VERBOSE_TEST_MEM
+  			printf("[IDMA TEST] Error!!! Read: %x, Test:%x, Index: %d \n ",read,test,i);
+#endif
+  			error++;
+      }
+    }
+  }
+
+  if (error == 0)
+    printf("[IDMA TEST] OOOOOOK!!!!!!\n");
+  else
+    printf("[IDMA TEST] NOT OK!!!!!\n");
+
+  return error;
+}
+
+
+int execute_idma(unsigned int len, test_type_t type, unsigned int ext_addr, unsigned int tcdm_addr){
+  volatile unsigned int id;
+  if (type == RX){
+#ifdef VERBOSE_EXEC_IDMA
+    printf ("STARTING TEST FOR RX %d OPERATION: \n", len);
+#endif
+    id = pulp_idma_memcpy(ext_addr, tcdm_addr, len, IDMA_PROT_AXI, IDMA_PROT_OBI);
+  } else {
+#ifdef VERBOSE_EXEC_IDMA
+    printf ("STARTING TEST FOR TX %d OPERATION: \n", len);
+#endif
+    id = pulp_idma_memcpy(tcdm_addr, ext_addr, len, IDMA_PROT_OBI, IDMA_PROT_AXI);
+  }
+  plp_dma_barrier();
+	return id;
+}
diff --git a/neureka/idma_pointwise/src/bias.c b/neureka/idma_pointwise/src/bias.c
new file mode 100644
index 0000000..55b2013
--- /dev/null
+++ b/neureka/idma_pointwise/src/bias.c
@@ -0,0 +1,10 @@
+#include "bias.h"
+
+#define BIAS_SIZE (39)
+PI_L1 int32_t bias[BIAS_SIZE] = {
+    0x36dfa, -0xec5f, 0x3ee6f, -0x23173, -0xb8b1, -0x236a3, -0x3fe84, 0x1285e, 0x9a43, -0xfb97, 
+    0x396a7, -0x16b65, 0x5527, 0xb8f, 0xe316, 0x3c2e1, 0x2bf5e, -0x1adf0, -0x486, 0x1ed7d, 
+    0x3a88e, 0x8e46, -0x3e8ef, 0xcbee, 0x65e5, 0x2426b, 0x298f4, 0x205dc, 0x2598, -0x2151c, 
+    0x390ef, -0x2318f, 0x1bc1, 0xe222, 0x1a09e, -0x1f3fd, 0x3d956, -0x12e4d, 0x33ee0
+};
+
diff --git a/neureka/idma_pointwise/src/input.c b/neureka/idma_pointwise/src/input.c
new file mode 100644
index 0000000..b6afde2
--- /dev/null
+++ b/neureka/idma_pointwise/src/input.c
@@ -0,0 +1,906 @@
+#include "input.h"
+
+#define INPUT_SIZE (8995)
+PI_L1 uint8_t input[INPUT_SIZE] = {
+    0xc2, 0xe1, 0x4d, 0x2d, 0x90, 0xf3, 0x32, 0xc8, 0x6d, 0x7a, 
+    0x72, 0x92, 0xc8, 0x2d, 0x28, 0x1b, 0x48, 0xe5, 0x2c, 0x55, 
+    0x62, 0x70, 0xd0, 0x6e, 0x3c, 0x2b, 0x77, 0x12, 0xdf, 0x21, 
+    0x02, 0x62, 0xc1, 0xae, 0x33, 0x74, 0xd1, 0xbd, 0x81, 0x35, 
+    0xb8, 0xf3, 0x0b, 0xb3, 0x5f, 0x10, 0x9c, 0xe5, 0x2d, 0x8d, 
+    0xa2, 0x92, 0x94, 0xbb, 0x8e, 0xda, 0x96, 0xf8, 0xe2, 0x45, 
+    0x7f, 0xb6, 0x04, 0x30, 0x4b, 0x00, 0x06, 0x95, 0x77, 0x44, 
+    0x1f, 0x66, 0xaf, 0x5d, 0x95, 0xb9, 0x4e, 0xcd, 0xde, 0x94, 
+    0x7f, 0x2e, 0x93, 0x3b, 0xec, 0x87, 0x51, 0xd8, 0x96, 0x02, 
+    0x91, 0x96, 0x18, 0x23, 0x4c, 0xd1, 0x04, 0x66, 0xde, 0xf6, 
+    0xc6, 0x32, 0xec, 0x4c, 0x59, 0x75, 0xd6, 0x30, 0x17, 0xc5, 
+    0xe6, 0xc4, 0xe7, 0x8e, 0x65, 0x02, 0x1d, 0x3a, 0xd9, 0x3c, 
+    0x8f, 0xe2, 0xf4, 0xcc, 0x88, 0xc6, 0xea, 0xc3, 0xaf, 0x4b, 
+    0xb6, 0x48, 0x01, 0xaf, 0x0f, 0xfa, 0xdc, 0x74, 0x1c, 0x29, 
+    0x31, 0xf8, 0x9c, 0x54, 0xa4, 0xf0, 0xb5, 0xbe, 0xe4, 0x43, 
+    0xaf, 0x5a, 0x62, 0x55, 0x69, 0x89, 0x53, 0x0c, 0xc7, 0x1a, 
+    0x70, 0xb0, 0x1f, 0xe7, 0x91, 0x60, 0x08, 0x51, 0xd8, 0xc6, 
+    0x8e, 0x37, 0x48, 0x64, 0x57, 0x90, 0xef, 0x85, 0xb6, 0x1d, 
+    0xc9, 0x95, 0x47, 0x9f, 0xd8, 0xe3, 0xd0, 0x77, 0x1a, 0x59, 
+    0x1c, 0x4f, 0x1e, 0x65, 0x08, 0xba, 0x3a, 0xc4, 0xd0, 0x29, 
+    0x98, 0x72, 0x2f, 0x54, 0x3f, 0x2d, 0xd0, 0x00, 0xbd, 0x4d, 
+    0xad, 0x8a, 0x99, 0xe8, 0x75, 0x42, 0x32, 0xc7, 0x01, 0xf3, 
+    0xf9, 0xbe, 0x66, 0x2a, 0x77, 0x93, 0x90, 0x69, 0x55, 0x34, 
+    0x6e, 0x48, 0x1a, 0x9b, 0xc6, 0xbb, 0x35, 0x5d, 0x64, 0xd8, 
+    0xe0, 0x86, 0x1a, 0x2f, 0xa1, 0x5f, 0xd9, 0x07, 0xb8, 0x42, 
+    0x49, 0xac, 0x6c, 0x0f, 0xdd, 0x27, 0x69, 0x36, 0x43, 0x14, 
+    0x99, 0x7f, 0x01, 0xdd, 0xad, 0x30, 0xf2, 0x46, 0xa4, 0x5b, 
+    0x74, 0xed, 0xac, 0x90, 0xbb, 0x75, 0xd1, 0x64, 0xe1, 0x5d, 
+    0x3d, 0xf1, 0x29, 0x6d, 0x44, 0xb2, 0x4a, 0x39, 0x8c, 0x9b, 
+    0x25, 0xa2, 0x89, 0x5c, 0x14, 0xba, 0xd9, 0xa0, 0x0b, 0x22, 
+    0x84, 0xfb, 0xd5, 0xe9, 0xe2, 0x42, 0x0c, 0xe8, 0x5d, 0x0a, 
+    0x99, 0xf7, 0x61, 0x35, 0xe5, 0x7b, 0x8c, 0x2c, 0x38, 0x1a, 
+    0xca, 0x94, 0x66, 0xae, 0x45, 0xa9, 0x4c, 0x52, 0x18, 0x1f, 
+    0x08, 0xf8, 0xfd, 0x9c, 0x03, 0xf2, 0xfb, 0xfe, 0xa9, 0x55, 
+    0xae, 0x69, 0x2c, 0x93, 0xd7, 0x8d, 0x2f, 0x63, 0xf7, 0x17, 
+    0x37, 0xe3, 0xeb, 0x0f, 0x2e, 0x2b, 0x33, 0x53, 0xcc, 0xfb, 
+    0x4d, 0x4f, 0x7f, 0x85, 0xcf, 0x16, 0x34, 0x88, 0x2d, 0x44, 
+    0x58, 0xda, 0x76, 0xca, 0xc3, 0xa8, 0x45, 0xc7, 0x9a, 0x25, 
+    0xd8, 0x59, 0x83, 0x5b, 0x21, 0xcd, 0x00, 0x2b, 0xcc, 0xa7, 
+    0x24, 0x45, 0x98, 0x02, 0x64, 0x7d, 0x18, 0x19, 0xe0, 0x03, 
+    0x0d, 0xc9, 0x1a, 0x5b, 0x29, 0x61, 0x46, 0xb4, 0x54, 0xe4, 
+    0x20, 0xb0, 0xcd, 0xa3, 0xf6, 0xb9, 0x43, 0x91, 0xcc, 0xc8, 
+    0x2b, 0xea, 0x49, 0xce, 0x83, 0x12, 0xb6, 0xc2, 0x38, 0x36, 
+    0x0b, 0x71, 0x22, 0x72, 0x78, 0x77, 0x01, 0xf0, 0xd3, 0x74, 
+    0x53, 0x70, 0x19, 0xf8, 0xac, 0x70, 0xf1, 0xd0, 0xdc, 0xea, 
+    0x71, 0x3b, 0xa8, 0xfc, 0x12, 0x3b, 0x06, 0x5c, 0x7d, 0x21, 
+    0x6e, 0x1b, 0xbf, 0xea, 0x79, 0x84, 0x19, 0x42, 0xf3, 0xbd, 
+    0xc1, 0x0e, 0xf8, 0x31, 0x3c, 0xfc, 0x34, 0x7f, 0xf6, 0x34, 
+    0xad, 0x0d, 0x01, 0x19, 0xa0, 0x61, 0x1e, 0xe8, 0xc6, 0x56, 
+    0x42, 0xf5, 0xb2, 0xa4, 0x90, 0x6e, 0x59, 0x3f, 0x89, 0x92, 
+    0x8d, 0x5d, 0x17, 0x69, 0x41, 0x19, 0x17, 0xca, 0xbd, 0x5b, 
+    0xb1, 0x4f, 0xd7, 0xcb, 0xe9, 0x68, 0xc3, 0x38, 0x5a, 0x5c, 
+    0x7f, 0x0b, 0xe9, 0xf3, 0x6e, 0x20, 0xcd, 0xca, 0x2d, 0x95, 
+    0xec, 0x85, 0xf0, 0xcd, 0xd4, 0xbd, 0x1f, 0xe5, 0xda, 0xf2, 
+    0x1b, 0xca, 0x43, 0x2e, 0x15, 0x06, 0xba, 0x29, 0x8e, 0xc0, 
+    0x99, 0x5e, 0x0f, 0x4f, 0x47, 0x41, 0xac, 0x69, 0xb4, 0xcd, 
+    0x00, 0x10, 0x21, 0xfc, 0x6b, 0xe2, 0x3e, 0x9b, 0xc6, 0xaa, 
+    0x5d, 0x8c, 0x28, 0x54, 0x36, 0xdc, 0x6a, 0xbc, 0xa6, 0x8d, 
+    0xf2, 0x07, 0x3b, 0x1c, 0x33, 0x05, 0x82, 0xd4, 0xaa, 0x5c, 
+    0xd4, 0xc6, 0x8d, 0x2d, 0x1e, 0x2c, 0xe4, 0xed, 0x12, 0x75, 
+    0x14, 0xbb, 0xf6, 0x04, 0x21, 0xc6, 0xdd, 0xc6, 0x52, 0xd5, 
+    0x71, 0x8b, 0x83, 0x0b, 0x89, 0x41, 0xe3, 0x36, 0x12, 0xef, 
+    0xdd, 0xce, 0x03, 0xdd, 0x7e, 0xef, 0x93, 0x22, 0x63, 0x25, 
+    0xb7, 0x1f, 0x4c, 0x17, 0x4e, 0x2e, 0x14, 0x1f, 0xc3, 0x6e, 
+    0xd7, 0x3c, 0x69, 0x15, 0xb7, 0xe8, 0x62, 0x2c, 0x7a, 0xbf, 
+    0x24, 0xe4, 0xa5, 0x50, 0xef, 0xa0, 0x53, 0x7c, 0x49, 0x20, 
+    0x37, 0x46, 0x27, 0x5f, 0x81, 0x1f, 0xb9, 0x08, 0x3b, 0x8f, 
+    0xcc, 0x13, 0xf6, 0x4c, 0x2f, 0x17, 0x68, 0x04, 0x7b, 0xd6, 
+    0x93, 0x1a, 0x50, 0x38, 0x2e, 0xda, 0x0a, 0x21, 0x4e, 0xf1, 
+    0x03, 0xea, 0xeb, 0x59, 0x63, 0xef, 0xf9, 0xf2, 0xcf, 0xa0, 
+    0x43, 0xc3, 0x6b, 0x9e, 0x07, 0x70, 0x05, 0x7d, 0xa5, 0x88, 
+    0x81, 0x45, 0xf1, 0x77, 0x5b, 0xd3, 0xc7, 0x3c, 0x0e, 0x3c, 
+    0xc5, 0x12, 0xc6, 0x37, 0xa9, 0xc7, 0x97, 0x84, 0xcf, 0x77, 
+    0xd1, 0x25, 0x98, 0x69, 0xf8, 0x1d, 0x8e, 0x97, 0xea, 0x8b, 
+    0xdb, 0xba, 0x02, 0x66, 0xee, 0x60, 0xea, 0x5b, 0x00, 0x7e, 
+    0xef, 0xed, 0x99, 0x1d, 0x97, 0xc5, 0x78, 0x53, 0x95, 0x7e, 
+    0xc6, 0xb2, 0x99, 0x51, 0xfa, 0xac, 0x55, 0xbe, 0x59, 0x1a, 
+    0xba, 0x87, 0x65, 0x19, 0x20, 0xe1, 0x3a, 0xa8, 0x52, 0xd1, 
+    0x3d, 0x77, 0xb9, 0x68, 0xdc, 0x63, 0x5a, 0xc0, 0xbc, 0x75, 
+    0x94, 0xc6, 0xa0, 0xc7, 0x31, 0xc2, 0x00, 0x63, 0x55, 0x74, 
+    0x5e, 0x79, 0x12, 0xdb, 0xc0, 0xbf, 0xd6, 0xbe, 0x5a, 0xa9, 
+    0xa9, 0xbd, 0xf2, 0xf8, 0x76, 0x1e, 0xfc, 0xc8, 0x77, 0x3b, 
+    0xb1, 0xc0, 0x4c, 0xa0, 0xa2, 0x92, 0xb5, 0x37, 0x63, 0x38, 
+    0x64, 0xf5, 0x06, 0xee, 0x53, 0x08, 0xc2, 0x70, 0xdc, 0x34, 
+    0x60, 0xa8, 0x23, 0x78, 0xb3, 0xed, 0xe6, 0x1d, 0x6a, 0x7e, 
+    0xbf, 0x88, 0x4c, 0x01, 0x9f, 0x52, 0x1a, 0xac, 0xa6, 0x57, 
+    0x6f, 0xb7, 0xae, 0x02, 0x4d, 0x96, 0x8b, 0xfe, 0x92, 0x09, 
+    0x01, 0xb5, 0xa2, 0x9e, 0x45, 0x09, 0x61, 0xbc, 0xa7, 0x70, 
+    0x4e, 0x5a, 0x0d, 0x88, 0xc2, 0xe0, 0xdd, 0x81, 0xf7, 0x6b, 
+    0xa8, 0x2d, 0xc7, 0xd9, 0x32, 0x8b, 0x29, 0x35, 0x65, 0x10, 
+    0xb2, 0x4a, 0xe9, 0xb0, 0x58, 0xd7, 0x78, 0xae, 0x66, 0x6c, 
+    0xd0, 0x6f, 0x4c, 0xfe, 0x5c, 0x09, 0xee, 0xb1, 0x99, 0xdd, 
+    0x67, 0xa2, 0x01, 0x87, 0xe8, 0xe1, 0xd5, 0x8e, 0xf3, 0x4c, 
+    0x68, 0x67, 0x4b, 0xb8, 0x73, 0x9a, 0xa9, 0x5c, 0x8a, 0x37, 
+    0x6b, 0xb9, 0xf0, 0x42, 0xb5, 0xbe, 0x69, 0x83, 0x60, 0x8c, 
+    0xdb, 0x33, 0xf0, 0x74, 0x57, 0x3b, 0x81, 0xee, 0xb2, 0x2b, 
+    0xf2, 0xf3, 0xad, 0xe0, 0x68, 0x94, 0xb4, 0xaa, 0x33, 0x43, 
+    0xe0, 0x3e, 0xbe, 0x7d, 0x58, 0xb9, 0xa2, 0x44, 0x80, 0x6f, 
+    0x63, 0xbe, 0xf9, 0xc3, 0x5a, 0x84, 0xfe, 0xbd, 0xbe, 0xcb, 
+    0xad, 0x3a, 0x77, 0x6c, 0x36, 0xc9, 0x94, 0x61, 0x17, 0xd8, 
+    0xa2, 0x5d, 0xe0, 0x60, 0xfa, 0xa5, 0xc3, 0x68, 0x73, 0xb8, 
+    0x81, 0xbb, 0x00, 0x6a, 0xc4, 0xdd, 0xb9, 0x10, 0x25, 0x75, 
+    0x93, 0xb6, 0x51, 0xca, 0xbf, 0x75, 0xca, 0x64, 0xdf, 0x67, 
+    0xe4, 0x62, 0xd8, 0x46, 0xed, 0x2a, 0xc2, 0xd7, 0x91, 0x99, 
+    0x27, 0x7a, 0xfc, 0x21, 0xaf, 0x3c, 0x7d, 0xc9, 0x2a, 0xfd, 
+    0x40, 0x6a, 0xca, 0xb4, 0x90, 0x78, 0x3c, 0xe7, 0xae, 0x62, 
+    0x04, 0x98, 0x7b, 0x01, 0xe8, 0xed, 0x6f, 0xe3, 0x3f, 0xed, 
+    0xc9, 0xee, 0x98, 0x66, 0x3a, 0xde, 0x89, 0xd6, 0xb8, 0x25, 
+    0xf0, 0x09, 0xd1, 0xa4, 0x77, 0xec, 0xa3, 0xd6, 0xdb, 0x15, 
+    0xe5, 0x3b, 0xdd, 0x36, 0x96, 0x69, 0x26, 0x99, 0x86, 0xcf, 
+    0x32, 0x6e, 0x4b, 0x57, 0x00, 0x52, 0x12, 0xa4, 0xf8, 0x37, 
+    0xf5, 0x27, 0x70, 0xc7, 0x4e, 0xb8, 0x09, 0xe5, 0x18, 0xaf, 
+    0x23, 0x59, 0xa6, 0xbe, 0xa5, 0x25, 0x37, 0xb1, 0xf0, 0x60, 
+    0x95, 0xcf, 0x6a, 0x8c, 0x96, 0xd8, 0x1c, 0x02, 0xd5, 0x5d, 
+    0x32, 0x44, 0x2b, 0x75, 0x52, 0xcf, 0x08, 0x74, 0x2e, 0x1c, 
+    0xfa, 0x91, 0x7e, 0xd5, 0x36, 0xe7, 0xfe, 0x85, 0x68, 0x09, 
+    0x45, 0xdb, 0x7e, 0x98, 0x2a, 0x7b, 0x6f, 0xe8, 0x04, 0x6c, 
+    0xf1, 0xb0, 0xda, 0x5d, 0x9c, 0xb1, 0x81, 0xe7, 0x1d, 0x16, 
+    0x33, 0xb6, 0xb8, 0x10, 0xf0, 0xfd, 0x25, 0x1f, 0xb8, 0xf1, 
+    0xea, 0x60, 0x58, 0xe5, 0x3a, 0xf0, 0x6a, 0xbd, 0x05, 0xc1, 
+    0xe1, 0x5c, 0x8e, 0x16, 0xdb, 0x61, 0xb8, 0x07, 0x97, 0x1d, 
+    0x84, 0x76, 0x44, 0x43, 0x55, 0xc5, 0x2f, 0x07, 0x39, 0x76, 
+    0xfb, 0x7d, 0x65, 0xb2, 0xdc, 0xdc, 0x9e, 0xed, 0x17, 0x38, 
+    0xdf, 0x1b, 0xb9, 0x27, 0xe1, 0xbb, 0x7d, 0x17, 0x23, 0xd1, 
+    0x9b, 0x49, 0x20, 0xb2, 0x3b, 0x86, 0x9e, 0xae, 0x86, 0xfd, 
+    0xc1, 0xbf, 0x4d, 0x31, 0x35, 0xc7, 0xd1, 0x49, 0xe7, 0x8d, 
+    0x83, 0xd2, 0xcc, 0x75, 0x3a, 0x25, 0x69, 0x82, 0x68, 0x72, 
+    0x46, 0x3f, 0x6d, 0xa4, 0xfa, 0x73, 0xa5, 0x30, 0x5c, 0x2f, 
+    0xd8, 0x61, 0xa0, 0x1e, 0x75, 0xcc, 0x8f, 0x56, 0x8c, 0x13, 
+    0xc6, 0x81, 0x69, 0x3b, 0x18, 0x86, 0x04, 0x08, 0x6b, 0xdd, 
+    0x89, 0x08, 0x6a, 0x09, 0x7f, 0xd6, 0x96, 0x86, 0x60, 0x1e, 
+    0x1a, 0x5b, 0x5d, 0xc3, 0x6a, 0x40, 0xc8, 0x4c, 0x49, 0x3b, 
+    0xed, 0xfa, 0xe0, 0x59, 0x5a, 0x85, 0x62, 0xa2, 0xce, 0x1b, 
+    0x02, 0xa0, 0x8f, 0x3b, 0xdb, 0x54, 0xe4, 0x83, 0x99, 0xb5, 
+    0xdd, 0xb0, 0xb0, 0x0b, 0xb0, 0xfa, 0x63, 0x5c, 0xab, 0x9b, 
+    0xe7, 0x21, 0xa4, 0x31, 0xa9, 0x59, 0x62, 0x07, 0x6f, 0x9c, 
+    0x01, 0x4a, 0x33, 0xe3, 0xa4, 0xba, 0xae, 0x8a, 0xf7, 0x4c, 
+    0x3f, 0x06, 0xa1, 0x61, 0x2b, 0xa5, 0xc3, 0xb1, 0xe0, 0xe6, 
+    0x42, 0x40, 0x21, 0x3b, 0x00, 0x3f, 0xbd, 0x2e, 0xfd, 0xfc, 
+    0xa8, 0x93, 0x2c, 0x9f, 0x33, 0x1b, 0x1b, 0xe0, 0x59, 0xa5, 
+    0xf6, 0xf4, 0x6b, 0x79, 0x07, 0x95, 0x55, 0x0e, 0x4c, 0x2c, 
+    0x9c, 0x65, 0xca, 0xb5, 0x0b, 0x5e, 0x8a, 0xaf, 0xf9, 0x48, 
+    0xad, 0xee, 0x75, 0x5d, 0xf4, 0x18, 0xf2, 0x24, 0x80, 0xd1, 
+    0x84, 0xd5, 0x72, 0x1e, 0xe9, 0x1a, 0xbc, 0x08, 0x44, 0xbd, 
+    0x4c, 0xf3, 0xa9, 0xaa, 0x83, 0xe2, 0xb8, 0x75, 0x89, 0x00, 
+    0x22, 0x84, 0x76, 0x0b, 0xf9, 0x11, 0xd9, 0x23, 0xd6, 0xc3, 
+    0x02, 0x5d, 0x80, 0xc9, 0x0d, 0xa0, 0x01, 0xf1, 0xfe, 0x5c, 
+    0x70, 0xad, 0x47, 0x58, 0xb4, 0x0b, 0x6e, 0xe6, 0xd6, 0xbc, 
+    0x21, 0xfd, 0xd6, 0x2d, 0x8c, 0x98, 0x56, 0x57, 0x6a, 0xaa, 
+    0x3f, 0xf2, 0x72, 0x68, 0x97, 0x8a, 0x38, 0x8b, 0xd4, 0xf6, 
+    0xe7, 0x97, 0x77, 0xf5, 0xfe, 0x3f, 0x74, 0x1b, 0xfa, 0xbc, 
+    0x43, 0xc9, 0xc5, 0x46, 0x3c, 0xa1, 0x1f, 0x20, 0x9a, 0xe0, 
+    0xcb, 0xa4, 0x92, 0x3e, 0xd4, 0xf7, 0xdb, 0x72, 0x17, 0x6a, 
+    0x7c, 0xb5, 0xed, 0x96, 0xf8, 0x08, 0xf9, 0x95, 0x1f, 0xa4, 
+    0xb1, 0xeb, 0x43, 0x7b, 0x54, 0x8b, 0x81, 0x03, 0xe9, 0xf0, 
+    0x27, 0x82, 0x12, 0x38, 0x2f, 0xdf, 0x2b, 0x84, 0x7f, 0x4b, 
+    0x21, 0x05, 0x73, 0x1c, 0x14, 0xfb, 0xd2, 0x8e, 0xfe, 0x26, 
+    0x56, 0xed, 0x69, 0x91, 0x04, 0x53, 0xea, 0x98, 0xde, 0x6a, 
+    0xd1, 0x36, 0xb5, 0x8e, 0x63, 0x85, 0xf1, 0x3c, 0xcb, 0xe5, 
+    0x10, 0x4e, 0xd3, 0x36, 0xb9, 0xab, 0x03, 0x39, 0x09, 0x8e, 
+    0xf5, 0xd3, 0xf7, 0x35, 0x70, 0x3b, 0x0b, 0x98, 0xb7, 0xb6, 
+    0x58, 0x31, 0x20, 0xed, 0x10, 0x7a, 0x63, 0x9a, 0x3c, 0x3d, 
+    0x4e, 0x34, 0x40, 0x1d, 0xdb, 0x43, 0x2f, 0x6f, 0x74, 0x51, 
+    0xc3, 0x77, 0x76, 0xe8, 0x5d, 0x0d, 0x0f, 0x22, 0xd7, 0x3f, 
+    0xfb, 0xf5, 0xce, 0x6a, 0x60, 0x81, 0x42, 0x64, 0x81, 0x45, 
+    0x24, 0xeb, 0x81, 0xf8, 0xe5, 0x13, 0xd7, 0xec, 0x07, 0x1d, 
+    0x5c, 0x42, 0xe2, 0x9b, 0x98, 0x9e, 0x57, 0x0f, 0xed, 0xd8, 
+    0x26, 0x91, 0x8c, 0x50, 0x69, 0x5b, 0x3a, 0x0a, 0x8e, 0x23, 
+    0x20, 0xaa, 0x3e, 0x3d, 0xd5, 0x83, 0x87, 0x8c, 0xe2, 0xe7, 
+    0xbc, 0x24, 0x90, 0x4e, 0x1a, 0x69, 0x18, 0x0a, 0x7e, 0x68, 
+    0x1c, 0x6f, 0xd2, 0x26, 0x76, 0x37, 0xe5, 0x1d, 0xe1, 0x96, 
+    0x42, 0x74, 0x72, 0x16, 0x8b, 0xa3, 0x90, 0xa1, 0x91, 0x48, 
+    0x4b, 0x56, 0x09, 0x2e, 0xea, 0xa3, 0x08, 0xd9, 0x2f, 0x04, 
+    0x51, 0x60, 0x65, 0x3b, 0x4b, 0x95, 0xba, 0xe8, 0x87, 0x16, 
+    0xe5, 0x3f, 0xb8, 0x42, 0x08, 0xdb, 0x82, 0x7b, 0xe7, 0xf6, 
+    0xb7, 0x6d, 0xc2, 0x82, 0xb6, 0x4f, 0xac, 0xcc, 0x70, 0x2f, 
+    0x56, 0xbc, 0xd4, 0x79, 0x96, 0xca, 0x02, 0x38, 0x40, 0xd3, 
+    0x0d, 0x8b, 0xd9, 0x53, 0x4d, 0xd7, 0xe6, 0x96, 0x70, 0x71, 
+    0x6a, 0xf5, 0xd3, 0xdb, 0xa7, 0x42, 0xe7, 0x6c, 0x04, 0xf1, 
+    0x8c, 0xe4, 0xf4, 0x57, 0x32, 0xf6, 0x94, 0x80, 0x5e, 0xa3, 
+    0xe2, 0x2e, 0x59, 0xb3, 0xf9, 0x4e, 0x40, 0x9a, 0x72, 0x7d, 
+    0xa5, 0xc8, 0x83, 0xc1, 0xae, 0x84, 0x41, 0x3e, 0x9b, 0xa6, 
+    0x9b, 0x5b, 0xf9, 0xc1, 0xe4, 0xad, 0x0f, 0xc7, 0x3a, 0x62, 
+    0xd0, 0x0b, 0x43, 0xd0, 0x59, 0x91, 0x25, 0x26, 0x91, 0x0a, 
+    0xe1, 0x44, 0xdf, 0xdf, 0xbb, 0xca, 0xec, 0x04, 0x8a, 0xb7, 
+    0x21, 0x18, 0x93, 0x5a, 0xac, 0x03, 0xe4, 0x6a, 0xf5, 0x2b, 
+    0x2b, 0xb5, 0xc9, 0x07, 0x62, 0x44, 0xf6, 0x2e, 0xc5, 0x03, 
+    0xb7, 0xb1, 0xd3, 0x18, 0xe9, 0xd3, 0x93, 0xd1, 0x52, 0x8c, 
+    0x36, 0xd1, 0x89, 0xfd, 0xca, 0x71, 0x42, 0x36, 0xd0, 0x3b, 
+    0x7b, 0x9b, 0xcb, 0x22, 0xf8, 0x52, 0xcc, 0x56, 0xb8, 0xfe, 
+    0x0d, 0xc2, 0x49, 0x0a, 0x4d, 0xc6, 0x66, 0x52, 0xa4, 0x31, 
+    0x3c, 0x79, 0x9c, 0xd4, 0x9b, 0x84, 0x13, 0xec, 0x8f, 0xa1, 
+    0x70, 0xd2, 0xf7, 0x73, 0x71, 0x47, 0x58, 0xef, 0x29, 0x40, 
+    0x48, 0x76, 0x11, 0x49, 0x9f, 0xf9, 0x6f, 0x93, 0xcd, 0xa4, 
+    0x3e, 0xea, 0xf3, 0xfd, 0x14, 0xa0, 0xce, 0x0e, 0x29, 0x53, 
+    0xfb, 0xbc, 0xfe, 0xcf, 0x1f, 0x62, 0x26, 0x65, 0xd1, 0x57, 
+    0x6e, 0x6f, 0xac, 0x29, 0x83, 0x44, 0x2e, 0x56, 0x48, 0xba, 
+    0xca, 0x1f, 0x60, 0x54, 0xd6, 0x50, 0xf9, 0xc8, 0xdf, 0xd0, 
+    0xa4, 0xe1, 0x49, 0x0e, 0x30, 0x1b, 0x7b, 0xb5, 0x78, 0x17, 
+    0xa4, 0x94, 0x0a, 0x5f, 0xd3, 0x12, 0x4a, 0x0c, 0xb1, 0x35, 
+    0xdd, 0x1c, 0x54, 0x83, 0xc3, 0x85, 0x06, 0x4b, 0x18, 0x46, 
+    0x20, 0xf4, 0x24, 0x77, 0x22, 0x34, 0x58, 0x93, 0xf2, 0xd8, 
+    0x03, 0xac, 0x64, 0x08, 0x88, 0xb8, 0x9a, 0x21, 0x94, 0xda, 
+    0xe1, 0x64, 0xc8, 0x2b, 0x30, 0xa8, 0xc4, 0xb9, 0x54, 0x6e, 
+    0xef, 0x5b, 0x5d, 0x62, 0x6e, 0xda, 0x76, 0xf6, 0x59, 0x54, 
+    0x1f, 0x62, 0x1c, 0xcf, 0xef, 0x84, 0x8e, 0x04, 0x6a, 0x62, 
+    0x28, 0x39, 0x24, 0xae, 0xa9, 0x33, 0x48, 0x69, 0x0f, 0x63, 
+    0x65, 0x01, 0x2a, 0xb6, 0x15, 0xf0, 0x73, 0x09, 0xce, 0x81, 
+    0xfb, 0x3d, 0x81, 0x2e, 0x46, 0x35, 0x33, 0xf9, 0x59, 0x99, 
+    0xd5, 0x33, 0x42, 0x08, 0xbd, 0xfe, 0x59, 0xe1, 0x98, 0x7d, 
+    0xdd, 0x46, 0x44, 0x6e, 0x6d, 0x08, 0xf2, 0xfa, 0xdd, 0xce, 
+    0xb1, 0x94, 0x09, 0x6c, 0x21, 0xb2, 0x37, 0xf4, 0x46, 0x36, 
+    0xdc, 0xb5, 0xcc, 0xd5, 0x68, 0x5d, 0x8a, 0x0e, 0xc3, 0xa4, 
+    0xab, 0xd5, 0xf0, 0x70, 0xe7, 0x80, 0xdc, 0xcb, 0xe2, 0xba, 
+    0x79, 0xa9, 0xed, 0x91, 0x4b, 0x20, 0x5e, 0xab, 0x6f, 0xb7, 
+    0xa2, 0x09, 0x18, 0x57, 0xad, 0xee, 0xc9, 0xa1, 0x4e, 0xa0, 
+    0x78, 0xb5, 0x37, 0x87, 0xad, 0xee, 0xf0, 0x5e, 0x28, 0x40, 
+    0x4b, 0xab, 0x86, 0xb4, 0x36, 0x83, 0x48, 0xe5, 0xc0, 0x9f, 
+    0x1d, 0x3f, 0xdc, 0x19, 0xf7, 0xd1, 0x7b, 0x9c, 0xe5, 0xe6, 
+    0x06, 0xbb, 0x95, 0xae, 0x8d, 0xea, 0x22, 0x3e, 0x49, 0x23, 
+    0xf0, 0x9c, 0xa4, 0x10, 0x41, 0x12, 0x59, 0x60, 0xad, 0xbb, 
+    0xa6, 0xf1, 0xec, 0xb2, 0xb5, 0xfe, 0x2e, 0x35, 0x26, 0x7e, 
+    0x6e, 0x06, 0xc7, 0xa2, 0x61, 0xe3, 0x22, 0x3f, 0x89, 0x8e, 
+    0x6a, 0xeb, 0xe3, 0x2c, 0xcf, 0xcd, 0xb0, 0xd6, 0x65, 0xd0, 
+    0x9b, 0xd5, 0x6a, 0x79, 0xd6, 0xf7, 0x3e, 0x87, 0xde, 0x15, 
+    0xb6, 0x19, 0x79, 0xc3, 0xbf, 0x7b, 0xc0, 0xf6, 0x90, 0x5c, 
+    0x50, 0x2a, 0x85, 0x03, 0x82, 0x56, 0x31, 0x62, 0x78, 0x84, 
+    0x14, 0x8e, 0x3e, 0x0c, 0x3f, 0x6d, 0x49, 0x9b, 0x33, 0x82, 
+    0x19, 0x39, 0x2f, 0x40, 0x8e, 0x20, 0x71, 0x64, 0x6a, 0x07, 
+    0x55, 0xed, 0xd4, 0xc9, 0x77, 0xa8, 0xf2, 0xb2, 0x40, 0xb3, 
+    0x91, 0xa1, 0xa5, 0xcf, 0x4d, 0x43, 0x3a, 0x7c, 0xa7, 0xa9, 
+    0xe0, 0x10, 0x60, 0xf3, 0x10, 0xb9, 0xb8, 0x8b, 0x93, 0x23, 
+    0x58, 0x18, 0x0f, 0x20, 0x87, 0x1f, 0x3f, 0x5e, 0xc5, 0x66, 
+    0xdd, 0x06, 0x13, 0x54, 0x8d, 0xfe, 0x83, 0xeb, 0x19, 0x3e, 
+    0xd3, 0x7b, 0x51, 0x11, 0xcd, 0xc8, 0x4d, 0x93, 0x50, 0x63, 
+    0x66, 0xa7, 0xcd, 0x61, 0x1a, 0x03, 0xdb, 0xcf, 0x7d, 0x4a, 
+    0x88, 0xd3, 0xa4, 0x10, 0x1f, 0x28, 0x31, 0xe0, 0x8a, 0xdf, 
+    0xd2, 0xac, 0x48, 0xd3, 0x8e, 0x79, 0x69, 0xd3, 0xf5, 0xc1, 
+    0x86, 0x79, 0x71, 0x73, 0x1f, 0x47, 0x7e, 0x8b, 0x3c, 0x12, 
+    0x24, 0xfe, 0x98, 0xdf, 0xc4, 0x10, 0xae, 0x84, 0xb8, 0xfe, 
+    0xa4, 0x49, 0x67, 0xf9, 0xbd, 0xa7, 0x03, 0x1b, 0xd5, 0x40, 
+    0x3a, 0xd5, 0xb7, 0x5c, 0xd4, 0x94, 0x1e, 0x96, 0x46, 0x01, 
+    0xf8, 0xb0, 0x20, 0xcf, 0x02, 0x59, 0xcf, 0x11, 0x37, 0xd2, 
+    0xe7, 0x56, 0xc9, 0x89, 0xc7, 0x77, 0x0b, 0x24, 0x04, 0xca, 
+    0xa6, 0x44, 0x43, 0xda, 0x28, 0x86, 0x9f, 0x5b, 0x06, 0x1b, 
+    0xc4, 0xcd, 0xa0, 0x9c, 0x89, 0x18, 0x03, 0x47, 0x50, 0x6d, 
+    0xe8, 0xcd, 0x18, 0x53, 0x10, 0xef, 0xa2, 0xa2, 0xf6, 0xcf, 
+    0xdf, 0x92, 0x1a, 0x50, 0x3f, 0xdb, 0x8e, 0x58, 0x40, 0x24, 
+    0xeb, 0xd0, 0xbd, 0x5f, 0x8b, 0x01, 0xb1, 0x65, 0x59, 0x18, 
+    0xd9, 0x62, 0xda, 0x70, 0x64, 0x10, 0x15, 0xce, 0x0b, 0x26, 
+    0xc0, 0x98, 0xd2, 0xa9, 0x19, 0x65, 0xdb, 0x22, 0x77, 0x8f, 
+    0x80, 0x79, 0x51, 0xc4, 0xa4, 0x67, 0xcf, 0x89, 0x8c, 0x9a, 
+    0xd3, 0x80, 0x2f, 0x18, 0x78, 0x26, 0x16, 0xeb, 0xe0, 0x17, 
+    0x69, 0xa5, 0x17, 0x9c, 0xa2, 0x8f, 0x08, 0x74, 0xad, 0x67, 
+    0x8f, 0xaa, 0x91, 0x0a, 0x9e, 0x56, 0x83, 0x14, 0x36, 0x34, 
+    0xcc, 0xef, 0xfe, 0x17, 0x9f, 0xc3, 0xdd, 0x12, 0x5c, 0x23, 
+    0xa7, 0x70, 0x58, 0x48, 0x06, 0x9c, 0x36, 0xe5, 0x2a, 0xfa, 
+    0x65, 0xe3, 0xd4, 0xdb, 0x48, 0xd7, 0xc6, 0xd0, 0xc8, 0x0a, 
+    0x29, 0x01, 0x76, 0x0a, 0x7f, 0x08, 0xd6, 0xda, 0x50, 0x0f, 
+    0x37, 0x30, 0xc4, 0xdd, 0x4e, 0x4a, 0x1f, 0x05, 0x0c, 0x30, 
+    0x6b, 0xa1, 0x55, 0x5a, 0xc8, 0x1d, 0x05, 0x81, 0xc6, 0x66, 
+    0xc0, 0x05, 0xd6, 0x65, 0xde, 0x29, 0x8e, 0xf3, 0xc7, 0x50, 
+    0x9a, 0x0e, 0x70, 0xbf, 0x32, 0xec, 0xa9, 0x3e, 0xf2, 0x09, 
+    0x36, 0xdb, 0xd9, 0x6f, 0xfc, 0x95, 0xb4, 0x8b, 0x4d, 0x4a, 
+    0x1a, 0xce, 0x31, 0x3f, 0xf3, 0x8b, 0x2f, 0x86, 0xb0, 0x34, 
+    0xa8, 0x0d, 0x5f, 0x19, 0x16, 0x18, 0x57, 0xf0, 0xa7, 0x4d, 
+    0x00, 0xc7, 0x81, 0x4a, 0x38, 0x5a, 0x3e, 0x03, 0x53, 0x87, 
+    0xd1, 0x55, 0xc8, 0x9b, 0xe7, 0xe5, 0x2f, 0xd8, 0x22, 0x3d, 
+    0xe2, 0xe2, 0xa1, 0x09, 0xf7, 0xc3, 0x77, 0x18, 0x06, 0xd6, 
+    0x45, 0x65, 0x3c, 0x53, 0xe0, 0x49, 0x80, 0x83, 0x52, 0x95, 
+    0x91, 0x51, 0x03, 0xb4, 0x06, 0x43, 0x84, 0x75, 0x90, 0xd9, 
+    0xfe, 0x3e, 0xc4, 0x0f, 0xee, 0x2d, 0x5e, 0xdc, 0x40, 0xaf, 
+    0x83, 0xb8, 0x77, 0x70, 0x7d, 0x68, 0xef, 0xe8, 0x58, 0x4a, 
+    0x2d, 0x15, 0xba, 0x41, 0x22, 0xd4, 0xc0, 0xf3, 0x00, 0x3f, 
+    0xb3, 0x13, 0x06, 0xe5, 0xe8, 0xfe, 0xfc, 0x40, 0xda, 0x39, 
+    0x38, 0x74, 0x87, 0xf2, 0x55, 0x01, 0x7c, 0x84, 0x0f, 0x16, 
+    0x52, 0x80, 0x0a, 0x91, 0xe8, 0x0d, 0xde, 0xc6, 0xeb, 0x5d, 
+    0x06, 0x5e, 0x89, 0x98, 0x60, 0xc9, 0x31, 0x38, 0x20, 0x29, 
+    0xe1, 0xcc, 0x1c, 0x29, 0xc8, 0xb6, 0xf3, 0x6b, 0x9b, 0x84, 
+    0xf6, 0x62, 0x4e, 0x5c, 0xa2, 0x2a, 0x7a, 0x71, 0xf6, 0x6f, 
+    0xb0, 0xa3, 0x8e, 0x37, 0x79, 0xa7, 0x35, 0x85, 0xbf, 0xc4, 
+    0x59, 0x51, 0xb8, 0x50, 0x1a, 0xd4, 0x30, 0x88, 0x14, 0x0d, 
+    0x21, 0x7f, 0x5c, 0xe0, 0xf3, 0x5d, 0x02, 0x92, 0xee, 0x68, 
+    0x8b, 0x07, 0x71, 0x71, 0xd9, 0x93, 0x51, 0x5e, 0xb4, 0xe5, 
+    0x50, 0x12, 0x1e, 0x14, 0x7b, 0x4d, 0xf8, 0xb0, 0x1a, 0x59, 
+    0x96, 0x9b, 0x8d, 0x3e, 0x6d, 0x6b, 0x8d, 0xf1, 0x1f, 0x4f, 
+    0x61, 0x22, 0x15, 0xa4, 0x61, 0x25, 0x8d, 0x4d, 0x86, 0xa8, 
+    0xfd, 0xf4, 0x7a, 0xdb, 0x50, 0x9c, 0xce, 0xe8, 0x9e, 0x19, 
+    0xe7, 0x4e, 0x28, 0x7b, 0x06, 0xab, 0x56, 0x22, 0xea, 0xa0, 
+    0xb2, 0x33, 0xb1, 0x4f, 0x6e, 0x29, 0xa4, 0x40, 0x3d, 0xa3, 
+    0x0e, 0x5b, 0x0e, 0xdc, 0x4b, 0x63, 0x66, 0xee, 0x0f, 0x1b, 
+    0xee, 0x74, 0x04, 0xd9, 0x23, 0x76, 0xf7, 0x01, 0x2e, 0x91, 
+    0x48, 0x2e, 0xc1, 0x02, 0x7b, 0x97, 0xde, 0x27, 0x81, 0x53, 
+    0x29, 0xd7, 0x6c, 0xe4, 0x9d, 0x42, 0xcf, 0xea, 0x3d, 0xf8, 
+    0xbb, 0xf2, 0x58, 0x7b, 0x06, 0x09, 0x57, 0xb8, 0xb2, 0x37, 
+    0x62, 0xc1, 0x92, 0xd9, 0xea, 0x0c, 0x54, 0xe6, 0xcd, 0x05, 
+    0x78, 0xe3, 0x90, 0x0d, 0x1e, 0xeb, 0x8e, 0x80, 0x9b, 0x14, 
+    0x0a, 0xae, 0xa7, 0x5c, 0xab, 0x91, 0x55, 0x6f, 0x8c, 0x24, 
+    0xe8, 0x13, 0x78, 0x39, 0x7f, 0x85, 0x22, 0xcc, 0xde, 0x6a, 
+    0x3a, 0x6a, 0x7b, 0x9e, 0x07, 0x2d, 0xb9, 0x7c, 0x01, 0x17, 
+    0x59, 0x2a, 0x7f, 0xd6, 0x0c, 0xe0, 0xdc, 0xc4, 0x55, 0xdd, 
+    0x1c, 0x30, 0x75, 0xdb, 0x81, 0x96, 0xe9, 0x90, 0xa6, 0x48, 
+    0x4a, 0x88, 0x76, 0xdb, 0x3a, 0xc7, 0xd0, 0x1b, 0x9f, 0x9d, 
+    0x39, 0xba, 0x70, 0x8a, 0xbe, 0xdc, 0x8f, 0xd0, 0x42, 0x6c, 
+    0x67, 0xf8, 0x7f, 0x49, 0x4e, 0x5d, 0x22, 0xee, 0x02, 0xec, 
+    0x29, 0xb0, 0xb1, 0xf8, 0x06, 0x16, 0x67, 0xd2, 0x3f, 0x8e, 
+    0x88, 0x25, 0x3b, 0x96, 0x89, 0x49, 0xf4, 0x58, 0xb0, 0xd4, 
+    0x05, 0xfa, 0x48, 0x8f, 0x40, 0xcd, 0xa5, 0xae, 0x3d, 0xec, 
+    0x3a, 0x26, 0xa4, 0x71, 0x51, 0x19, 0x1d, 0x5a, 0x2a, 0x08, 
+    0xc8, 0x12, 0x5a, 0x19, 0x66, 0x84, 0xa6, 0xb6, 0xd3, 0xc0, 
+    0x34, 0x22, 0xe9, 0xfe, 0x84, 0x03, 0x09, 0x87, 0xbe, 0xa9, 
+    0xb5, 0x2e, 0xdb, 0xa4, 0xdb, 0x2b, 0x94, 0xc4, 0xd1, 0x72, 
+    0xf3, 0x87, 0xad, 0x49, 0x10, 0x8d, 0x65, 0x27, 0x82, 0x12, 
+    0x96, 0x3d, 0x7c, 0x6f, 0xd2, 0xad, 0x9d, 0xfb, 0x0e, 0x51, 
+    0x3e, 0x5a, 0x4d, 0x0f, 0xb9, 0x3e, 0x67, 0xce, 0x2b, 0x90, 
+    0xbe, 0x11, 0x9f, 0xd2, 0x42, 0x15, 0x57, 0x31, 0x9a, 0x40, 
+    0x98, 0x60, 0x24, 0xce, 0x95, 0x9e, 0x06, 0xeb, 0x09, 0xd1, 
+    0x16, 0x9b, 0x3b, 0xbf, 0x8c, 0x55, 0x90, 0x90, 0x63, 0x4b, 
+    0x4e, 0x6b, 0xc2, 0x50, 0xe2, 0x5f, 0x48, 0x8d, 0xc2, 0x3f, 
+    0xc4, 0xf0, 0x19, 0xbe, 0x6c, 0xb2, 0x1b, 0xe7, 0xcd, 0xa5, 
+    0x1f, 0xa6, 0x10, 0xeb, 0x33, 0x41, 0xf9, 0x30, 0x4c, 0x8d, 
+    0x23, 0xc3, 0xa0, 0xee, 0x77, 0x90, 0x93, 0x75, 0x8d, 0x6b, 
+    0x8f, 0xea, 0x33, 0xf3, 0xc1, 0xea, 0x57, 0x72, 0x39, 0x2d, 
+    0x5c, 0xa6, 0xb8, 0xe6, 0x9e, 0x5d, 0x89, 0x2f, 0xbe, 0xf5, 
+    0x1f, 0x67, 0x8c, 0xf0, 0x06, 0x74, 0x3a, 0x83, 0x24, 0x63, 
+    0xf7, 0x41, 0x36, 0xa3, 0x6e, 0x4c, 0x3a, 0xe6, 0x45, 0x92, 
+    0xe7, 0x54, 0x28, 0xe2, 0x25, 0xaf, 0x3a, 0xd8, 0xe4, 0x29, 
+    0x32, 0x43, 0x67, 0x12, 0x34, 0x34, 0x3b, 0x34, 0xea, 0xed, 
+    0x7d, 0xd6, 0x1d, 0x86, 0x22, 0xd8, 0x89, 0xb5, 0xdc, 0x23, 
+    0x57, 0x75, 0x50, 0xc5, 0x08, 0x24, 0xc7, 0x1a, 0x94, 0xbe, 
+    0x30, 0x35, 0x7d, 0x53, 0x55, 0x5f, 0xe0, 0x16, 0x43, 0x1c, 
+    0x82, 0xa6, 0x83, 0xfb, 0x57, 0x5b, 0x38, 0x8f, 0x43, 0xc7, 
+    0x17, 0x0b, 0x33, 0x3c, 0x96, 0x7e, 0xce, 0x10, 0x8b, 0x60, 
+    0xab, 0xa2, 0x26, 0x6d, 0x78, 0x26, 0x5b, 0xb8, 0xd7, 0x58, 
+    0x8c, 0x09, 0x2f, 0x3a, 0x61, 0x1f, 0x65, 0xe1, 0xcb, 0xfd, 
+    0x9c, 0x5a, 0x31, 0x09, 0x67, 0x93, 0xed, 0x91, 0xa7, 0x4b, 
+    0xaa, 0xe5, 0xa1, 0xb9, 0x99, 0x15, 0x7f, 0xe0, 0xa6, 0x3b, 
+    0x6e, 0x79, 0x5d, 0xe2, 0x05, 0xd6, 0x25, 0xe8, 0xbb, 0x26, 
+    0xc1, 0x0f, 0x88, 0xce, 0xd7, 0xfd, 0xcf, 0x34, 0x89, 0xd1, 
+    0xb2, 0xdb, 0xf6, 0xba, 0xc0, 0x81, 0x67, 0x8f, 0x9a, 0xdf, 
+    0x05, 0xe2, 0x56, 0xdd, 0xdd, 0xee, 0xa5, 0x4c, 0x74, 0xc6, 
+    0x7b, 0x99, 0xec, 0xf5, 0x0b, 0x6c, 0xd1, 0xbc, 0x1c, 0x29, 
+    0x58, 0x0a, 0x13, 0xe3, 0x13, 0xb9, 0x77, 0x4c, 0xc1, 0x96, 
+    0xb7, 0xa8, 0x3f, 0x6d, 0x1a, 0x8c, 0x9d, 0xae, 0xa2, 0x08, 
+    0x09, 0x1c, 0x52, 0xe2, 0xa4, 0xd8, 0x67, 0x6a, 0xdf, 0x73, 
+    0xfc, 0x2a, 0x26, 0xce, 0x29, 0x5e, 0x06, 0x38, 0x43, 0xcc, 
+    0xb8, 0x34, 0xe0, 0x57, 0xbb, 0x15, 0x11, 0xf7, 0x64, 0x5c, 
+    0x36, 0x7a, 0x71, 0x92, 0x9d, 0xb2, 0x83, 0xb1, 0xa1, 0xdb, 
+    0xbd, 0xe5, 0xcd, 0xe1, 0x0b, 0x73, 0x71, 0x98, 0x34, 0xde, 
+    0x2d, 0x4d, 0x57, 0x10, 0x83, 0xe2, 0x6f, 0xe0, 0x77, 0xa1, 
+    0x7c, 0x75, 0x8e, 0x92, 0xea, 0xec, 0x40, 0x8d, 0xc2, 0x86, 
+    0x6b, 0x09, 0xd5, 0x30, 0x97, 0xb5, 0xc3, 0x1a, 0x79, 0x41, 
+    0x37, 0x87, 0x2e, 0xcc, 0x08, 0xb0, 0x04, 0xf2, 0x24, 0x04, 
+    0x23, 0xab, 0x6a, 0x0d, 0xf6, 0x5d, 0x41, 0xfb, 0x9d, 0x2b, 
+    0x0e, 0xf9, 0xee, 0x49, 0x53, 0x59, 0x8b, 0xc4, 0x9c, 0x42, 
+    0xcf, 0xec, 0x60, 0xb3, 0xd1, 0xe6, 0x74, 0x5a, 0xd3, 0x2e, 
+    0x85, 0xbe, 0x86, 0xb1, 0x94, 0x4e, 0x2b, 0xaf, 0x1b, 0xc3, 
+    0xe6, 0xdd, 0x0f, 0x5e, 0x6d, 0x46, 0x86, 0x8a, 0xe4, 0x32, 
+    0x93, 0x46, 0x32, 0x12, 0x48, 0x16, 0x2a, 0x81, 0x50, 0x95, 
+    0x40, 0xaf, 0x2a, 0xe6, 0x3b, 0x2e, 0xab, 0x91, 0xb2, 0x3d, 
+    0x81, 0x54, 0xd1, 0x8a, 0xc4, 0xb9, 0xed, 0x0a, 0x42, 0xdf, 
+    0xc5, 0x89, 0x1e, 0x3f, 0x61, 0xc5, 0x1e, 0x7c, 0xb3, 0x3d, 
+    0x0c, 0x95, 0x55, 0x34, 0xfa, 0x18, 0xce, 0x91, 0x5e, 0xb3, 
+    0x57, 0xfd, 0xcb, 0x5d, 0xaa, 0x38, 0x81, 0xae, 0xf4, 0xbf, 
+    0xf4, 0xc2, 0x38, 0x41, 0x68, 0x2c, 0x27, 0xeb, 0xfc, 0xc5, 
+    0xa2, 0x5c, 0x51, 0x92, 0x65, 0xf1, 0x45, 0xf9, 0xbd, 0x55, 
+    0x31, 0xb3, 0xa4, 0x56, 0x14, 0x92, 0xa7, 0x18, 0xcd, 0x54, 
+    0x80, 0x6e, 0x68, 0x39, 0xec, 0x7f, 0x3e, 0x91, 0x52, 0xfd, 
+    0xca, 0x05, 0x10, 0xb5, 0xe2, 0x23, 0x4e, 0x3f, 0xe7, 0xc6, 
+    0xef, 0xa0, 0x06, 0xe0, 0x6f, 0x36, 0x2b, 0xd1, 0xf7, 0x7d, 
+    0x07, 0xcd, 0xa9, 0xaf, 0xca, 0x46, 0xf2, 0x5b, 0x75, 0x25, 
+    0xa6, 0xba, 0x16, 0xa8, 0x28, 0xb5, 0xb3, 0x4f, 0xcc, 0x0a, 
+    0xf3, 0xaf, 0x60, 0xbd, 0x8d, 0x50, 0xb3, 0xbd, 0x29, 0xe2, 
+    0x43, 0xd8, 0xbf, 0x68, 0xec, 0x86, 0x3b, 0xac, 0xcc, 0xaa, 
+    0x70, 0xb8, 0x5b, 0xc3, 0xa4, 0xfc, 0xfb, 0x5d, 0x5c, 0x38, 
+    0x09, 0xcf, 0xee, 0x27, 0x87, 0x3c, 0x88, 0x48, 0xa0, 0x67, 
+    0x24, 0xf5, 0xcb, 0x60, 0xae, 0x8c, 0x7c, 0xe6, 0x23, 0x1d, 
+    0x23, 0x67, 0xa5, 0xb9, 0x20, 0xc1, 0xed, 0xae, 0xfd, 0xe9, 
+    0x35, 0x43, 0x99, 0x39, 0x27, 0x55, 0x7c, 0xe4, 0x30, 0x76, 
+    0xc7, 0x68, 0xc2, 0xb9, 0x95, 0xc6, 0x9a, 0x3f, 0x14, 0x1e, 
+    0xd5, 0x1a, 0x7e, 0xf9, 0x76, 0x99, 0x4b, 0x42, 0x5b, 0xfa, 
+    0x78, 0x4a, 0x15, 0xa3, 0xef, 0x2e, 0xae, 0x58, 0x2a, 0x81, 
+    0xb8, 0xf2, 0xb4, 0x49, 0x13, 0x32, 0x20, 0xd2, 0x8d, 0x66, 
+    0xc9, 0x21, 0xf9, 0x1a, 0x60, 0x75, 0x47, 0xc7, 0xab, 0x00, 
+    0xe8, 0xe7, 0xd8, 0xa2, 0x2b, 0x01, 0x48, 0x0d, 0x93, 0xfc, 
+    0x40, 0xc8, 0x88, 0x73, 0xd3, 0x29, 0x73, 0xda, 0x2d, 0x6d, 
+    0xc8, 0xd0, 0x66, 0x78, 0x69, 0x37, 0xc2, 0xd3, 0x1d, 0xab, 
+    0x5d, 0xa2, 0xbc, 0xd5, 0x0d, 0x74, 0xd4, 0xf4, 0x21, 0x86, 
+    0xc6, 0x00, 0x56, 0x28, 0x35, 0xc1, 0xfd, 0xad, 0x6b, 0x17, 
+    0xf0, 0xf4, 0x39, 0x82, 0x3b, 0xd6, 0xce, 0xbc, 0xa0, 0x59, 
+    0xcb, 0xec, 0xe6, 0xf2, 0xb9, 0xcb, 0x0e, 0x56, 0x58, 0x2e, 
+    0x11, 0xf6, 0x4e, 0x0b, 0xa6, 0xde, 0xb5, 0xde, 0x76, 0x32, 
+    0xbf, 0xa3, 0x52, 0x62, 0xbf, 0x3e, 0x13, 0x81, 0x4e, 0x2e, 
+    0xb7, 0x0b, 0xe4, 0xa3, 0x4e, 0xf1, 0x8f, 0x4d, 0x62, 0x2a, 
+    0x0a, 0x5e, 0xe9, 0xe3, 0xf7, 0x7b, 0x81, 0x52, 0xd5, 0x38, 
+    0x6d, 0x1f, 0x21, 0x41, 0x82, 0xbf, 0x58, 0x03, 0x94, 0xa0, 
+    0x89, 0x77, 0xac, 0x02, 0xe5, 0x51, 0x80, 0xd4, 0x37, 0xa7, 
+    0x90, 0x43, 0x47, 0x8d, 0x54, 0xf1, 0xe3, 0xce, 0x84, 0x28, 
+    0xb5, 0xb1, 0xef, 0xf2, 0x03, 0x26, 0xb9, 0x61, 0x8c, 0x74, 
+    0xed, 0x9b, 0xd7, 0x1e, 0xf3, 0x45, 0xe0, 0x8b, 0xa1, 0x8c, 
+    0xfc, 0xb4, 0x9b, 0x4e, 0x42, 0x92, 0x0c, 0xf2, 0x35, 0xf6, 
+    0x1b, 0x3f, 0x21, 0xea, 0x88, 0x8c, 0xce, 0x17, 0x1b, 0x99, 
+    0x13, 0x48, 0xe3, 0x66, 0x6a, 0x2f, 0xf6, 0x9b, 0x18, 0xf3, 
+    0xa3, 0x1c, 0x8c, 0x11, 0x93, 0xfe, 0x98, 0x85, 0x5c, 0x6e, 
+    0x84, 0x6e, 0x53, 0xbc, 0x46, 0xd7, 0x99, 0xea, 0xc1, 0xab, 
+    0x91, 0x61, 0x8f, 0xf5, 0xe6, 0xe7, 0xd1, 0x97, 0x43, 0x44, 
+    0xe7, 0x3a, 0x3c, 0x24, 0xe6, 0x30, 0x9a, 0x2d, 0x3f, 0x4f, 
+    0x67, 0x14, 0x73, 0x4c, 0x87, 0x70, 0xb8, 0x98, 0x03, 0xc8, 
+    0xf1, 0x83, 0x07, 0x93, 0xb4, 0x36, 0xdb, 0xd8, 0xce, 0x39, 
+    0x0d, 0xc3, 0x29, 0x50, 0x08, 0xe7, 0xb5, 0x51, 0xed, 0xb6, 
+    0xca, 0x0d, 0x67, 0x45, 0xed, 0x61, 0xdf, 0x39, 0x3a, 0x63, 
+    0xd6, 0x9a, 0x9c, 0x95, 0x13, 0x58, 0x0b, 0x80, 0x41, 0x4b, 
+    0x9e, 0x56, 0xd2, 0xb6, 0x81, 0x39, 0x2b, 0x94, 0x48, 0xc3, 
+    0x81, 0xbe, 0x09, 0x1a, 0xd3, 0x98, 0xb2, 0xfa, 0x47, 0x1b, 
+    0x26, 0xd8, 0xe2, 0x46, 0xe0, 0x46, 0x6e, 0xb9, 0x6a, 0x54, 
+    0x37, 0xa5, 0x67, 0x11, 0xfc, 0xca, 0x2b, 0xe3, 0xb6, 0xd1, 
+    0x4a, 0x99, 0x32, 0x28, 0x0a, 0xdb, 0xfd, 0x6a, 0xb1, 0xab, 
+    0xd5, 0x93, 0x39, 0x24, 0x0c, 0x7a, 0xb4, 0xb5, 0xe1, 0x1e, 
+    0x8c, 0x53, 0x65, 0xca, 0x2c, 0x17, 0x63, 0xba, 0x45, 0x00, 
+    0xdd, 0xbc, 0x60, 0xda, 0x76, 0x20, 0x7b, 0x99, 0xe2, 0xa2, 
+    0x03, 0x8b, 0xb5, 0x44, 0xa3, 0xae, 0x82, 0x8c, 0xdd, 0xfb, 
+    0xbd, 0xb3, 0xc6, 0x04, 0xa6, 0x52, 0x6e, 0x37, 0x3f, 0x92, 
+    0x64, 0x26, 0xc8, 0x0e, 0xd3, 0x8e, 0x71, 0x85, 0x0b, 0x6c, 
+    0xf2, 0x15, 0x78, 0x0a, 0x44, 0x84, 0xcf, 0xd8, 0x9c, 0x88, 
+    0x6a, 0x53, 0x1f, 0x74, 0xd5, 0xdd, 0xb1, 0x4a, 0x11, 0x18, 
+    0xb1, 0x72, 0xc4, 0x7c, 0xe9, 0x83, 0x30, 0x4f, 0xdd, 0xf5, 
+    0x74, 0xb5, 0x5c, 0xa7, 0x07, 0xfc, 0xe5, 0xd8, 0x1f, 0xf2, 
+    0x4a, 0x71, 0xe6, 0x8d, 0xf8, 0xbc, 0x91, 0x8e, 0x1d, 0x12, 
+    0x11, 0xca, 0xf2, 0x17, 0x76, 0x29, 0x03, 0x92, 0x24, 0x17, 
+    0xba, 0xcc, 0x1f, 0x3e, 0xb3, 0xdb, 0xbf, 0xe4, 0x04, 0xd6, 
+    0x12, 0xa2, 0xb2, 0x2e, 0x93, 0x89, 0x5e, 0x8f, 0xb6, 0x8c, 
+    0xf7, 0x80, 0xfc, 0x17, 0xf7, 0x3f, 0x28, 0xaa, 0xc5, 0x8b, 
+    0xa2, 0x84, 0x9a, 0x9f, 0xde, 0x05, 0x0f, 0xa1, 0x2d, 0x05, 
+    0x75, 0xed, 0x53, 0x96, 0xbc, 0xba, 0xcd, 0x99, 0xf7, 0xbb, 
+    0xe0, 0x47, 0x02, 0xb4, 0x4c, 0x10, 0xce, 0xd9, 0x6a, 0x27, 
+    0xbe, 0x35, 0x4d, 0xf8, 0x5c, 0x2b, 0xfb, 0x82, 0x72, 0x0c, 
+    0xec, 0x29, 0x95, 0xab, 0xa2, 0x52, 0xe4, 0x54, 0x30, 0x6c, 
+    0xad, 0xdf, 0x76, 0x29, 0x9c, 0x21, 0x0d, 0x06, 0x61, 0x6d, 
+    0x99, 0x3b, 0x51, 0xee, 0xb0, 0x47, 0x65, 0x61, 0xf8, 0x4b, 
+    0xc7, 0xf0, 0x28, 0x7f, 0xc8, 0x24, 0x80, 0x22, 0x6c, 0xc9, 
+    0x73, 0x39, 0x9b, 0x9f, 0xe5, 0x93, 0x14, 0xc4, 0x7c, 0x91, 
+    0x35, 0xec, 0x04, 0xa6, 0x5c, 0xaa, 0xe5, 0xfd, 0x35, 0x81, 
+    0x05, 0x83, 0x54, 0xc4, 0xad, 0x47, 0x0e, 0x60, 0x6c, 0x52, 
+    0x7d, 0x06, 0x76, 0x9e, 0xe6, 0x91, 0xc3, 0xf4, 0x5b, 0xf1, 
+    0xad, 0x1a, 0x51, 0xac, 0x49, 0x10, 0xb2, 0x18, 0x25, 0x06, 
+    0x90, 0x29, 0x46, 0xe3, 0xd5, 0xe1, 0xca, 0x56, 0x1a, 0x6b, 
+    0x24, 0x52, 0xf8, 0x08, 0xab, 0x52, 0x6c, 0x8d, 0xf7, 0x59, 
+    0xaa, 0x0a, 0xf5, 0x37, 0x31, 0x64, 0x3b, 0xb7, 0x12, 0x76, 
+    0xc4, 0xb4, 0x83, 0x9d, 0x76, 0x39, 0x1f, 0x64, 0x85, 0xed, 
+    0x67, 0x98, 0x97, 0x57, 0x76, 0xe9, 0x0d, 0xc8, 0x4f, 0xc4, 
+    0xc7, 0xf8, 0xad, 0x1d, 0x6f, 0xf3, 0xae, 0x4a, 0xf0, 0x3d, 
+    0x71, 0x73, 0xba, 0x74, 0xa7, 0xd0, 0xc7, 0x62, 0x69, 0xb0, 
+    0x5e, 0x25, 0x87, 0xe8, 0xb7, 0xa5, 0x54, 0x54, 0xe2, 0x7e, 
+    0x13, 0xd1, 0x6d, 0x70, 0x5b, 0x65, 0x49, 0x7c, 0x40, 0xdd, 
+    0xba, 0x95, 0x34, 0xc6, 0xa1, 0x01, 0x47, 0x67, 0xa9, 0x42, 
+    0xde, 0xc7, 0xb1, 0x1e, 0x2e, 0xd6, 0x83, 0x1e, 0x34, 0x30, 
+    0xd1, 0x76, 0x85, 0x33, 0x41, 0x72, 0xd5, 0x28, 0xc4, 0xdd, 
+    0x41, 0x78, 0x6a, 0x64, 0xee, 0x45, 0xa1, 0x90, 0x9e, 0xf2, 
+    0x16, 0xb1, 0x4d, 0xd1, 0xbc, 0x11, 0x2b, 0x0b, 0x19, 0xbf, 
+    0x7e, 0x63, 0xf1, 0x40, 0xbd, 0x16, 0x96, 0xb3, 0xa7, 0x1e, 
+    0x0e, 0x8a, 0x92, 0xbb, 0x7d, 0xcd, 0x03, 0x6e, 0x1c, 0x0a, 
+    0x58, 0x46, 0x88, 0xac, 0x80, 0x63, 0xd5, 0x92, 0xa9, 0xeb, 
+    0xec, 0x17, 0x9a, 0x0a, 0xe3, 0x75, 0xfe, 0x1f, 0xb8, 0x70, 
+    0xf5, 0x3c, 0x01, 0x2a, 0x10, 0x9d, 0xd7, 0xef, 0xa7, 0xb2, 
+    0xb5, 0xab, 0x5a, 0x77, 0x53, 0x61, 0x45, 0x9f, 0x1f, 0xca, 
+    0x95, 0xe0, 0x83, 0xf7, 0x11, 0xef, 0x52, 0x3d, 0xcc, 0x55, 
+    0xba, 0x7d, 0x71, 0x49, 0x0c, 0x26, 0x76, 0x89, 0x4c, 0x44, 
+    0x10, 0xbf, 0x64, 0xad, 0x56, 0x16, 0xb1, 0x83, 0xd0, 0x59, 
+    0xa5, 0x2a, 0x6a, 0x94, 0x78, 0xd4, 0x0f, 0x78, 0x1b, 0x28, 
+    0x15, 0xed, 0x05, 0x14, 0x38, 0xbe, 0x67, 0x76, 0x13, 0x83, 
+    0x6b, 0xca, 0x16, 0xd3, 0x38, 0xf2, 0x57, 0xb7, 0x72, 0x21, 
+    0xa6, 0x1e, 0x94, 0x3f, 0x61, 0x0b, 0x34, 0x9b, 0x30, 0x39, 
+    0x4e, 0x3a, 0x89, 0x03, 0x9c, 0x99, 0x2b, 0x9f, 0x28, 0x3f, 
+    0xde, 0x15, 0xbf, 0xb0, 0xfd, 0x3a, 0xda, 0x58, 0x73, 0x25, 
+    0x05, 0xab, 0x1d, 0xcf, 0xbf, 0xe5, 0x76, 0x30, 0xb5, 0x75, 
+    0x22, 0x3e, 0x6b, 0xfb, 0x59, 0x30, 0xe3, 0x3a, 0x62, 0x72, 
+    0xf8, 0xf5, 0x5a, 0xa7, 0x76, 0xbd, 0xef, 0xa4, 0x7b, 0x7e, 
+    0xdf, 0x4f, 0x94, 0x52, 0xd7, 0x1d, 0x12, 0xb2, 0x00, 0x56, 
+    0xa8, 0x55, 0x3c, 0x49, 0x7c, 0x48, 0x10, 0x9b, 0x97, 0x9a, 
+    0xd7, 0xa3, 0xfa, 0x54, 0xbc, 0xf6, 0x53, 0x56, 0x98, 0x7e, 
+    0xce, 0xab, 0x5b, 0x85, 0xe2, 0x2f, 0x95, 0x8c, 0x36, 0xbd, 
+    0x78, 0xe8, 0x13, 0xc5, 0x9b, 0x16, 0x72, 0xea, 0xba, 0x13, 
+    0x36, 0xfb, 0xda, 0xf2, 0x2c, 0x3f, 0x2d, 0x94, 0x1b, 0xf6, 
+    0x18, 0x83, 0x8d, 0x06, 0x30, 0xaa, 0x59, 0x52, 0x9b, 0xbc, 
+    0xbe, 0xe1, 0x79, 0x22, 0xb0, 0xe4, 0xd8, 0x37, 0xad, 0xe9, 
+    0x86, 0xc6, 0x67, 0x73, 0x1c, 0x2b, 0xda, 0x1f, 0xa5, 0x8e, 
+    0x48, 0x42, 0x52, 0x55, 0x34, 0x59, 0x12, 0xfa, 0x4f, 0xfb, 
+    0xb6, 0x6f, 0xb7, 0x9d, 0x7d, 0x2c, 0x0d, 0xc3, 0x2c, 0x25, 
+    0xd5, 0xb3, 0x90, 0xc6, 0x1a, 0x71, 0x3a, 0x8f, 0xa7, 0x5b, 
+    0xf4, 0xc7, 0xb0, 0x44, 0x98, 0x4a, 0x05, 0xa2, 0x57, 0x92, 
+    0x6e, 0xc1, 0xb4, 0xf8, 0xba, 0xd8, 0x47, 0xa4, 0x09, 0xe7, 
+    0xa1, 0x19, 0x67, 0x45, 0x1a, 0x00, 0x5c, 0x3a, 0x18, 0xf3, 
+    0x29, 0x45, 0x66, 0x3d, 0xfd, 0xc6, 0xa6, 0x67, 0x2f, 0xa5, 
+    0xc6, 0xcb, 0x6e, 0xe7, 0x01, 0xab, 0xd3, 0x39, 0xa9, 0x1e, 
+    0x52, 0x64, 0xd6, 0x30, 0x6c, 0x99, 0x4c, 0x24, 0x5e, 0xed, 
+    0x75, 0xe7, 0xb3, 0x94, 0x60, 0xf4, 0x75, 0xfe, 0x4f, 0xc1, 
+    0x6b, 0xe5, 0x2c, 0x11, 0xd3, 0x08, 0xa2, 0x3b, 0x86, 0x48, 
+    0x7a, 0xb6, 0xbf, 0x91, 0x08, 0x12, 0x71, 0x65, 0xa9, 0xbf, 
+    0x3e, 0xb8, 0xba, 0x9e, 0x07, 0x70, 0x66, 0xdc, 0xab, 0x92, 
+    0x88, 0x25, 0x3e, 0x71, 0xdb, 0x90, 0x29, 0x5f, 0xad, 0xa3, 
+    0x51, 0xc2, 0x82, 0x4c, 0xdb, 0x8f, 0xdf, 0x72, 0x1f, 0xd0, 
+    0x37, 0x75, 0x01, 0xa3, 0x38, 0xed, 0x36, 0x0a, 0x1b, 0xdf, 
+    0xa0, 0x1a, 0x60, 0x65, 0x54, 0xc7, 0x65, 0xd2, 0xc5, 0xe2, 
+    0x78, 0xce, 0xb6, 0x51, 0xe4, 0x84, 0xb0, 0x67, 0x90, 0x85, 
+    0x80, 0x30, 0xdb, 0x88, 0x17, 0x29, 0x0b, 0x36, 0x5a, 0x40, 
+    0x85, 0x80, 0xaa, 0x9c, 0xb8, 0x6d, 0x06, 0xc5, 0xd6, 0xcc, 
+    0xd7, 0x25, 0x83, 0xd1, 0xb2, 0x4d, 0x1e, 0x0d, 0x89, 0x68, 
+    0x26, 0x07, 0x94, 0x92, 0x97, 0x21, 0xda, 0xa4, 0x7e, 0xc7, 
+    0x05, 0x10, 0x1a, 0x98, 0x08, 0xae, 0xb4, 0xf3, 0x33, 0x20, 
+    0xe4, 0xd0, 0x65, 0x94, 0x4c, 0x05, 0xa9, 0xec, 0xec, 0x7f, 
+    0x76, 0x31, 0xe0, 0x33, 0x33, 0x17, 0x0b, 0x6f, 0xb8, 0x8c, 
+    0x8e, 0xb3, 0x54, 0x54, 0x39, 0x91, 0xfe, 0x6e, 0x90, 0xf5, 
+    0x5c, 0x73, 0xbc, 0x2b, 0xb5, 0x1e, 0x7b, 0x2a, 0x7d, 0x83, 
+    0x80, 0x26, 0x8e, 0x9e, 0x41, 0x5c, 0x30, 0x56, 0x12, 0xf8, 
+    0xdd, 0xb9, 0xbe, 0xd7, 0x73, 0xa9, 0xba, 0x7f, 0x24, 0xf5, 
+    0x11, 0xe4, 0xf1, 0x2e, 0xd1, 0x23, 0xc9, 0x4b, 0xa0, 0xfa, 
+    0x50, 0x49, 0x9c, 0x0a, 0xf5, 0xc0, 0x4d, 0xcf, 0x0a, 0xfd, 
+    0xa5, 0xcc, 0x2b, 0x4c, 0x62, 0x95, 0x8c, 0xd0, 0x6b, 0x28, 
+    0x20, 0xf3, 0xdc, 0xec, 0xa4, 0x7b, 0xd6, 0xea, 0xb2, 0x03, 
+    0xaf, 0x31, 0x68, 0x0f, 0xed, 0xe6, 0x05, 0x67, 0xf2, 0x35, 
+    0x87, 0xd6, 0xb2, 0xb0, 0x9b, 0x20, 0x1e, 0xf3, 0x1e, 0x0f, 
+    0x6d, 0x18, 0x91, 0xdd, 0x1f, 0x49, 0xb4, 0x63, 0x8d, 0x92, 
+    0x01, 0x2d, 0x14, 0x70, 0x81, 0x60, 0x60, 0x85, 0xe5, 0xdf, 
+    0x84, 0x10, 0x27, 0xfc, 0x83, 0x9e, 0x89, 0x06, 0x79, 0x0c, 
+    0x38, 0xbe, 0x4d, 0xca, 0xa5, 0xc7, 0x55, 0x82, 0x6c, 0x22, 
+    0x90, 0x11, 0x6c, 0xc6, 0x74, 0xe7, 0x19, 0x26, 0xb5, 0xc0, 
+    0x3b, 0x4e, 0x7c, 0x58, 0xf5, 0xd4, 0xce, 0x5e, 0xcd, 0x1b, 
+    0x8e, 0xb8, 0xe2, 0x17, 0xec, 0xf5, 0xaa, 0x00, 0x58, 0x1d, 
+    0x6f, 0xa0, 0x2a, 0xf4, 0x59, 0x0a, 0xa2, 0x48, 0xa0, 0x28, 
+    0xbb, 0x9a, 0x60, 0x84, 0x58, 0x88, 0x0f, 0xdd, 0x2e, 0x59, 
+    0x6b, 0x0d, 0x10, 0x04, 0x2b, 0x9a, 0x16, 0x5e, 0x2d, 0xed, 
+    0x87, 0x55, 0x4b, 0xd2, 0x41, 0x4d, 0xfe, 0x1e, 0x7e, 0x44, 
+    0xa9, 0xbf, 0xdf, 0x61, 0x9f, 0x8c, 0xad, 0x21, 0xfe, 0x03, 
+    0x84, 0xb5, 0xdc, 0x47, 0x02, 0x33, 0xc5, 0x54, 0xa5, 0xdc, 
+    0x99, 0x45, 0x3e, 0x20, 0x4d, 0xb3, 0x31, 0x37, 0x64, 0xd7, 
+    0xab, 0x7b, 0x3e, 0x1b, 0x76, 0xdc, 0x66, 0xa0, 0xa2, 0x9c, 
+    0x59, 0xbd, 0xe9, 0xcc, 0xe0, 0xae, 0x2a, 0x78, 0xc0, 0x10, 
+    0x85, 0x5a, 0x3c, 0x35, 0x75, 0x1e, 0xe6, 0x97, 0x68, 0x62, 
+    0x10, 0xb4, 0x77, 0xdf, 0xc2, 0x81, 0xe7, 0x53, 0x79, 0xdb, 
+    0x8e, 0xc3, 0x01, 0x19, 0xc3, 0xbc, 0xd9, 0x2c, 0x17, 0xb2, 
+    0xa9, 0x65, 0x5c, 0x75, 0xee, 0x1a, 0x9a, 0x65, 0x5b, 0x44, 
+    0xd5, 0xe2, 0xd2, 0xd5, 0x79, 0x27, 0x53, 0x2b, 0x5a, 0xe5, 
+    0x61, 0x8f, 0x21, 0x52, 0xb7, 0x7a, 0x27, 0x33, 0x94, 0x74, 
+    0x61, 0x0b, 0x82, 0x72, 0xc6, 0x76, 0xd7, 0x67, 0x1b, 0x18, 
+    0x88, 0x2a, 0x67, 0xf3, 0xea, 0x45, 0x8a, 0x38, 0x82, 0x61, 
+    0xbb, 0x02, 0x5b, 0x1f, 0x93, 0xa4, 0xa4, 0x55, 0x3c, 0x83, 
+    0x67, 0x43, 0xb2, 0x0f, 0xb6, 0x9c, 0x88, 0x9b, 0x4b, 0xa8, 
+    0x3c, 0xdc, 0xa1, 0x88, 0x08, 0xf7, 0x16, 0x29, 0xef, 0x36, 
+    0xab, 0x69, 0x09, 0x08, 0xa9, 0x4c, 0xb0, 0x0b, 0x21, 0xc9, 
+    0xc6, 0x12, 0xf3, 0xfa, 0x2a, 0x92, 0x17, 0xc3, 0xf4, 0xe7, 
+    0x71, 0x1e, 0xab, 0x4b, 0x78, 0xdd, 0x65, 0xa8, 0xb7, 0x63, 
+    0xf8, 0xe4, 0x82, 0x5e, 0x21, 0xa9, 0xc8, 0x87, 0x0f, 0x15, 
+    0xf3, 0xf2, 0x8e, 0x44, 0xf3, 0x73, 0x21, 0xfe, 0x52, 0x43, 
+    0x2b, 0xe7, 0xae, 0x14, 0x50, 0xaa, 0x3d, 0x07, 0x91, 0x1d, 
+    0xc1, 0x37, 0xcd, 0xd1, 0x45, 0x36, 0xa1, 0x68, 0xe9, 0x67, 
+    0x5c, 0x38, 0x9e, 0xcb, 0xf2, 0x09, 0x5c, 0xf8, 0x86, 0xfe, 
+    0x46, 0x53, 0xcc, 0x96, 0xbb, 0xd6, 0xca, 0xc1, 0x8d, 0x55, 
+    0xe2, 0xc6, 0x26, 0xba, 0x4f, 0xab, 0x65, 0x0a, 0x65, 0x55, 
+    0xf1, 0x0d, 0x6c, 0xbe, 0x31, 0x05, 0xab, 0xa1, 0xec, 0x1d, 
+    0xc9, 0x76, 0xf6, 0x26, 0xc7, 0x15, 0x8a, 0xf9, 0x8c, 0x23, 
+    0x7a, 0x0b, 0x66, 0xc5, 0xb5, 0x2a, 0xf2, 0x16, 0x84, 0xef, 
+    0x44, 0x94, 0x4b, 0xcb, 0x18, 0x96, 0x05, 0x44, 0xb8, 0xe3, 
+    0xa3, 0xe5, 0x5a, 0x98, 0x3b, 0xd3, 0x3f, 0x4a, 0x80, 0xbc, 
+    0x12, 0x51, 0x92, 0xc6, 0x1d, 0xeb, 0x44, 0x1d, 0x6e, 0x80, 
+    0xfe, 0x3c, 0x7f, 0x89, 0x6d, 0xaf, 0xc3, 0x71, 0xdb, 0xcf, 
+    0xc5, 0xa9, 0x81, 0x87, 0xec, 0xc4, 0x61, 0xf4, 0x3e, 0x16, 
+    0x25, 0x34, 0x0e, 0x15, 0x07, 0xc8, 0xe1, 0x74, 0xc4, 0x7c, 
+    0x8c, 0xfe, 0xa8, 0x1b, 0xd5, 0xad, 0xbb, 0x46, 0x72, 0x6b, 
+    0x46, 0x04, 0x90, 0xed, 0x5f, 0x64, 0xfd, 0x23, 0x35, 0xe0, 
+    0x23, 0xda, 0xa1, 0x76, 0xb0, 0x73, 0xb3, 0x7b, 0xf7, 0xa3, 
+    0x6a, 0xd2, 0xb8, 0x2f, 0xc2, 0x6b, 0xe3, 0x01, 0xd2, 0x28, 
+    0xcd, 0xf4, 0xd2, 0xc2, 0x6c, 0x75, 0xec, 0xc1, 0x76, 0x8f, 
+    0xd4, 0x4c, 0xc0, 0xe0, 0xd6, 0x33, 0x45, 0xbb, 0x21, 0xe7, 
+    0x1f, 0xf9, 0x9e, 0x66, 0x18, 0x52, 0xfa, 0x35, 0x7b, 0x33, 
+    0x75, 0x6d, 0x3b, 0x39, 0xf8, 0xce, 0x75, 0x22, 0x57, 0x65, 
+    0x6a, 0x34, 0x1e, 0xe9, 0x99, 0x66, 0x78, 0xa1, 0xbc, 0xfd, 
+    0xb7, 0xab, 0x98, 0xa7, 0x4b, 0x62, 0x45, 0x10, 0xc4, 0xf4, 
+    0xb0, 0xc2, 0xf1, 0x6d, 0xd3, 0x31, 0x19, 0x8b, 0xc4, 0xa3, 
+    0x36, 0x7e, 0xf4, 0x7f, 0xf4, 0x72, 0xb3, 0xae, 0x75, 0xdb, 
+    0x5b, 0xf6, 0x5d, 0xe7, 0xf3, 0x23, 0x90, 0xf3, 0x0d, 0x5c, 
+    0x0d, 0x54, 0x62, 0x51, 0xbc, 0x20, 0xf2, 0x45, 0x10, 0xb6, 
+    0x3c, 0x93, 0xb6, 0x01, 0x6c, 0x7e, 0x45, 0xc1, 0xad, 0xe2, 
+    0xe1, 0x1c, 0x15, 0x37, 0x30, 0x1f, 0x90, 0x62, 0x6e, 0x6d, 
+    0x6a, 0x83, 0xfd, 0xca, 0xd7, 0x76, 0x2b, 0xb6, 0xc5, 0xf9, 
+    0x1d, 0x54, 0x28, 0xd8, 0xcf, 0xa8, 0xac, 0xf8, 0xc1, 0x54, 
+    0xf8, 0x5c, 0x23, 0x48, 0x6e, 0xb8, 0x7a, 0x02, 0x72, 0x30, 
+    0x0e, 0x11, 0x83, 0xb5, 0x2d, 0x83, 0xad, 0xc3, 0x3e, 0xf6, 
+    0x8f, 0xf3, 0xb6, 0xa0, 0xd7, 0xbb, 0x43, 0xc1, 0x75, 0xfe, 
+    0x9e, 0x1d, 0x69, 0x2e, 0x83, 0x03, 0x43, 0x43, 0xf6, 0xfa, 
+    0x43, 0x54, 0xd3, 0xab, 0xfc, 0x53, 0xde, 0x87, 0x92, 0x7b, 
+    0xc4, 0x3a, 0x5a, 0xe4, 0xee, 0xef, 0x18, 0x96, 0x9c, 0xe0, 
+    0xea, 0xd1, 0xce, 0x3a, 0x67, 0x56, 0x0c, 0xe8, 0x23, 0x0d, 
+    0x34, 0x1e, 0x0d, 0x80, 0x03, 0x8f, 0x90, 0xc5, 0xba, 0x01, 
+    0x43, 0xfb, 0x94, 0x99, 0x3b, 0x19, 0x13, 0x30, 0x5f, 0xd8, 
+    0x0b, 0xb1, 0x10, 0x8d, 0x61, 0x20, 0x99, 0x3b, 0x2d, 0xbe, 
+    0x64, 0xf7, 0x7a, 0x97, 0x84, 0x12, 0xc0, 0x44, 0x85, 0x0b, 
+    0x4b, 0x2c, 0x1c, 0x83, 0x7f, 0x6a, 0xd7, 0x71, 0x6b, 0x5b, 
+    0x71, 0xf0, 0xc2, 0xbb, 0x9a, 0xfb, 0xad, 0x5c, 0x27, 0x07, 
+    0x02, 0x5f, 0x54, 0xd7, 0xb5, 0x5f, 0x95, 0xd1, 0x15, 0xb0, 
+    0x27, 0x1d, 0xda, 0x3f, 0x87, 0x5d, 0x8b, 0x20, 0x2c, 0x12, 
+    0x6b, 0x11, 0x67, 0xcb, 0xd7, 0xa7, 0xc5, 0xa1, 0x55, 0x91, 
+    0xed, 0x2a, 0x11, 0xc4, 0x84, 0x1c, 0xab, 0xeb, 0x68, 0x84, 
+    0x18, 0x94, 0x5b, 0xb1, 0x39, 0x65, 0xca, 0x6c, 0x14, 0xe1, 
+    0x65, 0xec, 0x18, 0xea, 0x7c, 0x5c, 0x7f, 0x4f, 0xda, 0xe9, 
+    0x4d, 0x30, 0xaf, 0x96, 0x47, 0x01, 0x7a, 0xfa, 0xd2, 0x97, 
+    0x37, 0x73, 0xa6, 0xd5, 0x4b, 0x36, 0xb4, 0xda, 0xa7, 0x69, 
+    0x43, 0x38, 0xa6, 0x59, 0x20, 0x4b, 0xd4, 0x1c, 0x2e, 0xbb, 
+    0x18, 0xed, 0xd3, 0x3e, 0xef, 0xd2, 0xac, 0xb1, 0xac, 0x51, 
+    0x23, 0x8b, 0xbc, 0xc3, 0x01, 0x99, 0xd4, 0x2c, 0x6c, 0x33, 
+    0xf5, 0xe4, 0x3d, 0x3e, 0x6d, 0xae, 0x05, 0x62, 0xeb, 0xb3, 
+    0xb6, 0x66, 0x64, 0x42, 0x32, 0xba, 0x81, 0xd0, 0xa1, 0x03, 
+    0xef, 0xc9, 0x4b, 0xee, 0x22, 0x0a, 0x64, 0x08, 0x8a, 0x99, 
+    0xf1, 0xc0, 0x4e, 0x9e, 0x79, 0x4a, 0xa0, 0x9b, 0xb1, 0xba, 
+    0x84, 0xbd, 0xe9, 0x5b, 0xec, 0x24, 0x4b, 0x34, 0x5d, 0x2a, 
+    0xcc, 0xd2, 0x5f, 0xc7, 0x58, 0xe6, 0x47, 0xbb, 0xb9, 0xf7, 
+    0xb4, 0xc0, 0x1d, 0x2b, 0x21, 0xc0, 0x7d, 0x76, 0xd5, 0xc2, 
+    0x40, 0xa5, 0xe4, 0x9b, 0x13, 0x0d, 0xb9, 0xb7, 0xb3, 0x2f, 
+    0xcd, 0xc4, 0xef, 0xef, 0x48, 0x28, 0xe7, 0x87, 0x11, 0x46, 
+    0xed, 0xac, 0xbf, 0xec, 0x69, 0x30, 0xa7, 0x61, 0x92, 0xcf, 
+    0x5b, 0xa2, 0x69, 0xe0, 0x14, 0xa2, 0xf9, 0x29, 0xf8, 0xc5, 
+    0x7e, 0xe2, 0xf0, 0xfc, 0xf0, 0x32, 0xf3, 0x0c, 0xab, 0x7d, 
+    0xbb, 0xbd, 0x89, 0x54, 0x98, 0x47, 0x6a, 0xb2, 0x87, 0x45, 
+    0xd6, 0x06, 0xd4, 0xbf, 0x77, 0x95, 0xf4, 0x61, 0x38, 0x96, 
+    0xad, 0x56, 0xfa, 0x71, 0x9b, 0xde, 0x15, 0x0f, 0xc3, 0x3d, 
+    0xf5, 0x21, 0x63, 0x9b, 0xfe, 0x70, 0x2d, 0xfe, 0xc2, 0xcf, 
+    0x26, 0xfd, 0xd6, 0xb6, 0x3e, 0xcc, 0x05, 0x36, 0x11, 0xe2, 
+    0xb7, 0xdd, 0xc4, 0x6f, 0xc7, 0xac, 0xfe, 0x6f, 0x5b, 0xb4, 
+    0xe6, 0x47, 0x53, 0x30, 0x69, 0x56, 0x5b, 0x5c, 0xee, 0xc4, 
+    0xd0, 0x5b, 0x1e, 0x1b, 0xb8, 0xbe, 0xcd, 0xde, 0x28, 0xe6, 
+    0x90, 0x2d, 0xcf, 0xa0, 0xfb, 0x11, 0x3b, 0xda, 0xba, 0x2e, 
+    0x05, 0xa8, 0x9e, 0x3b, 0x07, 0xce, 0x87, 0xf0, 0xc4, 0xb4, 
+    0x2c, 0x33, 0x32, 0xda, 0x5a, 0x86, 0xba, 0x97, 0x1a, 0xf4, 
+    0xbf, 0xbf, 0x7c, 0x0c, 0xa8, 0xf1, 0x17, 0x99, 0x59, 0xa5, 
+    0x32, 0xc2, 0x48, 0x2d, 0xa6, 0x02, 0x18, 0xc1, 0x97, 0x95, 
+    0x7b, 0x6f, 0x62, 0xf8, 0xbb, 0x80, 0x83, 0xb9, 0xe3, 0x27, 
+    0xdd, 0x38, 0xb0, 0x26, 0xea, 0xde, 0xa5, 0x45, 0xf3, 0x05, 
+    0xde, 0x51, 0x36, 0x97, 0xc2, 0x80, 0x13, 0x9d, 0xd8, 0x58, 
+    0xd8, 0xc3, 0xac, 0xaf, 0xfd, 0xf5, 0x19, 0x21, 0x26, 0x9d, 
+    0xe8, 0xeb, 0xc5, 0x94, 0xd2, 0xb2, 0x56, 0x7e, 0xe6, 0xab, 
+    0x3c, 0x9d, 0x17, 0xaa, 0x8a, 0x15, 0x6d, 0xd0, 0x01, 0x20, 
+    0x01, 0xe0, 0x76, 0x44, 0x9c, 0xd0, 0x36, 0x56, 0xbf, 0xc2, 
+    0x3d, 0xee, 0x5a, 0x11, 0x22, 0xbf, 0x05, 0xf5, 0x98, 0xfe, 
+    0x88, 0x86, 0x9f, 0x5b, 0x55, 0xce, 0x8b, 0x65, 0x49, 0x24, 
+    0x13, 0x30, 0x44, 0xb4, 0x32, 0x69, 0x23, 0x13, 0xab, 0xb9, 
+    0x76, 0x7c, 0x7d, 0xb1, 0x14, 0xdf, 0x35, 0x72, 0x82, 0xbd, 
+    0xda, 0x6a, 0xa2, 0x91, 0x5c, 0x9d, 0x2a, 0xe0, 0x89, 0x4d, 
+    0x89, 0xa7, 0xf4, 0x16, 0x7e, 0xc0, 0x74, 0x7e, 0x7d, 0xcc, 
+    0x2f, 0xf5, 0x63, 0x1a, 0xcb, 0x03, 0x18, 0x17, 0x2a, 0x21, 
+    0x10, 0x47, 0x83, 0x77, 0x71, 0x2a, 0xdb, 0x82, 0x09, 0xb8, 
+    0x0e, 0x5c, 0x2d, 0x9d, 0x56, 0x12, 0x20, 0x3d, 0x7c, 0xea, 
+    0xb1, 0x2f, 0x2a, 0x65, 0xc3, 0xb3, 0xc1, 0xc8, 0x03, 0x74, 
+    0xe4, 0xe9, 0x24, 0x3a, 0x31, 0xcf, 0x5b, 0x66, 0x3d, 0x23, 
+    0x1c, 0x39, 0x3b, 0x8b, 0x66, 0x46, 0x0b, 0xcc, 0x57, 0xe7, 
+    0xb5, 0xe7, 0x35, 0x7c, 0xd0, 0xe2, 0x94, 0x22, 0x33, 0x2a, 
+    0xb4, 0x01, 0x70, 0x1f, 0x03, 0x8c, 0xde, 0x3b, 0x70, 0x13, 
+    0x11, 0x1b, 0xe4, 0x22, 0xf3, 0x2b, 0x1f, 0x9d, 0x76, 0x08, 
+    0xec, 0x7c, 0x88, 0x19, 0xc0, 0x73, 0xef, 0x23, 0x94, 0xda, 
+    0xa5, 0xa2, 0x2d, 0xf4, 0xd9, 0xc5, 0x60, 0x87, 0x23, 0xb7, 
+    0x35, 0x29, 0x37, 0x04, 0xae, 0x7f, 0x7c, 0x81, 0x78, 0x49, 
+    0x30, 0xfc, 0x27, 0x98, 0x83, 0xa9, 0xf2, 0xcc, 0xe3, 0x45, 
+    0xa1, 0xe2, 0xe9, 0xf2, 0x36, 0x3b, 0x43, 0x8d, 0xe9, 0x39, 
+    0xa6, 0x2f, 0x00, 0xed, 0xf4, 0x88, 0x61, 0xd7, 0xfc, 0xad, 
+    0x5a, 0x94, 0x8f, 0x84, 0xc3, 0x38, 0xa0, 0x4e, 0xc3, 0x0d, 
+    0x30, 0xe1, 0x39, 0x36, 0xc8, 0x86, 0x41, 0xb8, 0x72, 0xd8, 
+    0x48, 0x4a, 0x08, 0x56, 0xa1, 0xbc, 0x52, 0x94, 0xee, 0x36, 
+    0x7f, 0xc4, 0xf3, 0x73, 0x59, 0x12, 0x0a, 0x03, 0x4c, 0xbe, 
+    0x10, 0x96, 0xdd, 0x37, 0xdd, 0x48, 0x41, 0xab, 0xcd, 0xf7, 
+    0x99, 0xc8, 0xac, 0x23, 0x23, 0x56, 0xba, 0x8b, 0x24, 0x0d, 
+    0x18, 0x15, 0x6d, 0x5b, 0x88, 0x01, 0x24, 0xe9, 0x99, 0x27, 
+    0x56, 0x3d, 0xdc, 0xe4, 0xf2, 0x41, 0x2b, 0x65, 0x1f, 0xe0, 
+    0x84, 0x2f, 0x31, 0xbd, 0x27, 0xd8, 0x0e, 0x99, 0xea, 0xc1, 
+    0xa9, 0x0b, 0x21, 0x2a, 0xe7, 0x57, 0xc0, 0xa3, 0x79, 0x1f, 
+    0xf7, 0xbb, 0x4e, 0x50, 0xf2, 0x37, 0x53, 0xed, 0x73, 0xb2, 
+    0xde, 0x08, 0xef, 0xa5, 0x9e, 0xcd, 0x38, 0x51, 0x55, 0xa1, 
+    0xf6, 0xac, 0x78, 0xf5, 0xd6, 0xd9, 0x7d, 0x92, 0x34, 0x57, 
+    0x78, 0xd1, 0x47, 0x17, 0xde, 0x10, 0x7e, 0xce, 0xf8, 0x1c, 
+    0x09, 0xcc, 0xe0, 0xea, 0x6d, 0x55, 0xec, 0xcb, 0x49, 0xaa, 
+    0x22, 0x5a, 0x18, 0xe7, 0x10, 0xb2, 0xb5, 0xf0, 0x9d, 0xdf, 
+    0xc5, 0x49, 0xe8, 0x4d, 0x9d, 0xcd, 0x51, 0x53, 0xf5, 0x95, 
+    0x37, 0x4b, 0xe4, 0x57, 0xca, 0x3f, 0x55, 0x68, 0x75, 0xc1, 
+    0xcb, 0xd6, 0x5b, 0xb7, 0x4b, 0x62, 0xb5, 0xce, 0x74, 0x98, 
+    0x84, 0x01, 0x28, 0xc8, 0x1c, 0x21, 0xc2, 0xc5, 0xa8, 0xdf, 
+    0x69, 0xd8, 0xd5, 0xe2, 0x4d, 0x72, 0xc1, 0xd5, 0x2f, 0x13, 
+    0x94, 0x6d, 0x1c, 0xc6, 0x46, 0xdf, 0x70, 0x57, 0xdd, 0x9b, 
+    0xb8, 0x90, 0x37, 0xad, 0xe9, 0xf0, 0x09, 0x1f, 0xfb, 0xb1, 
+    0x99, 0x6e, 0xa5, 0xe7, 0x9f, 0x94, 0xf1, 0xd4, 0xbd, 0x85, 
+    0x13, 0xfd, 0xbf, 0x18, 0x86, 0xc2, 0xb7, 0x95, 0xa9, 0x4c, 
+    0x26, 0x2f, 0x54, 0xcf, 0xc1, 0x64, 0xf6, 0x85, 0x13, 0xc9, 
+    0xd7, 0xd2, 0xef, 0xd7, 0xf0, 0x38, 0xf9, 0x19, 0x10, 0x88, 
+    0xd6, 0x13, 0x93, 0xc5, 0xb4, 0xc3, 0x58, 0x2e, 0x38, 0xbf, 
+    0xa9, 0xf7, 0x7d, 0xbc, 0x6d, 0xb6, 0x08, 0x3e, 0xd4, 0xab, 
+    0x7b, 0xd1, 0xe0, 0x68, 0x5c, 0xc7, 0xa8, 0x6a, 0xab, 0x9c, 
+    0x86, 0xe9, 0x0a, 0x7d, 0x71, 0x92, 0xdb, 0xd3, 0xa3, 0x8f, 
+    0x0e, 0x5c, 0xad, 0x86, 0xac, 0x5f, 0x9d, 0xd6, 0x75, 0x3b, 
+    0xf3, 0x64, 0x61, 0x27, 0xa5, 0x10, 0x74, 0x3f, 0x30, 0xce, 
+    0xa3, 0x40, 0x22, 0xb4, 0x29, 0xa3, 0x9b, 0x94, 0x10, 0xe5, 
+    0xc0, 0x8f, 0x7f, 0xe3, 0x51, 0x13, 0x93, 0x2f, 0xb2, 0xba, 
+    0x4e, 0xb8, 0x6e, 0xc2, 0x43, 0x25, 0xb6, 0xb1, 0x50, 0xe2, 
+    0xe6, 0x8a, 0xbe, 0x1c, 0xc6, 0x0a, 0x49, 0xcd, 0x87, 0xa4, 
+    0x36, 0x8b, 0xcf, 0x8c, 0xa1, 0xf3, 0x69, 0x3a, 0x73, 0x04, 
+    0x9d, 0xde, 0xf1, 0x8e, 0x7d, 0x62, 0x43, 0xf5, 0x38, 0x4b, 
+    0x81, 0x1f, 0xb7, 0xac, 0x06, 0x29, 0x08, 0xd3, 0x06, 0xd0, 
+    0x67, 0x7f, 0xf4, 0x25, 0x53, 0x54, 0x20, 0x2e, 0xdf, 0x6b, 
+    0x33, 0x03, 0x64, 0x04, 0x36, 0xd0, 0x25, 0x02, 0x2a, 0x6e, 
+    0xa6, 0xfc, 0xd5, 0x36, 0x6b, 0x55, 0x6c, 0x9f, 0x43, 0x34, 
+    0x29, 0xc7, 0xe9, 0xe2, 0xe4, 0x9a, 0x28, 0xbb, 0x40, 0x8d, 
+    0x4d, 0x99, 0x13, 0x84, 0x71, 0x31, 0x6d, 0xaa, 0x99, 0x34, 
+    0xb0, 0x9a, 0x7d, 0x16, 0xe6, 0xdf, 0x83, 0xaf, 0xbd, 0xf8, 
+    0x77, 0x88, 0xb1, 0x4b, 0x18, 0xa8, 0x62, 0xe2, 0xb5, 0xdb, 
+    0x17, 0x31, 0x47, 0xeb, 0xca, 0x9e, 0x9e, 0xe7, 0x19, 0x99, 
+    0x09, 0x00, 0x9a, 0x40, 0x55, 0x38, 0xc4, 0xea, 0xd7, 0xbf, 
+    0xae, 0x00, 0x95, 0x8b, 0x78, 0x01, 0x60, 0x01, 0xf9, 0xa3, 
+    0xac, 0x49, 0xc8, 0xdd, 0xc5, 0x78, 0xc2, 0x69, 0xf2, 0x2c, 
+    0x2b, 0x7a, 0x2f, 0x01, 0x09, 0x13, 0xe9, 0x85, 0x3e, 0xb9, 
+    0x63, 0xd9, 0x48, 0xea, 0xd1, 0x9e, 0x21, 0xda, 0x4b, 0x26, 
+    0xc7, 0xc9, 0xfd, 0xb5, 0x42, 0x37, 0x24, 0x1a, 0x97, 0x51, 
+    0x8c, 0x20, 0x51, 0xfe, 0x0d, 0x14, 0x0c, 0xc8, 0x50, 0x74, 
+    0xb5, 0x1d, 0xba, 0x8c, 0xd5, 0x8a, 0x80, 0xf4, 0x1e, 0x37, 
+    0xf2, 0x6a, 0x3c, 0x31, 0xf3, 0x90, 0xb4, 0x0b, 0xab, 0x7f, 
+    0x44, 0x2c, 0xe1, 0x09, 0x25, 0x42, 0xf9, 0x81, 0xd4, 0x49, 
+    0x4c, 0x5b, 0xd4, 0xc1, 0x14, 0x5e, 0x5e, 0x3c, 0xdc, 0xe8, 
+    0x7d, 0xee, 0x39, 0xe6, 0x10, 0x78, 0x7e, 0xf6, 0xb5, 0xb2, 
+    0xf1, 0x32, 0xac, 0xa2, 0x48, 0xc4, 0xc6, 0xbe, 0x0d, 0xeb, 
+    0x8a, 0x53, 0xf9, 0x0a, 0x6f, 0x50, 0x71, 0x36, 0x10, 0x88, 
+    0xc5, 0x8c, 0xda, 0xb9, 0x99, 0xe3, 0x56, 0xb6, 0xad, 0x59, 
+    0x1e, 0x42, 0x72, 0x97, 0x53, 0x75, 0x2f, 0x4e, 0x08, 0x43, 
+    0x53, 0xb1, 0x02, 0xbc, 0xd7, 0xc9, 0x13, 0x7a, 0x05, 0x84, 
+    0x8c, 0x0a, 0xf3, 0x62, 0x11, 0x5d, 0x0d, 0x56, 0x2a, 0x5a, 
+    0x7e, 0x05, 0x62, 0x57, 0xe5, 0x05, 0xb7, 0x5a, 0xf7, 0x69, 
+    0x39, 0xd7, 0xf2, 0x0d, 0x93, 0x82, 0x0c, 0x9a, 0xa4, 0xc8, 
+    0x77, 0x6c, 0xe3, 0x60, 0x21, 0x6c, 0x1b, 0x5e, 0xe2, 0xf1, 
+    0xb6, 0xbb, 0x69, 0x21, 0xca, 0xf6, 0xf0, 0x5f, 0xb4, 0xdb, 
+    0xeb, 0x20, 0x47, 0xa9, 0x82, 0x11, 0x0d, 0x72, 0xde, 0xd1, 
+    0xa4, 0xda, 0x6c, 0xfa, 0xaa, 0xe9, 0x2b, 0x30, 0xab, 0x6a, 
+    0x83, 0x48, 0xea, 0xfd, 0xf9, 0xaa, 0x76, 0x93, 0xe6, 0xe2, 
+    0x25, 0xb0, 0x6d, 0x4f, 0x11, 0xd4, 0xee, 0x9e, 0x6d, 0x72, 
+    0x20, 0x00, 0x8e, 0x90, 0x62, 0x0a, 0x57, 0x57, 0x08, 0x1f, 
+    0xfd, 0x24, 0x64, 0x69, 0x6e, 0x8e, 0x76, 0x8c, 0xbb, 0x8d, 
+    0x42, 0x8d, 0xd9, 0x94, 0xbf, 0xf2, 0x20, 0x6e, 0xcd, 0xca, 
+    0xdd, 0x0a, 0x4b, 0x5c, 0xb7, 0xa7, 0x36, 0x8f, 0xea, 0x51, 
+    0x2c, 0x3f, 0x59, 0x66, 0xa1, 0x8a, 0x8b, 0xba, 0x4c, 0x6c, 
+    0xc3, 0x63, 0x5b, 0x6d, 0x29, 0x2f, 0xc2, 0xd1, 0x99, 0x34, 
+    0x23, 0xf0, 0x81, 0x5c, 0xf7, 0x6e, 0xea, 0x84, 0x65, 0x9c, 
+    0x7f, 0x88, 0x1e, 0x8c, 0xf9, 0xaf, 0x57, 0x52, 0x58, 0x4a, 
+    0xd9, 0x3b, 0x6a, 0xf1, 0x02, 0xa6, 0xe0, 0xfe, 0x9b, 0xdf, 
+    0x7e, 0xa1, 0x74, 0x96, 0x37, 0x61, 0x95, 0x46, 0x81, 0xb2, 
+    0x37, 0x50, 0x47, 0x57, 0xf3, 0xc7, 0x07, 0xb1, 0x54, 0xe0, 
+    0x39, 0x0c, 0x48, 0xd7, 0x59, 0x17, 0x37, 0x0c, 0x08, 0x1b, 
+    0x84, 0xc7, 0xc1, 0xad, 0xdc, 0x3a, 0xd8, 0x0f, 0x77, 0x71, 
+    0x25, 0x92, 0x23, 0x1f, 0x51, 0xe6, 0x8d, 0x94, 0x8e, 0xc8, 
+    0xbe, 0x31, 0xbf, 0xce, 0x30, 0xc9, 0x3e, 0xe0, 0x65, 0x71, 
+    0xa2, 0x50, 0xb7, 0xab, 0x1a, 0x81, 0x3c, 0x8e, 0x3c, 0xe4, 
+    0xc3, 0xaa, 0x16, 0xe4, 0x0b, 0x41, 0xdd, 0x7f, 0x76, 0x06, 
+    0xad, 0x82, 0x19, 0x6e, 0x37, 0x9d, 0xc4, 0x3c, 0x26, 0x3a, 
+    0xa1, 0x13, 0x93, 0xbc, 0x9b, 0x6a, 0xd1, 0xb1, 0x03, 0x75, 
+    0xbd, 0x3a, 0x9d, 0x6c, 0x0f, 0xe3, 0x78, 0xf6, 0x12, 0x73, 
+    0x29, 0xef, 0x10, 0xdf, 0xfb, 0xfb, 0x58, 0xd6, 0xd8, 0x14, 
+    0xd7, 0xa2, 0xdb, 0x5f, 0x71, 0x49, 0x98, 0x5d, 0x3f, 0xb5, 
+    0xed, 0xc3, 0x1e, 0x77, 0xec, 0x3f, 0x5e, 0x99, 0x15, 0x41, 
+    0xc5, 0xb0, 0x82, 0xe5, 0x1c, 0x81, 0x85, 0x9e, 0xf1, 0x5a, 
+    0x58, 0x80, 0xb2, 0x27, 0x2b, 0x19, 0xb7, 0xf4, 0x29, 0x79, 
+    0x01, 0x41, 0x3e, 0x9c, 0x7a, 0x76, 0x49, 0x53, 0xaa, 0x8c, 
+    0xd1, 0x2f, 0xc5, 0x72, 0xdb, 0x2a, 0x4b, 0x2b, 0x9e, 0x76, 
+    0x20, 0x08, 0xfd, 0x3e, 0xb9, 0x97, 0xda, 0x6c, 0xdb, 0xf1, 
+    0x48, 0x6a, 0x3e, 0xbd, 0x8a, 0x15, 0x22, 0xbc, 0x13, 0x67, 
+    0xe1, 0xd4, 0x1a, 0xa3, 0xc1, 0xdb, 0xa4, 0x54, 0x16, 0x05, 
+    0x79, 0x13, 0x2d, 0x8e, 0x7c, 0x85, 0xc9, 0x4e, 0x4b, 0x00, 
+    0xc7, 0xc5, 0x7f, 0x47, 0xf0, 0xee, 0x3a, 0x36, 0x8f, 0x96, 
+    0xfe, 0x2a, 0x77, 0x48, 0x45, 0x3b, 0x52, 0x8d, 0x6d, 0x21, 
+    0x3f, 0x40, 0x52, 0x22, 0x83, 0x26, 0x06, 0xd3, 0x81, 0x53, 
+    0x18, 0x4f, 0xeb, 0xc8, 0x20, 0x1f, 0xc8, 0xab, 0x7b, 0x40, 
+    0x88, 0x9b, 0x6e, 0x5b, 0x96, 0x26, 0x64, 0xbf, 0x35, 0x33, 
+    0x06, 0x16, 0x7a, 0xf5, 0x43, 0xd1, 0xa7, 0xb9, 0x4c, 0x11, 
+    0x91, 0x6a, 0x52, 0x21, 0xa0, 0x1a, 0x23, 0x08, 0x64, 0x51, 
+    0x43, 0x3a, 0xb9, 0xf2, 0xa6, 0xa2, 0xc8, 0x46, 0xf1, 0xd0, 
+    0x8f, 0x0d, 0xe0, 0xd0, 0x1a, 0xf8, 0xcd, 0xcf, 0xaa, 0xaa, 
+    0x81, 0xa5, 0x69, 0xab, 0x53, 0xd6, 0xfb, 0x63, 0xe0, 0x9e, 
+    0xc6, 0x27, 0xc6, 0xb1, 0xf6, 0x49, 0x96, 0xea, 0x80, 0x19, 
+    0xf9, 0xe4, 0xb6, 0xec, 0x6b, 0x77, 0x14, 0xee, 0x93, 0xa3, 
+    0xa1, 0xd9, 0xac, 0x83, 0x63, 0xa1, 0x31, 0x1d, 0x5e, 0x8b, 
+    0x86, 0x1f, 0xa2, 0xc0, 0x97, 0x3a, 0x4a, 0xeb, 0xc3, 0xbb, 
+    0x1a, 0x30, 0xd3, 0xcd, 0x75, 0x8c, 0x93, 0x66, 0x3b, 0xdb, 
+    0x93, 0xfa, 0xf1, 0x62, 0xeb, 0xd7, 0x81, 0xb5, 0xe4, 0xba, 
+    0xdc, 0x7a, 0x11, 0x4c, 0x79, 0x6c, 0xe5, 0xdd, 0x69, 0x71, 
+    0x77, 0xf0, 0x4f, 0x90, 0x4b, 0x00, 0xa5, 0xd3, 0xe7, 0xba, 
+    0xad, 0x32, 0x5b, 0x34, 0x07, 0x0a, 0x70, 0x1f, 0xe1, 0xd8, 
+    0xb0, 0x9b, 0xc2, 0x35, 0x9c, 0x8c, 0x2b, 0xe2, 0xdd, 0x31, 
+    0x48, 0xcd, 0xf1, 0x8b, 0x18, 0xfa, 0x21, 0x9a, 0x2a, 0x3e, 
+    0xcd, 0xeb, 0x5e, 0x06, 0xfc, 0xf6, 0x9e, 0x41, 0x5f, 0xc8, 
+    0x7b, 0x13, 0xaf, 0x91, 0xac, 0xe6, 0xe9, 0xcd, 0xc1, 0x7c, 
+    0x84, 0x81, 0xaf, 0x75, 0x0a, 0xe9, 0x30, 0x6d, 0xfa, 0x02, 
+    0x2a, 0x1f, 0x13, 0x2e, 0x2d, 0x34, 0x7b, 0x87, 0x6c, 0x5a, 
+    0x43, 0xaa, 0x7d, 0x25, 0xd0, 0xc3, 0xcf, 0x32, 0x8b, 0x3f, 
+    0x28, 0xb5, 0x0b, 0xa8, 0x7d, 0x06, 0x8e, 0x04, 0x39, 0xd0, 
+    0x96, 0x6a, 0xa2, 0x26, 0x64, 0xe1, 0xdc, 0xcb, 0xcc, 0xd5, 
+    0xfe, 0x33, 0xef, 0x9c, 0xba, 0xdb, 0xb8, 0xe1, 0xf2, 0x2a, 
+    0xb8, 0x5c, 0x50, 0xb8, 0x2a, 0xc8, 0x82, 0xc0, 0x1c, 0xc9, 
+    0x9b, 0xca, 0x53, 0x0a, 0x52, 0x3d, 0xbe, 0x23, 0x71, 0xdc, 
+    0x5c, 0xf8, 0x41, 0x6e, 0xb1, 0x00, 0x17, 0x96, 0xec, 0xdc, 
+    0xfb, 0xac, 0x61, 0xb4, 0x62, 0xcd, 0x97, 0x5e, 0xc9, 0x82, 
+    0xf8, 0x00, 0xc7, 0x50, 0x2c, 0xf5, 0xe5, 0x6b, 0xd8, 0xe1, 
+    0x2c, 0x65, 0x50, 0x46, 0xbd, 0x1e, 0x4f, 0xe3, 0x5c, 0x3b, 
+    0x90, 0x22, 0xf4, 0x24, 0x4f, 0x4f, 0x8f, 0x08, 0x3a, 0x8e, 
+    0x02, 0x9c, 0x99, 0xf9, 0xe4, 0x0b, 0x34, 0x37, 0x66, 0x0b, 
+    0x29, 0x36, 0xa6, 0xde, 0x8a, 0x9d, 0x21, 0x2b, 0xab, 0x6a, 
+    0x0f, 0xe3, 0x55, 0x2c, 0xb9, 0xa7, 0x35, 0x68, 0x5e, 0xf4, 
+    0xaf, 0x38, 0x66, 0x1d, 0x2a, 0xcc, 0xe9, 0x26, 0x7a, 0x72, 
+    0xb4, 0x16, 0xde, 0x98, 0x4a, 0xe0, 0x01, 0xa4, 0x36, 0x11, 
+    0x00, 0x26, 0xc6, 0x08, 0xeb, 0x7e, 0xfd, 0x2f, 0x1a, 0xe7, 
+    0x8d, 0x17, 0xe2, 0xb4, 0xf3, 0x41, 0x53, 0xc5, 0x56, 0x64, 
+    0x92, 0x1f, 0xcf, 0x7f, 0xf2, 0xf6, 0x2c, 0x06, 0x10, 0x5e, 
+    0x21, 0xd1, 0xc6, 0x22, 0x66, 0x66, 0x5b, 0x70, 0x97, 0xbc, 
+    0xbf, 0x9f, 0x16, 0x92, 0x95, 0x66, 0xfb, 0xe3, 0x8b, 0xf6, 
+    0x36, 0xb7, 0x53, 0x8a, 0xd3, 0xbe, 0xa3, 0x3b, 0xa4, 0xc3, 
+    0x10, 0xc0, 0x7c, 0x28, 0xea, 0x7c, 0xa7, 0x35, 0x79, 0x0a, 
+    0x67, 0xe4, 0x47, 0xa0, 0x13, 0xea, 0xe8, 0x2e, 0x36, 0xbe, 
+    0xae, 0x83, 0x5f, 0xdc, 0xad, 0x35, 0x7b, 0xcd, 0x6d, 0x44, 
+    0x35, 0x40, 0xd7, 0xbf, 0x3a, 0xd0, 0xb9, 0x3d, 0xdc, 0xed, 
+    0x14, 0x4b, 0xcd, 0x34, 0x2c, 0x06, 0xd9, 0x61, 0x18, 0x80, 
+    0xe1, 0x06, 0x41, 0xb1, 0x9e, 0x46, 0x44, 0xa5, 0xb0, 0xaf, 
+    0xc9, 0x26, 0xa0, 0x7a, 0xca, 0xf5, 0x8a, 0xee, 0x37, 0xa6, 
+    0xc2, 0xd2, 0x04, 0x12, 0xc7, 0x1e, 0x0c, 0x69, 0x35, 0x61, 
+    0xe3, 0x61, 0x73, 0xdf, 0xad, 0xe8, 0x3d, 0x3b, 0xaa, 0xc2, 
+    0x19, 0x77, 0x84, 0xb8, 0x3d, 0x6d, 0x2b, 0xb5, 0xcf, 0xd9, 
+    0x8f, 0x75, 0x33, 0xc3, 0x04, 0xb5, 0xa4, 0xf8, 0xcc, 0xea, 
+    0x1b, 0x41, 0x11, 0x18, 0x09, 0x9d, 0xf4, 0x64, 0x8b, 0x59, 
+    0xee, 0x9e, 0xf8, 0x72, 0xdf, 0x97, 0x1d, 0xca, 0x4e, 0x55, 
+    0xca, 0x2b, 0xe0, 0x95, 0x9c, 0x0e, 0x91, 0xf5, 0xbf, 0x76, 
+    0x58, 0x57, 0xbc, 0xa0, 0xc0, 0xd1, 0x01, 0xef, 0x8f, 0x3b, 
+    0x21, 0x0d, 0x18, 0x4d, 0x1f, 0x98, 0x59, 0x2b, 0x94, 0xc4, 
+    0x0b, 0xb6, 0x26, 0x93, 0xcc, 0xe5, 0x43, 0x5c, 0xf5, 0xb5, 
+    0x48, 0x4d, 0x1e, 0xdd, 0x6c, 0x3f, 0x65, 0x5c, 0xb1, 0x00, 
+    0x62, 0x79, 0xf2, 0xf9, 0xaa, 0xb5, 0xbc, 0x36, 0x17, 0x0d, 
+    0x80, 0xc3, 0xcc, 0xc3, 0x13, 0x17, 0x5b, 0x90, 0xbc, 0xe8, 
+    0x6b, 0xbd, 0x9e, 0x8d, 0x15, 0x23, 0xd8, 0x2a, 0x54, 0x37, 
+    0x2d, 0x04, 0x3c, 0xeb, 0xe4, 0x22, 0x7c, 0x2f, 0x14, 0x8e, 
+    0xca, 0xbc, 0xbd, 0x8a, 0xb2, 0xd3, 0x85, 0xb4, 0xd3, 0x88, 
+    0x96, 0x9e, 0x45, 0xdb, 0xd0, 0xd9, 0xf4, 0x8e, 0x8a, 0x80, 
+    0x7b, 0xb0, 0xfb, 0x2a, 0xed, 0x9d, 0xb5, 0x96, 0xac, 0x96, 
+    0xcf, 0x0d, 0xfb, 0x82, 0x38, 0x41, 0xbb, 0x48, 0x38, 0x59, 
+    0x03, 0xb8, 0x41, 0xac, 0xbb, 0x91, 0x82, 0x9c, 0x2c, 0x17, 
+    0x8b, 0xe2, 0x71, 0x51, 0xdd, 0xe1, 0x14, 0x36, 0x7a, 0xc7, 
+    0xb0, 0x19, 0x5c, 0x75, 0xca, 0x6a, 0xf4, 0x65, 0x9e, 0xab, 
+    0xfb, 0x9e, 0xaf, 0x7b, 0x87, 0x96, 0xf5, 0xbf, 0x42, 0xf9, 
+    0x26, 0x69, 0xfd, 0xfb, 0x52, 0x11, 0x81, 0xb0, 0x52, 0x4a, 
+    0x24, 0xa8, 0x2d, 0xa0, 0x31, 0x2b, 0xcf, 0x7a, 0x0b, 0xd2, 
+    0x5b, 0x21, 0xfb, 0x20, 0x28, 0xac, 0x68, 0xd1, 0x11, 0x4e, 
+    0xe0, 0x35, 0x3c, 0xad, 0x51, 0x56, 0xdc, 0x5c, 0x8d, 0x8e, 
+    0x13, 0x87, 0xbc, 0x96, 0x3c, 0x03, 0x5f, 0xb2, 0x93, 0xbb, 
+    0x8c, 0x37, 0x9d, 0xb2, 0x0b, 0x87, 0x26, 0x07, 0x32, 0x43, 
+    0x73, 0xcb, 0xfc, 0x5c, 0xa6, 0x5d, 0x58, 0xa5, 0x3d, 0xe8, 
+    0x03, 0xfc, 0x6a, 0xdf, 0xe0, 0x67, 0x85, 0x05, 0xd1, 0x44, 
+    0x45, 0xbf, 0x2d, 0xd5, 0xa6, 0xd1, 0x7d, 0xa7, 0x5a, 0x82, 
+    0x44, 0x07, 0x02, 0x70, 0x07, 0x08, 0x86, 0x37, 0x23, 0x98, 
+    0x50, 0x54, 0x95, 0x99, 0x68, 0xa8, 0x3d, 0xb6, 0x23, 0x5c, 
+    0x7c, 0x38, 0x74, 0x77, 0x6b, 0xc4, 0xcf, 0xda, 0x6b, 0x7a, 
+    0x19, 0x20, 0xe1, 0x29, 0xdb, 0xf6, 0xe2, 0x3e, 0x43, 0x56, 
+    0x87, 0xeb, 0x12, 0x83, 0xb6, 0x0a, 0xcc, 0x20, 0x68, 0xdf, 
+    0x5b, 0xf8, 0xe0, 0x35, 0x65, 0x2e, 0xb8, 0xce, 0xa2, 0xa0, 
+    0x7d, 0xa0, 0x58, 0x36, 0xa6, 0x97, 0x4d, 0x3f, 0x55, 0x3f, 
+    0x30, 0x6f, 0x66, 0x06, 0x2f, 0x22, 0xd3, 0x6c, 0xab, 0x52, 
+    0x81, 0x51, 0x88, 0xdb, 0x9c, 0x3c, 0x2d, 0xce, 0xe0, 0x67, 
+    0x98, 0x00, 0x6e, 0xbb, 0xb7, 0xc6, 0x82, 0xb2, 0x7b, 0x41, 
+    0x41, 0xe6, 0x62, 0x99, 0x84, 0xe5, 0x58, 0xc5, 0xc9, 0xf0, 
+    0x5b, 0xc6, 0x2b, 0x2b, 0x75, 0x1d, 0xfe, 0x5b, 0x86, 0x9a, 
+    0x09, 0xdd, 0x60, 0x39, 0x50, 0xa0, 0xdb, 0xc5, 0x32, 0x7e, 
+    0x54, 0xc5, 0xad, 0xde, 0x65, 0x37, 0x1e, 0xf2, 0x71, 0x05, 
+    0x9a, 0x24, 0x24, 0xbb, 0x3f, 0x94, 0x6c, 0x7e, 0x46, 0x54, 
+    0x05, 0x8f, 0x49, 0x48, 0x29, 0x5e, 0x1a, 0x2b, 0x28, 0xca, 
+    0xbf, 0xb2, 0x56, 0x97, 0x01, 0x18, 0x7a, 0xf1, 0x22, 0xb6, 
+    0x72, 0x88, 0x76, 0xb2, 0x72, 0x70, 0x47, 0x6e, 0x74, 0x19, 
+    0x36, 0x76, 0x50, 0x99, 0x7e, 0x8e, 0x61, 0x54, 0x2c, 0x55, 
+    0xef, 0xf7, 0x27, 0x62, 0x2f, 0x7a, 0xa6, 0x0b, 0x33, 0xb3, 
+    0x21, 0xec, 0xde, 0xaf, 0x1e, 0x5d, 0x30, 0x42, 0x82, 0xd9, 
+    0x59, 0x5e, 0xc9, 0xcf, 0x5a, 0x5b, 0xf5, 0xf6, 0x77, 0xdf, 
+    0xc1, 0x55, 0x6d, 0x5a, 0x6f, 0x54, 0x9a, 0x1f, 0x2d, 0xc2, 
+    0x19, 0xe1, 0x54, 0x91, 0xe3, 0x1d, 0x01, 0xae, 0xcb, 0xd0, 
+    0x73, 0x99, 0x0f, 0x6a, 0x7d, 0x3a, 0x88, 0x81, 0x90, 0xf6, 
+    0x15, 0xb9, 0x80, 0xf8, 0x2f, 0x4b, 0x39, 0x85, 0x17, 0xc5, 
+    0xf9, 0x76, 0x1d, 0x35, 0x2e, 0xbf, 0x68, 0x12, 0x6a, 0xb2, 
+    0xee, 0xc2, 0x4f, 0xa0, 0x1b, 0xa8, 0xad, 0xb9, 0x58, 0x3f, 
+    0x3b, 0xbb, 0x38, 0x21, 0x29, 0xa4, 0xf5, 0x53, 0xc7, 0x10, 
+    0xb7, 0x64, 0x52, 0xa4, 0x8b, 0xda, 0x22, 0x0d, 0xc5, 0xbc, 
+    0xb0, 0xb5, 0xea, 0x3c, 0x15, 0x6e, 0x18, 0x3a, 0x03, 0x92, 
+    0x49, 0xd5, 0x07, 0xdc, 0x30, 0x93, 0xb8, 0x8c, 0x3b, 0x38, 
+    0xc4, 0x96, 0xef, 0x7f, 0xb6, 0x6f, 0x0a, 0xe4, 0x8e, 0x4c, 
+    0xbb, 0x15, 0x87, 0x2a, 0xa5, 0x97, 0xfe, 0xb1, 0xdf, 0x65, 
+    0xb4, 0xed, 0x9b, 0x1d, 0x7e, 0x6c, 0xa4, 0xa5, 0x20, 0x8f, 
+    0x09, 0xd6, 0x14, 0x33, 0xad, 0xcd, 0x25, 0xe8, 0x3b, 0x74, 
+    0x8e, 0x3b, 0x67, 0x07, 0x0f, 0x2b, 0x19, 0x35, 0x10, 0x98, 
+    0xab, 0x33, 0xaa, 0x8f, 0xa6, 0xf3, 0x9a, 0xc2, 0x89, 0x94, 
+    0xfd, 0x72, 0xb2, 0x94, 0x64, 0x77, 0x74, 0x84, 0xf7, 0x0a, 
+    0x47, 0xc9, 0x14, 0x5c, 0xe0, 0x19, 0xc6, 0xa8, 0xf0, 0xb9, 
+    0x13, 0x98, 0x11, 0xae, 0x76, 0x85, 0xc8, 0x38, 0xda, 0xe8, 
+    0x55, 0xb8, 0x1f, 0xe1, 0xf0, 0xfe, 0x77, 0x1c, 0x53, 0xb3, 
+    0xe9, 0x57, 0x1c, 0x2b, 0x08, 0x89, 0xe1, 0x88, 0x13, 0x24, 
+    0x57, 0x83, 0x9c, 0x85, 0xd1, 0x14, 0xf8, 0xf9, 0xdd, 0x29, 
+    0x30, 0x73, 0xae, 0xcd, 0xd0, 0x2a, 0xec, 0xa7, 0x44, 0x75, 
+    0x8a, 0x38, 0x3e, 0xd3, 0x5b, 0x24, 0x2b, 0xe0, 0x5a, 0x85, 
+    0x39, 0x8c, 0x0c, 0x51, 0x9b
+};
+
diff --git a/neureka/idma_pointwise/src/output.c b/neureka/idma_pointwise/src/output.c
new file mode 100644
index 0000000..0172aeb
--- /dev/null
+++ b/neureka/idma_pointwise/src/output.c
@@ -0,0 +1,165 @@
+#include "output.h"
+
+#define OUTPUT_SIZE (1365)
+PI_L1 uint8_t output[OUTPUT_SIZE];
+
+#define GOLDEN_OUTPUT_SIZE (1365)
+PI_L2 uint8_t golden_output[GOLDEN_OUTPUT_SIZE] = {
+    0xcd, 0x00, 0x00, 0x00, 0x63, 0x15, 0x00, 0x00, 0x00, 0x00, 
+    0x36, 0x6f, 0x00, 0x00, 0x27, 0x6d, 0x00, 0x2a, 0xff, 0x31, 
+    0xff, 0x8f, 0x00, 0x00, 0x01, 0x64, 0x55, 0x16, 0xb8, 0x0a, 
+    0x00, 0x00, 0x57, 0x00, 0x00, 0x00, 0x8c, 0x00, 0x00, 0x77, 
+    0x00, 0x1b, 0x00, 0x9b, 0x00, 0x00, 0x0d, 0x00, 0x00, 0x30, 
+    0x60, 0x9b, 0x37, 0x1b, 0xff, 0x00, 0x00, 0xff, 0x00, 0xff, 
+    0xbf, 0x00, 0x00, 0x00, 0xff, 0x3d, 0x22, 0x00, 0x00, 0x00, 
+    0x00, 0x4e, 0x00, 0x00, 0x45, 0xe0, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x93, 0x00, 0x00, 0x00, 0x00, 0x00, 0x3b, 0xff, 
+    0x00, 0x28, 0x00, 0xff, 0x00, 0x00, 0xff, 0x77, 0xf8, 0x4b, 
+    0x00, 0x00, 0x00, 0x8d, 0x37, 0x1c, 0x65, 0x00, 0x05, 0x00, 
+    0x8d, 0x00, 0x01, 0x50, 0xce, 0x00, 0x00, 0x8c, 0x00, 0x93, 
+    0x00, 0x17, 0x00, 0x00, 0x00, 0x00, 0x00, 0x3b, 0x2f, 0x6e, 
+    0x21, 0x00, 0xff, 0x00, 0x00, 0xff, 0x00, 0x4b, 0xcd, 0x00, 
+    0x00, 0x00, 0xff, 0x44, 0x0c, 0x5d, 0x00, 0x2f, 0x00, 0x7c, 
+    0x00, 0x00, 0x1d, 0xf9, 0x00, 0x00, 0x5b, 0x00, 0xd8, 0x00, 
+    0xf6, 0xdc, 0x00, 0x00, 0x00, 0x00, 0x33, 0x34, 0x00, 0x21, 
+    0x19, 0xe8, 0x95, 0x4e, 0xff, 0x00, 0xff, 0x9e, 0x00, 0x00, 
+    0x00, 0x9b, 0x41, 0x23, 0x47, 0x2c, 0x00, 0x00, 0x7d, 0x00, 
+    0x00, 0x00, 0x49, 0x00, 0x00, 0xee, 0x00, 0xee, 0x00, 0x86, 
+    0x71, 0x00, 0x00, 0x00, 0x00, 0x2e, 0x2d, 0xe3, 0x7a, 0x53, 
+    0xdc, 0x74, 0x14, 0xff, 0x66, 0xff, 0xcc, 0x13, 0x00, 0x00, 
+    0xce, 0x2e, 0x15, 0x93, 0x09, 0x24, 0x00, 0x70, 0x00, 0x00, 
+    0x00, 0xe9, 0x0e, 0x00, 0x3a, 0x00, 0x25, 0x00, 0x02, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x34, 0x93, 0xea, 0xc3, 0x59, 0x8a, 
+    0x00, 0xbb, 0xda, 0x3e, 0xff, 0xc3, 0x00, 0x00, 0x00, 0xff, 
+    0x4a, 0x12, 0x00, 0x34, 0x00, 0x00, 0x70, 0x00, 0x00, 0x00, 
+    0xeb, 0x00, 0x00, 0x00, 0x00, 0x1c, 0x08, 0x86, 0xb1, 0x14, 
+    0x14, 0x00, 0x00, 0x30, 0x5a, 0xad, 0x48, 0x00, 0xff, 0x00, 
+    0x00, 0xff, 0x00, 0xff, 0xda, 0x00, 0x00, 0x00, 0x62, 0x4b, 
+    0x22, 0x49, 0x00, 0x00, 0x00, 0x8a, 0x00, 0x00, 0x24, 0x64, 
+    0x00, 0x00, 0xff, 0x00, 0x9d, 0x00, 0x4b, 0xc1, 0x00, 0x00, 
+    0x00, 0x00, 0x2e, 0x00, 0x11, 0x2b, 0x8f, 0xff, 0x63, 0x00, 
+    0xd4, 0x12, 0xff, 0xcc, 0x00, 0x00, 0x00, 0x93, 0x5f, 0x1b, 
+    0x0c, 0x00, 0x43, 0x00, 0x6f, 0x00, 0x00, 0x00, 0xff, 0x00, 
+    0x00, 0x55, 0x00, 0x54, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x35, 0xff, 0x00, 0x00, 0x15, 0xf9, 0x00, 0x00, 0xe8, 
+    0x00, 0xff, 0xff, 0x00, 0x00, 0x02, 0x4f, 0x4a, 0x18, 0xc4, 
+    0x22, 0x09, 0x00, 0x6b, 0x00, 0x00, 0x08, 0xff, 0x00, 0x00, 
+    0x8c, 0x00, 0x00, 0x00, 0x00, 0x00, 0x74, 0x00, 0x00, 0x00, 
+    0x2f, 0xdf, 0x00, 0x00, 0x97, 0xff, 0x00, 0x00, 0xba, 0x00, 
+    0xff, 0xa5, 0x00, 0x00, 0x00, 0x9c, 0x13, 0x12, 0xbb, 0x00, 
+    0x00, 0x00, 0x4d, 0x00, 0x00, 0x00, 0xff, 0x00, 0x00, 0x11, 
+    0x00, 0x00, 0x00, 0x8a, 0x07, 0x00, 0x00, 0x00, 0x00, 0x36, 
+    0xd2, 0x8a, 0x6b, 0x00, 0xa2, 0x08, 0x33, 0xf1, 0x0d, 0x2a, 
+    0xc6, 0x00, 0x00, 0x00, 0x36, 0x54, 0x1a, 0x9e, 0x00, 0x10, 
+    0x00, 0x59, 0x00, 0x00, 0x00, 0xa0, 0x00, 0x00, 0x00, 0x00, 
+    0x4b, 0x00, 0x40, 0x00, 0x00, 0x00, 0x00, 0x00, 0x2e, 0xaf, 
+    0x56, 0x36, 0x18, 0xb5, 0x00, 0x00, 0x87, 0x00, 0x09, 0xaf, 
+    0x00, 0x00, 0x00, 0xec, 0x3b, 0x22, 0x64, 0x00, 0x47, 0x00, 
+    0xb7, 0x00, 0x00, 0x00, 0x9e, 0x15, 0x00, 0x3b, 0x00, 0x00, 
+    0x00, 0x63, 0xa5, 0x00, 0x18, 0x00, 0x00, 0x34, 0x97, 0x00, 
+    0x9e, 0x23, 0xa2, 0x64, 0x57, 0xff, 0x00, 0xf2, 0xef, 0x00, 
+    0x00, 0x00, 0xf2, 0x4f, 0x1b, 0x5e, 0x06, 0x00, 0x00, 0x8c, 
+    0x00, 0x00, 0x0a, 0xd6, 0x00, 0x00, 0x38, 0x00, 0x61, 0x00, 
+    0x90, 0xbf, 0x00, 0x00, 0x00, 0x00, 0x2d, 0x00, 0xc8, 0x80, 
+    0x00, 0xa9, 0x01, 0x01, 0xfb, 0x4b, 0xff, 0xcb, 0x20, 0x00, 
+    0x00, 0xff, 0x4e, 0x13, 0x94, 0x00, 0x00, 0x00, 0x54, 0x00, 
+    0x00, 0x00, 0xff, 0x00, 0x00, 0xe9, 0x00, 0x14, 0x31, 0x0c, 
+    0xd1, 0x00, 0x00, 0x00, 0x00, 0x35, 0x97, 0x46, 0x00, 0x0d, 
+    0xff, 0x00, 0x00, 0xff, 0x00, 0xb7, 0xbf, 0x00, 0x00, 0x00, 
+    0xa0, 0x3d, 0x11, 0x9f, 0x00, 0x51, 0x00, 0x7f, 0x00, 0x00, 
+    0x00, 0x96, 0x00, 0x00, 0x82, 0x00, 0x6a, 0x00, 0x51, 0xae, 
+    0x00, 0x00, 0x00, 0x00, 0x31, 0xff, 0x00, 0xae, 0x00, 0xf2, 
+    0x79, 0x00, 0xd7, 0x00, 0xff, 0xef, 0x05, 0x00, 0x00, 0xf0, 
+    0x44, 0x20, 0xff, 0xed, 0x4d, 0x00, 0x63, 0x00, 0x00, 0x81, 
+    0xff, 0x00, 0x00, 0xda, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x35, 0x95, 0xff, 0x9d, 0x00, 0xff, 0x00, 
+    0x00, 0xff, 0x62, 0xff, 0xbb, 0x00, 0x00, 0x00, 0x9e, 0x3c, 
+    0x1e, 0x97, 0x1f, 0x52, 0x00, 0x53, 0x00, 0x00, 0x43, 0xff, 
+    0x00, 0x00, 0xa6, 0x00, 0x72, 0x00, 0xfb, 0x4b, 0x01, 0x00, 
+    0x00, 0x00, 0x31, 0x86, 0x00, 0x36, 0x00, 0xf9, 0x00, 0xa3, 
+    0xff, 0xed, 0xff, 0xb5, 0x00, 0x00, 0x00, 0x6a, 0x3c, 0x19, 
+    0x8c, 0x63, 0x00, 0x00, 0x57, 0x00, 0x00, 0x00, 0xff, 0x00, 
+    0x00, 0xff, 0x00, 0x00, 0x00, 0x00, 0xc9, 0x00, 0x00, 0x00, 
+    0x00, 0x34, 0x9e, 0x11, 0x2b, 0x00, 0x36, 0x00, 0x00, 0xd6, 
+    0x30, 0xff, 0x96, 0x00, 0x00, 0x00, 0x4b, 0x27, 0x1d, 0x31, 
+    0x00, 0x25, 0x00, 0x53, 0x00, 0x00, 0x00, 0xad, 0x00, 0x6c, 
+    0x70, 0x00, 0x0a, 0x00, 0x62, 0x28, 0x00, 0x00, 0x00, 0x00, 
+    0x32, 0xba, 0x49, 0x3e, 0x13, 0xeb, 0x27, 0x14, 0xff, 0x11, 
+    0x09, 0xdd, 0x00, 0x00, 0x00, 0xff, 0x3e, 0x1b, 0x92, 0x00, 
+    0x00, 0x00, 0x9b, 0x00, 0x00, 0x00, 0xf2, 0x00, 0x00, 0xf5, 
+    0x00, 0x16, 0x00, 0x00, 0xb4, 0x00, 0x00, 0x00, 0x00, 0x33, 
+    0x45, 0x01, 0x37, 0x96, 0xff, 0x00, 0xa8, 0xff, 0x00, 0xff, 
+    0xd5, 0x00, 0x00, 0x00, 0xb0, 0x2e, 0x19, 0xbe, 0x00, 0x00, 
+    0x00, 0x80, 0x00, 0x00, 0x3e, 0xf9, 0x00, 0x00, 0x69, 0x00, 
+    0x3d, 0x00, 0x7e, 0x00, 0x00, 0x00, 0x00, 0x00, 0x36, 0xd3, 
+    0x0e, 0x2a, 0x3c, 0xc3, 0x00, 0x32, 0xff, 0x71, 0xff, 0x77, 
+    0x00, 0x00, 0x00, 0xff, 0x56, 0x23, 0x9c, 0x00, 0x52, 0x00, 
+    0x7d, 0x00, 0x16, 0x5d, 0xa2, 0x00, 0x00, 0x77, 0x00, 0x16, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x33, 0x46, 0x00, 
+    0x2c, 0x17, 0xdc, 0x0a, 0x00, 0xff, 0x00, 0xff, 0xba, 0x00, 
+    0x00, 0x00, 0x9e, 0x33, 0x27, 0x5f, 0x3f, 0x00, 0x00, 0x75, 
+    0x00, 0x00, 0xbd, 0x54, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x36, 0xe6, 0x00, 0x00, 
+    0x00, 0xc2, 0xbb, 0x00, 0xff, 0x2a, 0xff, 0xff, 0x00, 0x00, 
+    0x00, 0x47, 0x51, 0x15, 0x71, 0x00, 0x00, 0x00, 0x8a, 0x00, 
+    0x00, 0x00, 0xff, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0xff, 
+    0x6b, 0x00, 0x00, 0x00, 0x00, 0x30, 0x3e, 0x5d, 0x8b, 0x32, 
+    0xde, 0x00, 0x00, 0xff, 0x49, 0xff, 0x92, 0x00, 0x00, 0x00, 
+    0x08, 0x4f, 0x12, 0xb3, 0x00, 0x71, 0x00, 0x83, 0x00, 0x0c, 
+    0x00, 0xe1, 0x00, 0x00, 0x00, 0x00, 0x3f, 0x00, 0x00, 0x57, 
+    0x00, 0x18, 0x00, 0x00, 0x2f, 0xd2, 0x00, 0x00, 0x55, 0xd9, 
+    0x5c, 0x26, 0xff, 0x00, 0xff, 0xc0, 0x00, 0x00, 0x00, 0xe2, 
+    0x3c, 0x23, 0x18, 0x0e, 0x28, 0x00, 0x66, 0x00, 0x00, 0x00, 
+    0xff, 0x00, 0x00, 0x26, 0x00, 0xb4, 0x00, 0x5f, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x35, 0xc6, 0x84, 0x01, 0x53, 0xff, 0x00, 
+    0x00, 0xff, 0x00, 0xfe, 0x7f, 0x10, 0x00, 0x00, 0xed, 0x2d, 
+    0x1d, 0xb2, 0x55, 0x00, 0x00, 0x5e, 0x00, 0x26, 0xd2, 0xff, 
+    0x00, 0x00, 0x03, 0x00, 0x5b, 0x00, 0x2e, 0x20, 0x00, 0x00, 
+    0x00, 0x00, 0x34, 0xff, 0x9a, 0x3e, 0x00, 0xff, 0x00, 0x00, 
+    0xff, 0x00, 0xff, 0xff, 0x00, 0x00, 0x00, 0x26, 0x55, 0x08, 
+    0x6a, 0x00, 0x1d, 0x00, 0x83, 0x00, 0x00, 0x7a, 0xff, 0x00, 
+    0x00, 0x28, 0x00, 0xe7, 0x00, 0x08, 0x0b, 0x0b, 0x00, 0x00, 
+    0x00, 0x30, 0xab, 0x00, 0x00, 0x00, 0x64, 0x0c, 0x1e, 0xe4, 
+    0x27, 0xff, 0x93, 0x05, 0x00, 0x00, 0xff, 0x00, 0x2b, 0x8a, 
+    0x65, 0x0a, 0x00, 0x5f, 0x00, 0x00, 0x30, 0xff, 0x00, 0x00, 
+    0x45, 0x00, 0x7d, 0x00, 0x19, 0x6c, 0x00, 0x00, 0x00, 0x00, 
+    0x3b, 0xf9, 0x00, 0x00, 0x06, 0xa7, 0x00, 0x6f, 0xfd, 0x40, 
+    0xff, 0xb7, 0x00, 0x00, 0x00, 0xff, 0x22, 0x1e, 0xb5, 0x00, 
+    0x15, 0x00, 0x92, 0x00, 0x00, 0x54, 0xff, 0x00, 0x00, 0x66, 
+    0x00, 0x00, 0x00, 0x62, 0x00, 0x00, 0x00, 0x00, 0x00, 0x33, 
+    0x5b, 0x82, 0x26, 0x60, 0xff, 0x00, 0x00, 0xff, 0x6c, 0xff, 
+    0xfe, 0x00, 0x00, 0x00, 0x89, 0x63, 0x14, 0x69, 0xe5, 0x00, 
+    0x00, 0x80, 0x00, 0x00, 0x2b, 0xcf, 0x00, 0x00, 0x00, 0x00, 
+    0x3d, 0x00, 0x84, 0x3d, 0x00, 0x00, 0x00, 0x00, 0x37, 0xaf, 
+    0x00, 0x00, 0x3e, 0xff, 0x00, 0x42, 0xff, 0x00, 0xff, 0xff, 
+    0x00, 0x00, 0x04, 0x09, 0x3f, 0x1e, 0xaf, 0x17, 0x00, 0x00, 
+    0x5d, 0x00, 0x00, 0x22, 0xff, 0x00, 0x1c, 0x56, 0x00, 0x00, 
+    0x00, 0xd5, 0x19, 0x00, 0x00, 0x00, 0x00, 0x35, 0xc9, 0xff, 
+    0x00, 0x00, 0x88, 0x00, 0x0b, 0xff, 0x23, 0xff, 0xff, 0x00, 
+    0x00, 0x00, 0x0b, 0x4e, 0x17, 0x3c, 0x00, 0x00, 0x00, 0x61, 
+    0x00, 0x00, 0x00, 0x52, 0x00, 0x00, 0xb8, 0x00, 0xff, 0x00, 
+    0x45, 0x07, 0x00, 0x00, 0x00, 0x00, 0x2d, 0xe5, 0x00, 0x47, 
+    0x01, 0xff, 0x00, 0x00, 0xf7, 0x27, 0xff, 0x44, 0x00, 0x00, 
+    0x00, 0x4e, 0x5e, 0x17, 0xb2, 0x00, 0x00, 0x00, 0x5a, 0x00, 
+    0x00, 0x00, 0xff, 0x00, 0x36
+};
+
+int check_output() {
+        printf("Checking the output vector:\n");
+
+        int n_err = 0;
+        for (int i = 0; i < OUTPUT_SIZE; i++) {
+            if (output[i] != golden_output[i]) {
+                printf("ERROR: wrong value of output @ %d: %d vs. golden: %d\n", i, output[i], golden_output[i]);
+                n_err++;
+            }
+        }
+
+        if (n_err == 0)
+            printf("> Success! No errors found.\n");
+        else
+            printf("> Failure! Found %d/%d errors.\n", n_err, OUTPUT_SIZE);
+        return n_err;
+    }
+
+    
\ No newline at end of file
diff --git a/neureka/idma_pointwise/src/scale.c b/neureka/idma_pointwise/src/scale.c
new file mode 100644
index 0000000..5eb2c42
--- /dev/null
+++ b/neureka/idma_pointwise/src/scale.c
@@ -0,0 +1,10 @@
+#include "scale.h"
+
+#define SCALE_SIZE (39)
+PI_L1 uint8_t scale[SCALE_SIZE] = {
+    0x1e, 0x02, 0x1a, 0x1d, 0x15, 0x1b, 0x13, 0x07, 0x15, 0x0e, 
+    0x01, 0x11, 0x1b, 0x11, 0x0e, 0x11, 0x19, 0x13, 0x0d, 0x13, 
+    0x1e, 0x0f, 0x0b, 0x0b, 0x05, 0x1b, 0x04, 0x02, 0x0c, 0x14, 
+    0x0c, 0x04, 0x05, 0x0f, 0x07, 0x0f, 0x0f, 0x16, 0x1c
+};
+
diff --git a/neureka/idma_pointwise/src/weight.c b/neureka/idma_pointwise/src/weight.c
new file mode 100644
index 0000000..140952b
--- /dev/null
+++ b/neureka/idma_pointwise/src/weight.c
@@ -0,0 +1,1130 @@
+#include "weight.h"
+
+#define WEIGHT_SIZE (11232)
+PI_L1 uint8_t weight[WEIGHT_SIZE] = {
+    0x9f, 0x3b, 0x8a, 0x72, 0xf7, 0x43, 0x00, 0x69, 0xa1, 0xc9, 
+    0x09, 0x17, 0xa7, 0xab, 0xc5, 0xf7, 0xa7, 0xfa, 0x97, 0xa8, 
+    0xa5, 0x92, 0x11, 0xdf, 0xa7, 0x92, 0x91, 0xff, 0x58, 0x6d, 
+    0x6e, 0x00, 0x18, 0xb6, 0x1f, 0x16, 0x49, 0x91, 0x15, 0x60, 
+    0x06, 0x38, 0x2d, 0x55, 0x2d, 0xa3, 0x6e, 0x7e, 0xd2, 0x59, 
+    0x80, 0x79, 0x8f, 0x11, 0x4a, 0x7e, 0x8f, 0x11, 0x4a, 0x7e, 
+    0x70, 0xee, 0xb5, 0x81, 0x68, 0x39, 0xc7, 0xb5, 0x59, 0x5c, 
+    0xcd, 0x89, 0xfa, 0x75, 0x35, 0xdd, 0x0d, 0x89, 0x5c, 0xcd, 
+    0x9a, 0xee, 0x05, 0xba, 0x3d, 0xbc, 0x05, 0xca, 0x1d, 0xbc, 
+    0x05, 0xca, 0xe2, 0x43, 0xfa, 0x35, 0xee, 0xf7, 0xfa, 0x2d, 
+    0x88, 0xfb, 0x79, 0x5b, 0x2a, 0xc9, 0x15, 0x24, 0xe4, 0xbe, 
+    0x84, 0x12, 0x1d, 0x2d, 0x28, 0x42, 0xc6, 0xac, 0x00, 0xb2, 
+    0xc6, 0xac, 0x00, 0x12, 0x39, 0x53, 0xff, 0xed, 0xe0, 0x32, 
+    0xc6, 0x63, 0x0c, 0xea, 0x91, 0xf4, 0x00, 0x85, 0x05, 0x7f, 
+    0x6c, 0x88, 0x55, 0x3e, 0xfd, 0x98, 0x18, 0x1d, 0x6d, 0x9e, 
+    0x3b, 0x25, 0x6d, 0x9e, 0x3b, 0x25, 0x92, 0x61, 0xc4, 0xda, 
+    0x1e, 0x15, 0x6a, 0x82, 0x3d, 0x68, 0xeb, 0x2f, 0x6e, 0x76, 
+    0x1d, 0x53, 0xbb, 0x91, 0x91, 0xa7, 0x5c, 0x4a, 0xfa, 0x3a, 
+    0xc9, 0x01, 0x2a, 0x1b, 0xc9, 0x01, 0x6a, 0x1b, 0x36, 0xfe, 
+    0x95, 0xe4, 0x13, 0x69, 0xdd, 0xf5, 0xdc, 0xed, 0xb0, 0x39, 
+    0xdc, 0xda, 0xf6, 0x76, 0xaf, 0x02, 0xaa, 0xc4, 0x42, 0x3a, 
+    0xe9, 0x4f, 0x4a, 0x1b, 0xff, 0x4f, 0xca, 0x1a, 0xfb, 0x4f, 
+    0x35, 0xe5, 0x04, 0xb0, 0x2f, 0x58, 0xab, 0xfa, 0x4c, 0x98, 
+    0x7f, 0x23, 0x21, 0xa9, 0xfd, 0xce, 0x34, 0xf0, 0x3a, 0x67, 
+    0x5b, 0x89, 0x6f, 0x3b, 0x01, 0x88, 0x26, 0x3b, 0x01, 0x88, 
+    0x2e, 0xbb, 0xfe, 0x77, 0xd1, 0x44, 0x00, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0xf7, 0xd1, 
+    0x59, 0xb4, 0x26, 0x04, 0x6e, 0x6b, 0xb0, 0x5c, 0x36, 0xaf, 
+    0xde, 0x83, 0x05, 0xfe, 0x12, 0xc3, 0xe5, 0x71, 0xf6, 0x8f, 
+    0x0c, 0xcd, 0xf6, 0x8f, 0x24, 0xcd, 0x09, 0x70, 0xdb, 0x32, 
+    0xb0, 0x48, 0x23, 0x45, 0xb7, 0x93, 0xae, 0xcc, 0x66, 0x56, 
+    0x86, 0x3f, 0xa5, 0xe8, 0x76, 0xbf, 0xcf, 0x08, 0xa1, 0xed, 
+    0xff, 0x08, 0x1d, 0xad, 0xff, 0x08, 0x15, 0xad, 0x00, 0xf7, 
+    0xea, 0x52, 0x78, 0x44, 0x14, 0xba, 0x06, 0x42, 0x66, 0xcb, 
+    0xc3, 0xe8, 0x9d, 0x75, 0x48, 0x9b, 0x03, 0x94, 0x7b, 0x80, 
+    0xff, 0x8a, 0xfb, 0x11, 0xce, 0x96, 0x5b, 0x11, 0xce, 0x92, 
+    0xa4, 0xee, 0x31, 0x6d, 0xe5, 0xe4, 0x8d, 0x49, 0x4d, 0xb4, 
+    0x11, 0x98, 0x0c, 0x8e, 0x94, 0xbd, 0x3e, 0xfc, 0xbc, 0x24, 
+    0xe0, 0x7c, 0xfa, 0x85, 0x95, 0xcc, 0x9d, 0x87, 0x94, 0x4c, 
+    0x9d, 0x87, 0x6b, 0xb3, 0x62, 0x78, 0x01, 0x54, 0xa5, 0xaa, 
+    0xd9, 0xdd, 0x03, 0xe6, 0xfc, 0x51, 0xce, 0x8e, 0xd4, 0x0c, 
+    0x6a, 0xe7, 0x4c, 0xf4, 0xfd, 0x13, 0x58, 0x74, 0x87, 0xb3, 
+    0x58, 0x74, 0x87, 0xbb, 0xa7, 0x8b, 0x78, 0x44, 0x8a, 0x0d, 
+    0x4a, 0xe6, 0x03, 0x6e, 0x2c, 0xfe, 0xb5, 0x8b, 0x96, 0xbd, 
+    0xc1, 0xdc, 0xf7, 0xb1, 0x51, 0xc7, 0x5f, 0x03, 0x51, 0xf7, 
+    0x7b, 0x95, 0x51, 0xe7, 0x7b, 0x95, 0xae, 0x18, 0x84, 0x6a, 
+    0x54, 0x66, 0x1d, 0xda, 0x22, 0x63, 0xdd, 0x34, 0x7a, 0x6a, 
+    0x2e, 0x56, 0x99, 0x82, 0x9a, 0xc7, 0xf2, 0x7e, 0x90, 0x5e, 
+    0x7a, 0x4e, 0xda, 0x76, 0xfa, 0x4e, 0xda, 0x76, 0x05, 0xb1, 
+    0x25, 0x89, 0x0d, 0xef, 0x72, 0xd8, 0x3c, 0xe2, 0x2a, 0x60, 
+    0x9d, 0xd5, 0xf0, 0x43, 0x2b, 0x50, 0x86, 0x30, 0xea, 0x59, 
+    0x9d, 0x45, 0xe8, 0x18, 0xd9, 0xe7, 0xe8, 0x18, 0xd9, 0xe7, 
+    0x17, 0xe7, 0x26, 0x18, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x69, 0x4a, 0x80, 0xce, 
+    0xd0, 0x4e, 0x9b, 0x8a, 0xba, 0xa5, 0xd9, 0x53, 0x75, 0x4c, 
+    0x98, 0x23, 0xcc, 0xe4, 0xd7, 0x31, 0x3d, 0x00, 0xfd, 0x0b, 
+    0x7d, 0x00, 0xdd, 0x0b, 0x82, 0xff, 0x22, 0xf4, 0x75, 0x49, 
+    0x5f, 0xcc, 0xdb, 0xae, 0xde, 0xab, 0xdd, 0x30, 0x8b, 0x2c, 
+    0xa6, 0xc4, 0xb5, 0xf1, 0x97, 0x01, 0x7a, 0x23, 0xa3, 0x95, 
+    0x0e, 0x2d, 0xa3, 0x95, 0x2e, 0x2d, 0x5c, 0x6a, 0xd1, 0xd2, 
+    0x65, 0xf0, 0x53, 0x99, 0xf0, 0x91, 0xef, 0x32, 0x96, 0xf7, 
+    0x12, 0xc3, 0xe0, 0x17, 0x13, 0xac, 0x3c, 0x53, 0x2f, 0x1f, 
+    0x54, 0x77, 0x33, 0xde, 0x14, 0x77, 0x33, 0x9e, 0xeb, 0x88, 
+    0xcc, 0x61, 0x02, 0x02, 0xca, 0xed, 0xa2, 0xa4, 0xcf, 0x20, 
+    0xf7, 0x92, 0xdb, 0x88, 0x11, 0x86, 0xee, 0xcf, 0x9a, 0xe0, 
+    0xc9, 0x6f, 0x9b, 0x82, 0x88, 0xee, 0x9b, 0x82, 0xc8, 0xee, 
+    0x64, 0x7d, 0x37, 0x11, 0x46, 0x01, 0xa3, 0xbc, 0x0e, 0xc4, 
+    0x72, 0xd1, 0xc3, 0x85, 0xad, 0xcb, 0x11, 0x25, 0x56, 0x89, 
+    0x4e, 0x27, 0xcd, 0xcd, 0xa6, 0xa7, 0x6c, 0x7d, 0x86, 0x27, 
+    0xec, 0xfd, 0x79, 0xd8, 0x13, 0x02, 0xa0, 0x26, 0xfe, 0xe8, 
+    0xfc, 0xf5, 0x64, 0x2d, 0x59, 0xa7, 0x92, 0xd4, 0x5a, 0xd3, 
+    0xbe, 0x77, 0x08, 0x4e, 0x04, 0x86, 0x50, 0xce, 0x36, 0x84, 
+    0x50, 0xce, 0x36, 0x84, 0xaf, 0x31, 0xc9, 0x7b, 0xbb, 0x25, 
+    0x67, 0x70, 0x0c, 0x20, 0x92, 0x7f, 0x30, 0xc2, 0xd0, 0x88, 
+    0x27, 0xaa, 0x32, 0xaf, 0x65, 0xd0, 0x19, 0x4f, 0x83, 0x98, 
+    0x1e, 0xb3, 0x03, 0x98, 0x1e, 0xb3, 0xfc, 0x67, 0xe1, 0x4c, 
+    0x73, 0x1e, 0x1f, 0xd8, 0xb7, 0x16, 0x15, 0x8b, 0x96, 0xc5, 
+    0x15, 0xa3, 0xa5, 0x2b, 0xae, 0x7a, 0x30, 0xb5, 0x0f, 0x89, 
+    0x2f, 0x92, 0x37, 0xa8, 0x27, 0x92, 0x17, 0xa8, 0xd8, 0x6d, 
+    0xe8, 0x57, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x4b, 0x8a, 0x05, 0xdc, 0x97, 0x61, 
+    0xcf, 0x05, 0xb4, 0x98, 0x55, 0x98, 0xc4, 0xc5, 0xd8, 0xb1, 
+    0x28, 0x49, 0x74, 0x9d, 0x89, 0x8b, 0x74, 0xbd, 0x89, 0xcb, 
+    0x74, 0xbd, 0x76, 0x34, 0x8b, 0x42, 0xe3, 0x67, 0x8e, 0x28, 
+    0x5e, 0x2e, 0x32, 0x1f, 0xdf, 0x9e, 0xa1, 0x1f, 0xcf, 0x0f, 
+    0x1b, 0x16, 0xaa, 0x3e, 0x33, 0x31, 0x83, 0x37, 0xb1, 0x75, 
+    0x83, 0x36, 0xb1, 0x75, 0x7c, 0xc9, 0x4e, 0x8a, 0xe4, 0x3f, 
+    0xc5, 0x8b, 0xd2, 0x10, 0xf2, 0xc2, 0x99, 0x4f, 0x15, 0xe1, 
+    0x72, 0xfc, 0x83, 0xf4, 0x79, 0x03, 0x05, 0x83, 0x79, 0x0f, 
+    0xc7, 0x80, 0x79, 0x0f, 0xc7, 0x80, 0x86, 0xf0, 0x38, 0x7f, 
+    0x91, 0x81, 0xee, 0x27, 0xf3, 0x13, 0xbd, 0x47, 0x55, 0xbf, 
+    0xe1, 0x5e, 0xa4, 0x47, 0xed, 0xf3, 0x1b, 0x7b, 0x41, 0x5b, 
+    0x26, 0x5b, 0xe5, 0x7b, 0x26, 0x5b, 0x65, 0x7b, 0xd9, 0xa4, 
+    0x9a, 0x84, 0x3b, 0x70, 0x1e, 0x38, 0x31, 0x21, 0x9c, 0x48, 
+    0xe7, 0x9c, 0x1b, 0x8d, 0x67, 0xc9, 0x9c, 0x89, 0x26, 0x1a, 
+    0xc3, 0x2f, 0x27, 0xd8, 0xf3, 0xbb, 0x27, 0x98, 0xb3, 0xab, 
+    0xd8, 0x67, 0x4c, 0x54, 0x48, 0xbd, 0xf9, 0x4f, 0x2c, 0x37, 
+    0xc9, 0xdd, 0x4d, 0xef, 0x59, 0x09, 0x73, 0x9d, 0xb8, 0x36, 
+    0x9b, 0x06, 0x11, 0xc0, 0xb3, 0x16, 0x5d, 0x09, 0xb3, 0x16, 
+    0x59, 0x09, 0x4c, 0xe9, 0xa6, 0xf6, 0x25, 0xf5, 0xf8, 0x08, 
+    0x9b, 0x75, 0x87, 0x32, 0x0a, 0xd0, 0x64, 0x04, 0xbb, 0x9d, 
+    0x5c, 0x30, 0x75, 0xa3, 0x9e, 0x55, 0xb7, 0x94, 0xcd, 0x01, 
+    0xb7, 0x97, 0xcd, 0x01, 0x48, 0x68, 0x32, 0xfe, 0x4f, 0x22, 
+    0x33, 0xd1, 0x46, 0x2a, 0x5d, 0xc7, 0x69, 0x59, 0xf0, 0xc2, 
+    0x1e, 0xc6, 0xb6, 0xca, 0xa0, 0x1f, 0xd6, 0x1f, 0x2b, 0xc6, 
+    0xdc, 0xfb, 0x29, 0xc6, 0xd4, 0xfb, 0xd6, 0x39, 0x2b, 0x04, 
+    0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x78, 0xdb, 0x23, 0x5a, 0xc8, 0xac, 0xb1, 0x69, 
+    0x2d, 0x8a, 0xc1, 0x6a, 0x7b, 0x9c, 0x59, 0x1c, 0xc6, 0x09, 
+    0x3e, 0x88, 0xfe, 0x00, 0x5b, 0x68, 0xfe, 0x00, 0x5b, 0x68, 
+    0x01, 0xff, 0xa4, 0x97, 0x62, 0xd2, 0xf9, 0xdf, 0x04, 0x54, 
+    0x66, 0xcf, 0x69, 0x13, 0x3e, 0x79, 0xc4, 0xa3, 0x28, 0x1c, 
+    0x81, 0xb4, 0x86, 0xa5, 0xc4, 0xa4, 0xa8, 0x11, 0xc4, 0xa4, 
+    0xa0, 0x11, 0x3b, 0x5b, 0x5f, 0xee, 0xbe, 0x03, 0x85, 0x09, 
+    0x4c, 0x70, 0xa5, 0x8a, 0x8a, 0xe9, 0x64, 0x34, 0x4d, 0x6b, 
+    0x08, 0xe0, 0x6f, 0x60, 0xdc, 0x8c, 0x4f, 0x72, 0xc8, 0xa4, 
+    0x4f, 0x70, 0xc8, 0xac, 0xb0, 0x8f, 0x37, 0x53, 0xd0, 0xd6, 
+    0xbe, 0x82, 0x0b, 0x60, 0xbc, 0x0b, 0xea, 0xc9, 0x00, 0xb8, 
+    0x4d, 0xf7, 0xc5, 0x2b, 0x43, 0xc2, 0xcd, 0x2c, 0x4b, 0x4a, 
+    0x85, 0x16, 0x4b, 0xc8, 0xc5, 0x16, 0xb4, 0x37, 0x3a, 0xe9, 
+    0x1f, 0x07, 0x35, 0x29, 0x54, 0x16, 0x57, 0xfc, 0xac, 0x94, 
+    0x3b, 0xf2, 0xb7, 0x54, 0xa6, 0x89, 0xc4, 0xa5, 0x94, 0x63, 
+    0x85, 0x87, 0xe3, 0xe7, 0x85, 0x87, 0xa2, 0xe3, 0x7a, 0x78, 
+    0x5d, 0x1c, 0x11, 0xa7, 0xf8, 0x78, 0x8c, 0x11, 0xa1, 0xe8, 
+    0x75, 0xbf, 0x6e, 0x3f, 0x05, 0xe7, 0x2d, 0xae, 0x9e, 0xed, 
+    0x9a, 0x32, 0xd1, 0xef, 0xdb, 0x22, 0xd1, 0xef, 0x9b, 0x22, 
+    0x2e, 0x10, 0x64, 0xdd, 0x00, 0xb6, 0x03, 0x5d, 0x88, 0x5b, 
+    0x8f, 0xc7, 0x71, 0x40, 0x83, 0x28, 0x6f, 0x1d, 0xf3, 0xf3, 
+    0x7c, 0x40, 0xd4, 0xe9, 0x4c, 0x51, 0xc8, 0xe3, 0x4c, 0x51, 
+    0xc4, 0xe3, 0xb3, 0xae, 0x3b, 0x1c, 0xa8, 0x94, 0x08, 0x8f, 
+    0x88, 0x6d, 0x96, 0x3d, 0xcd, 0x50, 0xa6, 0xf4, 0x58, 0x41, 
+    0xd7, 0xa1, 0x58, 0x3a, 0x2d, 0x4e, 0x74, 0x78, 0x2f, 0xcc, 
+    0x74, 0x78, 0x2f, 0xcc, 0x8b, 0x87, 0xd0, 0x33, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0xca, 0xd3, 0x0c, 0x83, 0x34, 0xb6, 0x8c, 0x65, 0xb7, 0x0f, 
+    0x20, 0x7d, 0xf7, 0xaa, 0x57, 0xc3, 0x1b, 0x4a, 0xb4, 0xcf, 
+    0x6b, 0x2a, 0xb4, 0x76, 0x7b, 0x2a, 0xb4, 0x76, 0x84, 0xd5, 
+    0x4b, 0x89, 0x47, 0xe6, 0x1e, 0xd2, 0x99, 0x2c, 0x18, 0xf7, 
+    0x80, 0x02, 0x61, 0x8a, 0x12, 0xfc, 0x89, 0xa3, 0x99, 0x93, 
+    0x64, 0xec, 0x8d, 0xb2, 0xe1, 0xf7, 0x89, 0xb3, 0xe1, 0xf7, 
+    0x76, 0x4c, 0x1e, 0x08, 0x1f, 0xff, 0x88, 0x4b, 0x72, 0x31, 
+    0x0e, 0x14, 0xd1, 0x68, 0x91, 0x9d, 0x2d, 0x90, 0xa8, 0x8d, 
+    0xc1, 0x85, 0xf8, 0x6e, 0x14, 0x80, 0xb2, 0x6d, 0x04, 0x80, 
+    0xb8, 0x6d, 0xfb, 0x7f, 0x47, 0x92, 0x6c, 0x25, 0xfe, 0xbb, 
+    0x22, 0x74, 0xac, 0x32, 0x08, 0xd0, 0xb9, 0x98, 0x8a, 0x2c, 
+    0xc9, 0xba, 0xc4, 0x03, 0x70, 0x24, 0xcc, 0x07, 0xbd, 0x38, 
+    0xcc, 0x07, 0xbc, 0x38, 0x33, 0xf8, 0x43, 0xc7, 0x03, 0x8f, 
+    0x12, 0x85, 0x7e, 0xc2, 0xa2, 0x58, 0x2e, 0xdf, 0xee, 0xd9, 
+    0x5a, 0x4e, 0x35, 0x0f, 0xdc, 0x19, 0x68, 0xc3, 0x56, 0x9a, 
+    0x01, 0x5b, 0x5e, 0x1a, 0x41, 0x5b, 0xa1, 0xe5, 0xbe, 0xa4, 
+    0x61, 0x55, 0xb5, 0x36, 0x7c, 0xaf, 0x79, 0x28, 0xee, 0x65, 
+    0x4c, 0x17, 0x29, 0xa6, 0x9a, 0xe1, 0xdb, 0xd6, 0xfa, 0x83, 
+    0x5f, 0xe8, 0xac, 0x29, 0x5d, 0xc6, 0xa8, 0x29, 0xa2, 0x39, 
+    0x57, 0xd6, 0xf4, 0xa1, 0x69, 0x75, 0x8b, 0x66, 0x66, 0xbd, 
+    0xe9, 0xd5, 0x3f, 0xda, 0x69, 0x1e, 0x0a, 0x95, 0x47, 0xe9, 
+    0x00, 0x5b, 0xa3, 0xe0, 0x79, 0x0e, 0xa3, 0xe8, 0x58, 0x1a, 
+    0x5c, 0x17, 0xa7, 0xe5, 0xc2, 0x17, 0xe0, 0xda, 0x9e, 0x76, 
+    0x68, 0x43, 0x21, 0xd1, 0xff, 0x6c, 0x24, 0x85, 0x79, 0x50, 
+    0x75, 0xf4, 0x02, 0xcb, 0xe5, 0xfc, 0x44, 0xe7, 0x65, 0xf4, 
+    0x40, 0xee, 0x9a, 0x0b, 0xbf, 0x11, 0x01, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x07, 0x5f, 
+    0x4b, 0x87, 0xac, 0x0d, 0xca, 0xed, 0x8e, 0x1c, 0x34, 0xe7, 
+    0xaa, 0x88, 0xa2, 0x36, 0x5c, 0x47, 0xa8, 0xd2, 0x00, 0x04, 
+    0xeb, 0x96, 0x04, 0x04, 0xea, 0x96, 0xfb, 0xfb, 0x15, 0x69, 
+    0x0c, 0x78, 0xa1, 0x71, 0x29, 0x4b, 0x5e, 0x7f, 0x2a, 0xdd, 
+    0x8e, 0x99, 0xc7, 0x3a, 0x91, 0x94, 0xee, 0x9b, 0x1d, 0x25, 
+    0xbe, 0x11, 0x15, 0x00, 0xbe, 0x19, 0x1d, 0x00, 0x41, 0xe6, 
+    0xe2, 0xff, 0x88, 0xdb, 0x6b, 0x02, 0x47, 0xe6, 0x35, 0xe8, 
+    0xa7, 0x9b, 0x5b, 0x99, 0xc6, 0xb8, 0xd7, 0xa2, 0x08, 0x03, 
+    0x76, 0x40, 0xe0, 0x97, 0xfb, 0xba, 0xc0, 0x97, 0xfb, 0xaa, 
+    0x3f, 0x68, 0x04, 0x55, 0x4a, 0x42, 0x8e, 0xa3, 0x33, 0x66, 
+    0xa3, 0x39, 0x4e, 0xc0, 0xf9, 0x7e, 0x05, 0x86, 0x00, 0x34, 
+    0x8b, 0x7f, 0x5e, 0x3b, 0x80, 0x6f, 0x5c, 0xbb, 0x81, 0x6f, 
+    0x5c, 0xbb, 0x7e, 0x90, 0xa3, 0x44, 0xea, 0xb7, 0xac, 0x08, 
+    0x62, 0x5d, 0x4a, 0x25, 0x38, 0xd8, 0xef, 0xb8, 0x15, 0xa7, 
+    0xe9, 0xda, 0x80, 0xcf, 0x71, 0x23, 0x16, 0xa6, 0x3b, 0x73, 
+    0x14, 0xae, 0x7b, 0x73, 0xeb, 0x51, 0x84, 0x8c, 0x7a, 0x61, 
+    0x8d, 0x92, 0x48, 0x9f, 0xd7, 0x0e, 0x62, 0x8e, 0x5a, 0x58, 
+    0x47, 0x60, 0x35, 0x9c, 0x5a, 0x6e, 0x10, 0x2f, 0x0a, 0x19, 
+    0x1c, 0x3f, 0x0a, 0x19, 0x1c, 0x3f, 0xf5, 0xe6, 0xe3, 0xc0, 
+    0x07, 0x93, 0x9e, 0x91, 0x7b, 0x6d, 0x0a, 0x4a, 0x2a, 0x21, 
+    0xdb, 0xfb, 0x40, 0x1c, 0x96, 0xff, 0xa1, 0x8e, 0x14, 0xde, 
+    0x12, 0xff, 0x30, 0x7e, 0x02, 0xff, 0x12, 0xfe, 0xfd, 0x00, 
+    0xed, 0x01, 0x42, 0x11, 0x91, 0xaa, 0xfe, 0x89, 0x1f, 0x1a, 
+    0x9b, 0xb7, 0xb2, 0xf6, 0xef, 0x73, 0x92, 0x49, 0x8d, 0x84, 
+    0xcc, 0x21, 0x98, 0x86, 0x98, 0x06, 0x98, 0x86, 0x98, 0x02, 
+    0x67, 0x79, 0x67, 0xfd, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x38, 0xcd, 0x21, 0x7b, 
+    0x6a, 0xaf, 0x7c, 0xf0, 0xc4, 0x58, 0x07, 0x7f, 0x2d, 0xfb, 
+    0xdb, 0xfe, 0xa7, 0x81, 0x9d, 0x89, 0x8d, 0xd9, 0xd9, 0xca, 
+    0xad, 0xd9, 0xd9, 0xca, 0x52, 0x26, 0x26, 0x35, 0x26, 0xe5, 
+    0xf5, 0xfe, 0x3b, 0xa4, 0x71, 0x1b, 0x44, 0xcb, 0xfd, 0xb6, 
+    0x9a, 0xf3, 0x27, 0x22, 0xc8, 0xfa, 0x78, 0x6c, 0xe9, 0xfa, 
+    0x6e, 0x74, 0xe8, 0xfa, 0x6e, 0x74, 0x17, 0x05, 0x91, 0x8b, 
+    0x93, 0x49, 0xa2, 0x64, 0xfc, 0x78, 0x4b, 0xb5, 0x20, 0x51, 
+    0x90, 0xb5, 0x4e, 0x65, 0x9e, 0xa3, 0x28, 0xa0, 0x04, 0xba, 
+    0x40, 0xa6, 0x48, 0x39, 0x40, 0xa6, 0x48, 0x39, 0xbf, 0x59, 
+    0xb7, 0xc6, 0x2f, 0xf4, 0xdf, 0x7b, 0x46, 0x8f, 0x30, 0x10, 
+    0x65, 0x59, 0x79, 0x8e, 0xc7, 0x96, 0xe4, 0xed, 0x3f, 0x2b, 
+    0x6d, 0x9f, 0xae, 0xad, 0xe9, 0x99, 0xae, 0xaf, 0x69, 0x9d, 
+    0x51, 0x50, 0x96, 0x62, 0xb9, 0x23, 0x8e, 0xa3, 0x0b, 0x15, 
+    0x87, 0x3a, 0x9b, 0x56, 0x2d, 0x71, 0x0e, 0x35, 0xb6, 0xea, 
+    0x8d, 0xce, 0xa1, 0x29, 0x8e, 0x2a, 0xfa, 0x0b, 0x8e, 0x0a, 
+    0xea, 0x09, 0x71, 0xf5, 0x15, 0xf6, 0x61, 0x8e, 0xce, 0xa5, 
+    0x47, 0xce, 0x20, 0x70, 0x3b, 0xb3, 0x8b, 0xa1, 0x0b, 0xbe, 
+    0x46, 0x5f, 0x2a, 0x45, 0xdc, 0x1b, 0x8b, 0x9a, 0x10, 0xbf, 
+    0x8b, 0x9a, 0x90, 0xbf, 0x74, 0x65, 0x6f, 0x40, 0xfd, 0x99, 
+    0x2b, 0x2c, 0x34, 0xd2, 0x98, 0x6c, 0xa2, 0x96, 0x7f, 0xeb, 
+    0x01, 0x7f, 0xce, 0xf3, 0x57, 0xde, 0x3a, 0xbc, 0x73, 0xbe, 
+    0x1d, 0xd9, 0x73, 0xbe, 0x1f, 0xd9, 0x8c, 0x41, 0xe0, 0x26, 
+    0xce, 0xa8, 0xec, 0x37, 0xd6, 0x13, 0x47, 0xb2, 0x58, 0xbd, 
+    0x88, 0xa5, 0x22, 0xbb, 0x85, 0x61, 0xaf, 0xbd, 0x6d, 0x5a, 
+    0x32, 0x89, 0xed, 0x6b, 0x3a, 0x99, 0xed, 0x6b, 0xc5, 0x66, 
+    0x12, 0x94, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x7c, 0x06, 0x63, 0x8a, 0x5e, 0x18, 
+    0xbb, 0x88, 0x3a, 0x7a, 0xf9, 0xd3, 0x4d, 0x85, 0xfe, 0x2a, 
+    0x19, 0x8f, 0x50, 0xc3, 0xcd, 0x23, 0x9f, 0xcc, 0xcd, 0x2b, 
+    0xdf, 0xce, 0x32, 0xd4, 0x20, 0x31, 0x81, 0x32, 0xe7, 0x3e, 
+    0x93, 0x3a, 0x17, 0x89, 0x21, 0x48, 0xf5, 0xd7, 0x60, 0x44, 
+    0x37, 0x04, 0x78, 0x53, 0xb3, 0xed, 0xb0, 0xcc, 0x3e, 0xad, 
+    0xb0, 0xc4, 0x36, 0xad, 0x4f, 0x3b, 0xc9, 0x52, 0x3a, 0x01, 
+    0x6b, 0xa2, 0xf2, 0x11, 0x50, 0x3d, 0x30, 0xb4, 0x5e, 0xfe, 
+    0x87, 0x19, 0x58, 0x7c, 0x7d, 0x90, 0xa4, 0x7a, 0xb9, 0x3c, 
+    0xb8, 0xf6, 0xb9, 0x3c, 0xb8, 0xfe, 0x46, 0xc3, 0x47, 0x01, 
+    0xbb, 0xb5, 0xfc, 0x72, 0xc6, 0x36, 0x7c, 0xbd, 0x08, 0x16, 
+    0xf8, 0x39, 0x0e, 0xcf, 0xed, 0xa8, 0x1a, 0x8f, 0xd0, 0x19, 
+    0x53, 0xe7, 0x55, 0x39, 0x13, 0xe7, 0x55, 0x39, 0xec, 0x18, 
+    0xaa, 0xc6, 0xc2, 0x10, 0x66, 0xbf, 0x83, 0xbf, 0x7a, 0x46, 
+    0x84, 0xe2, 0x90, 0xf9, 0x16, 0xf6, 0x0d, 0x2e, 0x1c, 0x8f, 
+    0xb0, 0x81, 0x08, 0x96, 0xf4, 0x1b, 0x18, 0x96, 0xf4, 0x1b, 
+    0xe7, 0x69, 0x0b, 0xe4, 0x7f, 0x88, 0xfd, 0xa9, 0xaa, 0x9c, 
+    0xb8, 0x31, 0x91, 0xf0, 0xe1, 0xa6, 0xe2, 0x4e, 0x8d, 0x99, 
+    0x15, 0xbe, 0x00, 0x0c, 0x79, 0xfa, 0x69, 0x97, 0x79, 0xfe, 
+    0x49, 0x95, 0x86, 0x01, 0xb6, 0x6a, 0xc8, 0xde, 0x64, 0x2c, 
+    0x07, 0x35, 0xdf, 0x26, 0x62, 0x1f, 0xa3, 0xc3, 0x95, 0xfc, 
+    0x2d, 0x04, 0x46, 0x90, 0x31, 0xbd, 0x25, 0x7d, 0x73, 0x0c, 
+    0x25, 0x7c, 0x73, 0x0c, 0xda, 0x83, 0x8c, 0xf3, 0xfe, 0x6b, 
+    0xd7, 0x46, 0x86, 0xf6, 0x0b, 0x5e, 0x55, 0xde, 0x4c, 0x5d, 
+    0x67, 0xec, 0x0f, 0x6c, 0x7c, 0x97, 0x6a, 0x2a, 0xe6, 0x8f, 
+    0x2a, 0xa3, 0x66, 0xdf, 0x2a, 0xab, 0x99, 0x20, 0xd5, 0x54, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x12, 0x00, 0x77, 0x8b, 0xee, 0x27, 0x42, 0x81, 
+    0x7c, 0x43, 0x31, 0xf3, 0x45, 0x89, 0xed, 0x09, 0x18, 0x79, 
+    0xda, 0x83, 0xce, 0xf8, 0x4f, 0x03, 0xce, 0xf8, 0xdb, 0x03, 
+    0x31, 0x07, 0x24, 0xfc, 0x84, 0xb3, 0x04, 0x7b, 0x22, 0xf6, 
+    0xa6, 0x45, 0x8a, 0x6a, 0xcb, 0x6f, 0x5e, 0x6b, 0x31, 0x55, 
+    0xe7, 0x2d, 0x17, 0x2a, 0xf4, 0x03, 0x33, 0x3b, 0xf6, 0x43, 
+    0x33, 0x3b, 0x09, 0xbc, 0xcc, 0xc4, 0x16, 0x4c, 0x18, 0xd8, 
+    0xc8, 0xa2, 0xa8, 0x45, 0xed, 0xd9, 0xc8, 0x6a, 0xe1, 0x39, 
+    0x8c, 0x26, 0x2c, 0x21, 0x8e, 0x26, 0x40, 0xf1, 0x86, 0x04, 
+    0x78, 0x71, 0x8e, 0x26, 0x87, 0x8e, 0x71, 0xd9, 0x61, 0x1e, 
+    0xb0, 0x82, 0x4b, 0x54, 0xb5, 0xde, 0x8d, 0xba, 0x26, 0xc7, 
+    0xbd, 0x54, 0xc7, 0x6f, 0x2c, 0x1f, 0x50, 0x38, 0x18, 0x5d, 
+    0x40, 0x3e, 0x18, 0x5d, 0x40, 0x3e, 0xe7, 0xa2, 0xbf, 0xc1, 
+    0x2d, 0x17, 0x5a, 0xb0, 0xa9, 0xdd, 0x60, 0x10, 0xe9, 0xd3, 
+    0x15, 0x8a, 0x7d, 0x94, 0x20, 0xf1, 0x8e, 0x0c, 0xdb, 0x48, 
+    0xef, 0x4b, 0xab, 0x9e, 0xef, 0x4b, 0xab, 0xde, 0x10, 0xb4, 
+    0x54, 0x21, 0x2c, 0x3e, 0x3a, 0x45, 0x53, 0x0e, 0x7f, 0x2f, 
+    0x44, 0x31, 0x55, 0x48, 0xec, 0x31, 0x2b, 0x38, 0xb4, 0xa0, 
+    0xe1, 0x0a, 0xbc, 0x0b, 0x61, 0x2b, 0xbc, 0x03, 0x61, 0x6b, 
+    0x43, 0xfc, 0x9e, 0xd4, 0x6c, 0x85, 0xa1, 0x59, 0x6c, 0x25, 
+    0x95, 0x9c, 0x63, 0xd9, 0xbc, 0x35, 0x18, 0x03, 0x89, 0x6b, 
+    0xc7, 0xd2, 0xf5, 0xee, 0xe3, 0x71, 0xc5, 0xaa, 0xe3, 0x51, 
+    0xc5, 0xea, 0x1c, 0xae, 0x3a, 0x15, 0x05, 0x1b, 0x2c, 0xd0, 
+    0x30, 0xc2, 0x94, 0x31, 0x54, 0xc8, 0x8b, 0x8f, 0xde, 0x62, 
+    0x1d, 0x58, 0x99, 0x77, 0x2c, 0xc9, 0x1c, 0xf3, 0x34, 0x4d, 
+    0x1c, 0xf3, 0x3c, 0x4d, 0xe3, 0x0c, 0xc3, 0xb2, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x14, 0x2e, 0x50, 0x55, 0x21, 0x4f, 0xb4, 0x93, 0x75, 0x95, 
+    0x3f, 0x2f, 0x6e, 0x70, 0x02, 0x44, 0x9e, 0xd4, 0xee, 0x10, 
+    0xeb, 0x75, 0x42, 0x4c, 0xeb, 0x75, 0x42, 0x44, 0x14, 0x8a, 
+    0xbd, 0xbb, 0xef, 0x2d, 0xb8, 0x77, 0x9f, 0x2c, 0x1d, 0x29, 
+    0xe0, 0x05, 0x9c, 0x5a, 0xcf, 0xf6, 0xf6, 0x79, 0x4b, 0x06, 
+    0xac, 0xef, 0xaa, 0x94, 0xac, 0xed, 0xca, 0x14, 0xac, 0xef, 
+    0x35, 0xeb, 0x53, 0x10, 0xb9, 0x22, 0x21, 0xcc, 0x95, 0x41, 
+    0x36, 0xec, 0x7c, 0x83, 0x66, 0xe0, 0x36, 0x48, 0x65, 0xe4, 
+    0x7a, 0xf2, 0xe2, 0x56, 0xb6, 0xf0, 0xe0, 0xee, 0xb6, 0xf0, 
+    0xe0, 0xee, 0x49, 0x0f, 0x1f, 0x11, 0x7d, 0x76, 0xd8, 0xa0, 
+    0xc1, 0x10, 0xb9, 0x4c, 0xcd, 0xb2, 0xa4, 0x27, 0x53, 0x97, 
+    0xe3, 0x7e, 0xe0, 0xfb, 0x95, 0x97, 0x61, 0x6a, 0x96, 0x1b, 
+    0x61, 0x7b, 0x96, 0x9b, 0x9e, 0x84, 0x69, 0x64, 0x3e, 0x37, 
+    0xa1, 0xb8, 0x3a, 0x09, 0xee, 0x34, 0x2a, 0xd8, 0x7a, 0x51, 
+    0x8b, 0x6c, 0xe8, 0x7d, 0x9c, 0x51, 0x8f, 0xc3, 0x90, 0x1d, 
+    0xbe, 0xc4, 0x90, 0x19, 0xbf, 0xc4, 0x6f, 0xe6, 0x40, 0x3b, 
+    0x13, 0xa1, 0x45, 0x1b, 0x0b, 0x8b, 0x22, 0xec, 0x34, 0x25, 
+    0x8e, 0x44, 0x8c, 0x9b, 0x72, 0x12, 0xb8, 0x8b, 0x5f, 0x54, 
+    0xb0, 0x70, 0x1a, 0x74, 0xb0, 0xba, 0x1a, 0x14, 0x4f, 0x45, 
+    0xe5, 0xeb, 0x52, 0xdc, 0x5b, 0x77, 0xdb, 0xa7, 0xac, 0x22, 
+    0x15, 0xf9, 0xd9, 0xa4, 0xc9, 0x6a, 0xf6, 0x9f, 0xb3, 0x62, 
+    0xd4, 0x57, 0xb1, 0x6a, 0xa9, 0xc5, 0xb1, 0x6a, 0xb9, 0xc5, 
+    0x4e, 0x95, 0x46, 0x3a, 0x82, 0xa5, 0x81, 0x62, 0x47, 0xa5, 
+    0xa6, 0xca, 0x90, 0x67, 0xee, 0x22, 0x57, 0x4d, 0x11, 0x48, 
+    0xe9, 0x55, 0x2d, 0xf7, 0x7a, 0x17, 0x29, 0x8e, 0x7a, 0x55, 
+    0x29, 0xce, 0x85, 0xaa, 0xd6, 0x31, 0x01, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x1c, 0xea, 
+    0x42, 0x9b, 0x87, 0x51, 0xd4, 0xb1, 0x0d, 0xa3, 0x13, 0x96, 
+    0xed, 0xf5, 0x90, 0xce, 0x49, 0x6b, 0xd9, 0xb8, 0x19, 0x62, 
+    0x70, 0xb2, 0x19, 0x63, 0x50, 0xb2, 0xe6, 0x9c, 0xaf, 0x4d, 
+    0x25, 0x1a, 0x7e, 0x0d, 0x88, 0x4b, 0x2e, 0xb3, 0xea, 0x33, 
+    0x81, 0xe3, 0xb2, 0x90, 0xcb, 0x97, 0xd5, 0x33, 0x13, 0x84, 
+    0xb4, 0x03, 0x11, 0x31, 0xb4, 0x03, 0x11, 0x10, 0x4b, 0xfc, 
+    0xee, 0xef, 0x5a, 0x56, 0xd7, 0x2a, 0xc2, 0x83, 0x8a, 0x07, 
+    0xa1, 0x6f, 0xf9, 0x84, 0x26, 0x21, 0x67, 0xa5, 0xe2, 0x8a, 
+    0xc6, 0xb3, 0x52, 0x0e, 0xce, 0x21, 0x52, 0x0e, 0xce, 0x21, 
+    0xad, 0xf1, 0x31, 0xde, 0x28, 0x6a, 0x17, 0x5b, 0x92, 0xd0, 
+    0xf5, 0x33, 0xfa, 0x22, 0x54, 0xf7, 0x82, 0xc3, 0x71, 0x73, 
+    0x80, 0x48, 0x1f, 0x7a, 0x9c, 0x70, 0x74, 0xba, 0x98, 0x70, 
+    0x55, 0xba, 0x67, 0x8f, 0xaa, 0x45, 0x28, 0x9f, 0xb7, 0xcc, 
+    0x72, 0x83, 0x2a, 0xe7, 0x9c, 0xa6, 0xa3, 0x8f, 0x38, 0xe7, 
+    0x6d, 0x7d, 0x3d, 0xb0, 0x83, 0xfb, 0x23, 0xf1, 0x82, 0xff, 
+    0x2b, 0xb1, 0x83, 0xff, 0xd4, 0x4e, 0x7c, 0x00, 0x7c, 0x79, 
+    0x36, 0xcc, 0x2d, 0xf5, 0xae, 0x00, 0x8d, 0x7b, 0xa7, 0xcf, 
+    0xb2, 0xde, 0xa2, 0x7c, 0xf2, 0xd5, 0x95, 0x0d, 0x94, 0xb0, 
+    0x75, 0x26, 0x96, 0xf0, 0x35, 0x24, 0x69, 0x0f, 0xca, 0xdb, 
+    0x20, 0x9d, 0x8f, 0x24, 0xc9, 0x97, 0x2b, 0x1e, 0xf5, 0xaf, 
+    0x7a, 0x22, 0xa6, 0xa7, 0x1e, 0x76, 0xf4, 0x95, 0x7e, 0x7a, 
+    0x3d, 0x97, 0x3e, 0xda, 0xbd, 0x97, 0x3e, 0xfa, 0x42, 0x68, 
+    0xc1, 0x05, 0x78, 0xe2, 0x57, 0x31, 0xcd, 0xe2, 0x53, 0x25, 
+    0x2a, 0x66, 0x68, 0x7d, 0x3d, 0x09, 0xcf, 0x8a, 0x29, 0xb7, 
+    0xe3, 0x4c, 0x0b, 0x77, 0x09, 0x01, 0x09, 0x77, 0x49, 0x01, 
+    0xf6, 0x88, 0xb6, 0xfe, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x98, 0xac, 0x21, 0x34, 
+    0xf3, 0xbb, 0x7f, 0x8f, 0x29, 0x3a, 0x0f, 0x0a, 0xb4, 0xb8, 
+    0x5d, 0x00, 0x89, 0x59, 0xd7, 0x43, 0x0b, 0x3a, 0x64, 0x19, 
+    0x09, 0x3a, 0x65, 0x08, 0xf6, 0xc5, 0x9a, 0xf7, 0xb2, 0x0d, 
+    0xb7, 0xbd, 0xbc, 0x0d, 0x50, 0xca, 0xd4, 0x1d, 0x93, 0x44, 
+    0xc0, 0xa0, 0x14, 0x0f, 0xfe, 0x7a, 0x3c, 0x1b, 0x37, 0xba, 
+    0x61, 0x77, 0x37, 0xba, 0x70, 0x7f, 0xc8, 0x45, 0x8f, 0x80, 
+    0x2e, 0x33, 0xd8, 0x88, 0x61, 0xa4, 0xce, 0x30, 0xe1, 0xea, 
+    0x8c, 0x6f, 0xf2, 0x1f, 0x2d, 0xaa, 0xb0, 0x7b, 0x9c, 0x43, 
+    0xd1, 0xff, 0xed, 0x17, 0xd0, 0xff, 0xad, 0x13, 0x2f, 0x00, 
+    0x52, 0xec, 0x9c, 0xda, 0x41, 0x4d, 0xaa, 0x56, 0xab, 0x95, 
+    0x5e, 0x54, 0xb2, 0xae, 0x30, 0xcd, 0x2c, 0x5b, 0x12, 0x29, 
+    0xe0, 0xe4, 0x02, 0x41, 0xbb, 0x65, 0x02, 0x49, 0xb9, 0x64, 
+    0xfd, 0xb6, 0x46, 0x9b, 0x35, 0xf0, 0xb1, 0x56, 0xa9, 0x8e, 
+    0x45, 0xa5, 0xe7, 0x9e, 0xf1, 0xc1, 0x65, 0xa3, 0x38, 0x96, 
+    0x73, 0x29, 0xa8, 0x51, 0x54, 0x77, 0x31, 0x4a, 0x65, 0x77, 
+    0xb1, 0x43, 0x9a, 0x88, 0x4e, 0xbc, 0x32, 0x24, 0x99, 0x19, 
+    0xd4, 0x08, 0x13, 0x2f, 0xbf, 0x5b, 0x34, 0xfb, 0x2f, 0x93, 
+    0xd1, 0x9b, 0xdc, 0x70, 0x6e, 0x19, 0x56, 0x53, 0x44, 0xf3, 
+    0x5e, 0x53, 0x44, 0x7b, 0xa1, 0xac, 0xbb, 0x84, 0x08, 0xd9, 
+    0x8e, 0x10, 0xb1, 0x88, 0x4e, 0x33, 0x86, 0xdd, 0xc1, 0xbc, 
+    0x0a, 0x83, 0x84, 0x68, 0x75, 0x6f, 0x67, 0x77, 0x44, 0x85, 
+    0x07, 0x76, 0x44, 0x8f, 0x07, 0x77, 0xbb, 0x70, 0xf8, 0x88, 
+    0x83, 0x18, 0x11, 0xdb, 0xf2, 0xea, 0x5e, 0x78, 0x9d, 0xb4, 
+    0xa7, 0xd0, 0x77, 0x32, 0xfd, 0xb2, 0x90, 0xad, 0xb7, 0xdf, 
+    0x30, 0x02, 0xff, 0xd0, 0x20, 0x00, 0xff, 0xd0, 0xcf, 0xff, 
+    0x00, 0x2f, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x80, 0xde, 0x9b, 0x24, 0x79, 0xde, 
+    0x52, 0x83, 0xf4, 0x1f, 0x16, 0x0a, 0xa5, 0xf0, 0x7a, 0xfc, 
+    0xd9, 0x0d, 0x5f, 0x0d, 0xf9, 0x9a, 0x33, 0x4b, 0xf9, 0x9a, 
+    0x33, 0x09, 0x06, 0x65, 0xcc, 0xf6, 0x35, 0x08, 0x23, 0xaa, 
+    0x06, 0xb1, 0x5b, 0x50, 0xdf, 0xa4, 0xa3, 0xcd, 0xc0, 0x08, 
+    0xe6, 0x85, 0xcd, 0xb1, 0x2d, 0x1a, 0xdf, 0x32, 0xe6, 0x11, 
+    0xcf, 0x30, 0xa7, 0x11, 0x30, 0xcf, 0x58, 0xee, 0x4c, 0xad, 
+    0x9a, 0x46, 0xe1, 0x56, 0x69, 0x88, 0x9a, 0xbe, 0xe3, 0x93, 
+    0xa5, 0x8d, 0x59, 0x33, 0x0a, 0x86, 0x13, 0x46, 0x08, 0xa6, 
+    0x93, 0x86, 0x08, 0xa6, 0x93, 0x06, 0xf7, 0x59, 0x6c, 0xf9, 
+    0xc0, 0x96, 0x57, 0x1e, 0xef, 0xc6, 0x33, 0x25, 0x5d, 0x30, 
+    0x18, 0x77, 0x04, 0x7f, 0x63, 0x76, 0xc3, 0x81, 0x5d, 0x55, 
+    0xdf, 0xe5, 0xda, 0x74, 0xd7, 0xe5, 0xda, 0x74, 0x28, 0x1a, 
+    0x25, 0x8b, 0x58, 0xa9, 0xfa, 0x1c, 0xd9, 0xb2, 0xaa, 0xa8, 
+    0x52, 0x74, 0xba, 0x35, 0x97, 0xc3, 0x11, 0x61, 0x0d, 0x89, 
+    0x5b, 0x70, 0x11, 0xed, 0x57, 0x73, 0x15, 0xed, 0x53, 0x71, 
+    0xea, 0x12, 0xac, 0x8e, 0xaf, 0x5b, 0xae, 0xc5, 0x6d, 0x5b, 
+    0x68, 0xc1, 0x77, 0x6d, 0xba, 0x91, 0xfb, 0xd0, 0x12, 0x23, 
+    0xbe, 0x02, 0x12, 0xfc, 0x2a, 0xb2, 0x88, 0xfc, 0xaa, 0x92, 
+    0x98, 0xfc, 0x55, 0x6d, 0x67, 0x03, 0xaf, 0x1b, 0xbf, 0x3e, 
+    0xe4, 0x3c, 0x17, 0x8b, 0x4d, 0x1f, 0x23, 0xa6, 0x59, 0x00, 
+    0x3b, 0x4e, 0x54, 0x67, 0xe3, 0xc9, 0x5d, 0x46, 0x01, 0x4b, 
+    0x5d, 0x44, 0x21, 0x4b, 0xa2, 0xbb, 0xde, 0xb4, 0xcf, 0x48, 
+    0x2a, 0x76, 0x2d, 0xb7, 0x31, 0x7f, 0xc7, 0x35, 0xb1, 0x88, 
+    0x94, 0x5f, 0xa7, 0xe9, 0xde, 0xb7, 0xbf, 0x5a, 0x16, 0x37, 
+    0xb7, 0xdd, 0x96, 0x37, 0xb7, 0xd9, 0x69, 0xc8, 0x48, 0x26, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x35, 0x0b, 0x81, 0x6e, 0x7b, 0xa4, 0x51, 0x57, 
+    0x5a, 0xb1, 0xff, 0xd2, 0x53, 0x0a, 0xdc, 0xfc, 0x94, 0x89, 
+    0x24, 0x1a, 0x00, 0x9e, 0x78, 0x3b, 0x10, 0x9b, 0x38, 0x3b, 
+    0xef, 0x64, 0xc7, 0xc4, 0xa4, 0x4f, 0xee, 0x1a, 0x58, 0xb4, 
+    0x30, 0x49, 0x03, 0xc1, 0x50, 0xdc, 0xb1, 0xec, 0x5b, 0x94, 
+    0x83, 0x06, 0x8e, 0x2c, 0x43, 0x84, 0xd9, 0xac, 0x03, 0x84, 
+    0xda, 0xac, 0xfc, 0x7b, 0x25, 0x53, 0x0c, 0x64, 0x40, 0xb9, 
+    0x49, 0x48, 0x41, 0x1d, 0xca, 0x4a, 0x67, 0x57, 0x1c, 0x2e, 
+    0x2b, 0x2d, 0xb4, 0x15, 0x0b, 0xb5, 0x7c, 0x80, 0x96, 0x9d, 
+    0x7c, 0x80, 0x9a, 0x9d, 0x83, 0x7f, 0x65, 0x62, 0x57, 0x29, 
+    0x49, 0x40, 0xbf, 0x1a, 0xbc, 0xa9, 0xbb, 0xa4, 0xb0, 0xa9, 
+    0xe1, 0x39, 0x3b, 0xd6, 0xc7, 0x04, 0x4f, 0xc1, 0x41, 0x2e, 
+    0x2f, 0x85, 0x41, 0x2c, 0x2f, 0xc5, 0xbe, 0xd3, 0xd0, 0x3a, 
+    0xbd, 0x99, 0x11, 0x63, 0xd0, 0xbd, 0x9f, 0xf4, 0xda, 0x71, 
+    0xf1, 0xbd, 0xae, 0xc2, 0x1f, 0xf7, 0x4e, 0x12, 0x29, 0x54, 
+    0x20, 0x52, 0x79, 0xde, 0x28, 0x52, 0x79, 0xd6, 0xd7, 0xad, 
+    0x86, 0x29, 0x3d, 0xe0, 0xe2, 0x58, 0xe4, 0xb3, 0x40, 0x94, 
+    0x24, 0xe9, 0x10, 0xea, 0x70, 0x36, 0x37, 0x84, 0x37, 0xf6, 
+    0xd6, 0x32, 0x35, 0xfc, 0x0a, 0xba, 0x35, 0xfc, 0x1a, 0xba, 
+    0xca, 0x03, 0xe5, 0x45, 0x5c, 0xd7, 0x39, 0x71, 0x09, 0x3b, 
+    0x31, 0x93, 0x18, 0x99, 0x9b, 0x43, 0x33, 0x39, 0x62, 0x7e, 
+    0xd9, 0x02, 0x22, 0xed, 0xd9, 0x08, 0x2a, 0xfe, 0xd9, 0x08, 
+    0x22, 0xff, 0x26, 0xf7, 0xdd, 0x00, 0xf8, 0x80, 0xe6, 0x7a, 
+    0x12, 0xe5, 0x3b, 0x51, 0x04, 0xf9, 0xcc, 0x3f, 0xcb, 0x42, 
+    0xeb, 0x28, 0xae, 0xee, 0x02, 0xd5, 0x0d, 0x18, 0x27, 0x15, 
+    0x0f, 0x18, 0x26, 0x15, 0xf0, 0xe7, 0xd9, 0xea, 0x01, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x4b, 0x1d, 0x31, 0xeb, 0x9e, 0xea, 0x2f, 0x38, 0x01, 0x30, 
+    0xbe, 0x26, 0x94, 0x22, 0x25, 0x73, 0x29, 0x04, 0xe7, 0x8a, 
+    0x85, 0x80, 0x77, 0xfe, 0x84, 0x80, 0x77, 0xfe, 0x7b, 0x7f, 
+    0x88, 0x01, 0x9d, 0x4e, 0xd9, 0xec, 0xa2, 0xd8, 0xd5, 0xa9, 
+    0x8d, 0x42, 0x0f, 0x5e, 0x6f, 0xc6, 0xee, 0xfe, 0xca, 0x2c, 
+    0xd6, 0x5f, 0x66, 0xcd, 0xcf, 0xf2, 0x66, 0x8d, 0xcf, 0xde, 
+    0x99, 0x72, 0x30, 0x21, 0xc2, 0xfb, 0xea, 0x85, 0xb3, 0x92, 
+    0xf2, 0xf9, 0xa3, 0xc8, 0x45, 0xd3, 0x7d, 0x19, 0xf6, 0xcd, 
+    0xd4, 0xc3, 0x45, 0xe7, 0x54, 0xe3, 0x47, 0xdf, 0x54, 0xc3, 
+    0x47, 0xdf, 0xab, 0x3c, 0xb8, 0x20, 0x21, 0x36, 0xfa, 0xa6, 
+    0x35, 0x15, 0x0e, 0xf0, 0x28, 0x44, 0xba, 0x93, 0x70, 0xc8, 
+    0x9e, 0xaf, 0x3c, 0x25, 0x31, 0x0c, 0xdc, 0x55, 0xf0, 0xee, 
+    0x1c, 0x45, 0xf0, 0xee, 0xe3, 0xba, 0x0f, 0x11, 0xe0, 0x28, 
+    0xf4, 0x7e, 0xe5, 0x04, 0xa8, 0xe5, 0x5e, 0x48, 0x1b, 0xc0, 
+    0x8e, 0x17, 0x05, 0xa3, 0x2d, 0xd0, 0x3c, 0xf0, 0x6d, 0x89, 
+    0xb4, 0xe0, 0x6d, 0x88, 0xb4, 0xe0, 0x92, 0x77, 0x4b, 0x1f, 
+    0xc2, 0x0f, 0x82, 0x73, 0xb7, 0x27, 0x88, 0x2e, 0x27, 0x3e, 
+    0x89, 0xfb, 0x61, 0x37, 0xdd, 0x66, 0xb7, 0x19, 0x1e, 0x36, 
+    0x03, 0x59, 0x1c, 0xee, 0x23, 0x19, 0x1c, 0xa6, 0xdc, 0xe6, 
+    0xe3, 0x59, 0xa0, 0x0c, 0x00, 0xd8, 0x55, 0xd9, 0x56, 0x47, 
+    0x6e, 0x8b, 0xc9, 0xb5, 0xd7, 0x39, 0xfb, 0x1a, 0x4d, 0xbc, 
+    0xe5, 0x65, 0x4c, 0x53, 0xc5, 0x60, 0x4c, 0x19, 0xe5, 0x60, 
+    0xb3, 0xe6, 0x1a, 0x9f, 0xf7, 0x25, 0x77, 0x24, 0x51, 0x25, 
+    0x46, 0x11, 0x38, 0x62, 0xc7, 0x6b, 0x52, 0x9a, 0x56, 0x9b, 
+    0x22, 0x56, 0xfa, 0xd4, 0xaa, 0xb2, 0x53, 0x0d, 0x2a, 0xb2, 
+    0x53, 0x0d, 0xd5, 0x4d, 0xac, 0xf2, 0x00, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x0e, 0x64, 
+    0x3e, 0x18, 0x3b, 0xcb, 0xeb, 0xf2, 0x02, 0x44, 0x7b, 0x05, 
+    0x15, 0x20, 0x2e, 0x45, 0x8f, 0x10, 0x33, 0xb6, 0x83, 0xe0, 
+    0xb2, 0x9d, 0x83, 0xe0, 0xb2, 0x9d, 0x7c, 0x1f, 0x4d, 0x62, 
+    0x8e, 0x8a, 0x04, 0x60, 0x17, 0x58, 0xc2, 0x72, 0x0f, 0x5a, 
+    0x00, 0x4e, 0xcc, 0xc9, 0x4f, 0xbf, 0xda, 0x59, 0x02, 0x39, 
+    0x4b, 0xd5, 0x76, 0x29, 0x4b, 0xd1, 0x56, 0x29, 0xb4, 0x2e, 
+    0xa9, 0xd6, 0x9f, 0x51, 0x50, 0xd6, 0xf7, 0x24, 0x06, 0xa5, 
+    0x0f, 0x2a, 0x25, 0xe3, 0x7b, 0xf1, 0x23, 0xb6, 0x2e, 0xa1, 
+    0xf2, 0xeb, 0xbe, 0xed, 0x60, 0xcf, 0xbe, 0xed, 0x60, 0x8f, 
+    0x41, 0x12, 0x9f, 0x70, 0x4a, 0xb0, 0x50, 0x89, 0xe3, 0xe3, 
+    0xd0, 0x22, 0x7e, 0x8f, 0x69, 0x3e, 0x35, 0x72, 0x52, 0xf7, 
+    0x35, 0x03, 0xed, 0xbf, 0xb5, 0x02, 0xcc, 0xbf, 0x35, 0x02, 
+    0xcc, 0xbf, 0xca, 0xfd, 0x33, 0x40, 0xb2, 0x0d, 0x23, 0xcc, 
+    0xf6, 0xcf, 0x89, 0xb3, 0x07, 0x63, 0x92, 0xad, 0x47, 0x3d, 
+    0x7d, 0xd3, 0xb3, 0x7e, 0xf0, 0x2c, 0xa7, 0x93, 0xe7, 0x98, 
+    0xa7, 0xbf, 0xf4, 0x89, 0x58, 0x40, 0x0b, 0x76, 0xda, 0x21, 
+    0x50, 0xb3, 0x08, 0x36, 0xcb, 0x8d, 0xd7, 0x74, 0xec, 0xaf, 
+    0x56, 0x85, 0x2c, 0xf6, 0x28, 0xab, 0xbf, 0xf8, 0xf4, 0x09, 
+    0x8b, 0xe2, 0xf4, 0x09, 0xaf, 0xe2, 0x0b, 0xf6, 0x50, 0x1d, 
+    0x56, 0x3e, 0xd4, 0x03, 0x57, 0x23, 0xf4, 0xa3, 0x10, 0xb0, 
+    0x8d, 0x06, 0x19, 0x69, 0x35, 0x6f, 0xfa, 0xc8, 0x37, 0x69, 
+    0x3f, 0x5d, 0x35, 0x6c, 0x3f, 0x88, 0x35, 0x6d, 0xc0, 0x77, 
+    0xca, 0x92, 0xc8, 0x3f, 0xce, 0x0c, 0xba, 0x1f, 0x44, 0xb7, 
+    0x68, 0xfb, 0xb8, 0x4a, 0xfa, 0x9a, 0x94, 0x72, 0x98, 0x40, 
+    0x89, 0xe2, 0x4f, 0x55, 0x98, 0xc2, 0xcf, 0x53, 0x98, 0xc2, 
+    0x30, 0xac, 0x67, 0x3d, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x0c, 0x43, 0x91, 0x3d, 
+    0xe4, 0x75, 0xe0, 0xb3, 0x42, 0x9e, 0x5c, 0x75, 0x28, 0x41, 
+    0x25, 0xfc, 0x3e, 0xa4, 0xc7, 0x67, 0x23, 0xc8, 0x44, 0x67, 
+    0x23, 0xc8, 0x44, 0x67, 0xdc, 0x37, 0xbb, 0x98, 0x18, 0xb7, 
+    0x3f, 0x2f, 0x0c, 0x34, 0x2e, 0xc8, 0x9f, 0x42, 0xa1, 0xc0, 
+    0x5b, 0x8a, 0xb0, 0x48, 0xc0, 0xbe, 0x45, 0x80, 0x4c, 0x3e, 
+    0xa0, 0x44, 0x4c, 0x3e, 0xa0, 0x40, 0xb3, 0xc1, 0x5f, 0xbf, 
+    0xb8, 0x25, 0xf4, 0xa7, 0x68, 0xb3, 0x3f, 0x24, 0x90, 0xf4, 
+    0xf8, 0x21, 0x7f, 0x86, 0x5d, 0xa6, 0xb4, 0xc9, 0x2b, 0x64, 
+    0xb9, 0xd0, 0xed, 0xac, 0xb9, 0xd0, 0xe9, 0x2c, 0x46, 0x2f, 
+    0x16, 0xd3, 0xd3, 0x0d, 0xa5, 0xf5, 0xb9, 0x05, 0xea, 0x11, 
+    0xfb, 0x92, 0xf6, 0x73, 0xc8, 0xea, 0xcd, 0xe8, 0x0e, 0xe0, 
+    0x65, 0x8f, 0xfc, 0x61, 0x14, 0xcc, 0xfc, 0x61, 0x54, 0xcc, 
+    0x03, 0x9e, 0xab, 0x33, 0xda, 0x88, 0xe9, 0xf2, 0xfd, 0xc8, 
+    0x19, 0x34, 0xde, 0x55, 0x52, 0x7b, 0xe8, 0xbe, 0xd3, 0xf8, 
+    0xa4, 0x55, 0x24, 0xc4, 0xe9, 0x85, 0xe7, 0xf9, 0xe9, 0x05, 
+    0x67, 0xfd, 0x16, 0xfa, 0x98, 0x02, 0xdd, 0xa3, 0x23, 0xec, 
+    0xd5, 0xf7, 0x75, 0x13, 0x5b, 0x8d, 0x54, 0x02, 0x6c, 0xc2, 
+    0x1e, 0x59, 0xc0, 0xfb, 0x5b, 0xc6, 0xf9, 0x5a, 0x77, 0xad, 
+    0xf8, 0xda, 0x7f, 0xe9, 0x07, 0x25, 0x80, 0x16, 0xaa, 0x2f, 
+    0xe7, 0xf3, 0x3b, 0x7d, 0x9d, 0x37, 0x97, 0x81, 0xa0, 0xa8, 
+    0x10, 0x43, 0xe4, 0x3a, 0x4f, 0xed, 0x38, 0xf2, 0x87, 0xe5, 
+    0x27, 0x36, 0x87, 0xe5, 0x25, 0x32, 0x78, 0x1a, 0xda, 0xcd, 
+    0x5d, 0x74, 0xe6, 0xf6, 0x2a, 0x7b, 0xd2, 0x57, 0x76, 0x48, 
+    0xe5, 0x26, 0x6b, 0x86, 0x70, 0x86, 0x56, 0xe3, 0x6a, 0xd7, 
+    0x12, 0x60, 0xce, 0x26, 0x16, 0x60, 0xea, 0x26, 0xe9, 0x9f, 
+    0x15, 0xd9, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x39, 0xda, 0x11, 0x7f, 0x2c, 0x42, 
+    0xe4, 0x4b, 0xb8, 0xe0, 0x90, 0xd0, 0xbb, 0x18, 0x04, 0xc2, 
+    0x7b, 0x8f, 0x86, 0x4d, 0xa5, 0x8d, 0xa6, 0x50, 0xa9, 0x8d, 
+    0x86, 0x40, 0x56, 0x72, 0x79, 0xbf, 0xe5, 0x18, 0x44, 0x8b, 
+    0x1d, 0x18, 0xd4, 0x81, 0xa8, 0xed, 0xc3, 0x76, 0x96, 0x3b, 
+    0xd8, 0x31, 0x74, 0x15, 0x2c, 0x58, 0x46, 0xf5, 0x3e, 0x55, 
+    0x46, 0x55, 0x3c, 0x55, 0xb9, 0xaa, 0xc3, 0xaa, 0xcf, 0xa1, 
+    0xf7, 0x76, 0x8d, 0x9c, 0xef, 0x9a, 0xde, 0x2b, 0x7a, 0x49, 
+    0xc1, 0xc1, 0x38, 0xc8, 0x63, 0x45, 0x42, 0xeb, 0x67, 0x8d, 
+    0xb7, 0x79, 0x67, 0x85, 0x32, 0x79, 0x98, 0x7a, 0xcd, 0x86, 
+    0x05, 0xc0, 0xc6, 0xa4, 0xd0, 0xb0, 0x70, 0xd0, 0xed, 0x54, 
+    0xea, 0x94, 0xf4, 0xaa, 0x1d, 0x75, 0x91, 0x86, 0x3e, 0x6e, 
+    0x1c, 0xeb, 0x0e, 0x6e, 0x94, 0xeb, 0x2e, 0x6e, 0x6b, 0x14, 
+    0xd1, 0x91, 0x6a, 0xde, 0x6a, 0xcb, 0x82, 0x57, 0xf1, 0x95, 
+    0x49, 0xff, 0xe4, 0xf4, 0xd4, 0x20, 0xcc, 0xbd, 0x2c, 0xa4, 
+    0x80, 0x77, 0x05, 0x0c, 0xdb, 0xbe, 0x05, 0x0c, 0xd9, 0x3e, 
+    0xfa, 0xf3, 0x26, 0xc1, 0x9f, 0x92, 0x00, 0x8b, 0x5c, 0xeb, 
+    0x6a, 0xd2, 0x90, 0xab, 0x56, 0x2a, 0xf7, 0xc0, 0x64, 0x11, 
+    0xc6, 0xbc, 0x97, 0x60, 0x90, 0xbc, 0x54, 0x40, 0x90, 0xbc, 
+    0x54, 0x40, 0x6f, 0x43, 0xab, 0xbf, 0x8a, 0xc4, 0xce, 0x3f, 
+    0xae, 0x5b, 0xc5, 0x96, 0x94, 0xa0, 0x67, 0xa2, 0xc7, 0x75, 
+    0x8d, 0x84, 0x30, 0xdb, 0xa1, 0x8b, 0x4e, 0xfb, 0xe1, 0xd8, 
+    0x4e, 0xfb, 0xa1, 0x88, 0xb1, 0x04, 0x5e, 0x77, 0x8a, 0x55, 
+    0xbf, 0xfb, 0x04, 0xfe, 0x21, 0x99, 0x89, 0xbe, 0x4a, 0x99, 
+    0xb5, 0x42, 0x06, 0x84, 0xbc, 0xe0, 0xf0, 0xfc, 0x85, 0x00, 
+    0x05, 0xc3, 0x85, 0x00, 0x04, 0xc3, 0x7a, 0xff, 0xfb, 0x3c, 
+    0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x22, 0xda, 0x8d, 0x48, 0xe1, 0x99, 0x28, 0x49, 
+    0x90, 0x69, 0x60, 0xbe, 0xc5, 0x88, 0x99, 0x9a, 0x1f, 0x3a, 
+    0x78, 0xf0, 0xcb, 0x31, 0x68, 0xc3, 0xcb, 0x31, 0x68, 0xc1, 
+    0x34, 0xce, 0x97, 0x3e, 0x34, 0xff, 0xe7, 0x3a, 0xe6, 0xe1, 
+    0xff, 0xe7, 0xaf, 0x75, 0x3d, 0xb9, 0x2a, 0x06, 0x5b, 0xae, 
+    0x87, 0x59, 0x16, 0xd3, 0xc3, 0x4c, 0x37, 0x53, 0xc3, 0x4c, 
+    0x37, 0xd3, 0x3c, 0xb3, 0xc8, 0x2c, 0x08, 0xdd, 0x74, 0x06, 
+    0x33, 0xf9, 0x61, 0xcc, 0x7c, 0x79, 0xfd, 0x3a, 0x6d, 0x4a, 
+    0x40, 0xb5, 0xea, 0xf0, 0x78, 0x98, 0xec, 0xda, 0x3d, 0xaa, 
+    0xec, 0xda, 0x3d, 0xaa, 0x13, 0x25, 0xc2, 0x55, 0xf2, 0x87, 
+    0x14, 0x8b, 0xb6, 0x29, 0x44, 0xe1, 0xb1, 0x10, 0x9b, 0x13, 
+    0x2f, 0xae, 0xb6, 0x63, 0x8f, 0x8c, 0xce, 0xab, 0x4e, 0x0f, 
+    0x12, 0xab, 0x4f, 0x0c, 0x92, 0xab, 0xb0, 0xf3, 0x6d, 0x54, 
+    0x55, 0x95, 0x9b, 0x2a, 0xf5, 0x17, 0xd0, 0x79, 0x5c, 0xc5, 
+    0x3a, 0xbe, 0xae, 0xf6, 0xb2, 0x68, 0x0a, 0x20, 0xc8, 0xd6, 
+    0x08, 0xc2, 0xcb, 0x68, 0x0a, 0xc0, 0xcb, 0x68, 0xf5, 0x3f, 
+    0x34, 0x97, 0x4c, 0x52, 0x17, 0x9b, 0x91, 0xcb, 0x84, 0x19, 
+    0x7e, 0x61, 0x1a, 0x45, 0x0b, 0x1f, 0x6c, 0x8a, 0xb4, 0x03, 
+    0xb9, 0x8f, 0x95, 0xa3, 0xba, 0x8d, 0x95, 0x23, 0xba, 0x8d, 
+    0x6a, 0xdc, 0x45, 0x72, 0x0d, 0xf0, 0x5b, 0x74, 0xe2, 0x5a, 
+    0x5b, 0xd0, 0x7b, 0x9b, 0x64, 0xe8, 0xb5, 0xbf, 0x72, 0x25, 
+    0xba, 0xd7, 0x2a, 0x2c, 0xbf, 0xa5, 0xc1, 0xfc, 0xbf, 0xa7, 
+    0x40, 0x7c, 0x40, 0x58, 0xbf, 0x83, 0x00, 0xbc, 0xc0, 0xc5, 
+    0xfc, 0xcc, 0x87, 0xf6, 0x82, 0xdb, 0x7f, 0x04, 0x38, 0xbe, 
+    0x87, 0xee, 0xae, 0x59, 0xc0, 0x5a, 0x3c, 0x5d, 0xc8, 0x56, 
+    0x3c, 0x5d, 0xc8, 0x5e, 0xc3, 0xa2, 0x37, 0xa1, 0x00, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0xc1, 0x7c, 0xa3, 0x86, 0x61, 0x37, 0x8d, 0xb0, 0xf0, 0xb9, 
+    0x6b, 0x2c, 0x31, 0xfa, 0xdf, 0xe6, 0x75, 0xbd, 0x0e, 0x31, 
+    0xb9, 0xb9, 0x9b, 0x20, 0xf1, 0xb9, 0x9f, 0x20, 0x0e, 0x46, 
+    0x60, 0xdf, 0xd2, 0x5e, 0xd3, 0xa5, 0xd5, 0xf2, 0xc8, 0xef, 
+    0xb3, 0x18, 0x09, 0x91, 0xaf, 0x4b, 0x40, 0x3b, 0x7a, 0xd1, 
+    0x73, 0xfb, 0x6e, 0x19, 0xf5, 0xe1, 0x7e, 0x19, 0xf5, 0xe1, 
+    0x81, 0xe6, 0x0a, 0x1e, 0x2a, 0x1e, 0x4e, 0x81, 0x9c, 0xf4, 
+    0x27, 0x36, 0x19, 0xba, 0xea, 0x28, 0x92, 0x1e, 0x34, 0x6b, 
+    0xfb, 0x5b, 0x58, 0x2e, 0x62, 0xaa, 0x3a, 0x34, 0x62, 0xaa, 
+    0x38, 0x34, 0x9d, 0x55, 0xc7, 0xcb, 0x89, 0x87, 0xa7, 0x3b, 
+    0xd6, 0x28, 0x54, 0xee, 0x3d, 0x26, 0x3d, 0x67, 0x47, 0x88, 
+    0xe6, 0x9a, 0xf4, 0x5a, 0x1a, 0x10, 0x27, 0xc0, 0x16, 0x9a, 
+    0x27, 0x48, 0x16, 0x9a, 0xd8, 0xb7, 0xe9, 0x65, 0x82, 0x75, 
+    0x6c, 0xad, 0xd8, 0xfe, 0x32, 0x42, 0x39, 0x26, 0x17, 0x9b, 
+    0x35, 0xf4, 0x6d, 0x26, 0x45, 0x9f, 0x6b, 0x5c, 0x50, 0x1d, 
+    0x22, 0x55, 0x50, 0x1d, 0x62, 0x45, 0xaf, 0xe2, 0x9d, 0xba, 
+    0x1d, 0x4e, 0xf8, 0xcb, 0x37, 0x49, 0x44, 0xcd, 0x29, 0xe3, 
+    0xf8, 0xfd, 0xa3, 0x23, 0x64, 0x27, 0x05, 0x49, 0xee, 0x3e, 
+    0xa5, 0x26, 0xec, 0x5f, 0xa5, 0x26, 0xec, 0x5f, 0x5a, 0xd9, 
+    0x13, 0xa0, 0xcb, 0xf5, 0x5a, 0xd2, 0x63, 0xec, 0xfb, 0xba, 
+    0x79, 0x40, 0x19, 0x76, 0xcc, 0xed, 0x49, 0x94, 0x2f, 0x43, 
+    0x01, 0x61, 0xa5, 0x45, 0x27, 0x76, 0xa5, 0x45, 0x07, 0x72, 
+    0x5a, 0xba, 0xf8, 0x8d, 0xdc, 0x78, 0x48, 0xb6, 0x79, 0xf7, 
+    0x3f, 0x87, 0x1c, 0xb1, 0x2c, 0xc6, 0xef, 0x6c, 0x62, 0x37, 
+    0x05, 0x08, 0xe7, 0xcc, 0xe5, 0x08, 0xae, 0x81, 0xc5, 0x08, 
+    0xae, 0x84, 0x3a, 0xf7, 0x51, 0x7b, 0x00, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x6a, 0x83, 
+    0xce, 0x4a, 0xa5, 0x2a, 0xf8, 0xbb, 0x6d, 0x45, 0x91, 0x4d, 
+    0x62, 0x8a, 0xd2, 0xb0, 0x02, 0xbc, 0x85, 0xc4, 0x53, 0x0a, 
+    0x16, 0x59, 0x53, 0x08, 0x96, 0x59, 0xac, 0xf7, 0x69, 0xa6, 
+    0xe9, 0x7e, 0xd1, 0x6c, 0x31, 0x53, 0x3f, 0xf9, 0xdd, 0x63, 
+    0x8c, 0x31, 0x2f, 0xb4, 0xaf, 0xe2, 0x85, 0x43, 0x81, 0x3e, 
+    0xd5, 0x66, 0x29, 0x6a, 0xd5, 0x66, 0x29, 0x6a, 0x2a, 0x99, 
+    0xd6, 0x95, 0x01, 0xea, 0x7a, 0x41, 0x90, 0x79, 0xdd, 0x28, 
+    0x88, 0x12, 0x58, 0x28, 0x7f, 0x3c, 0x30, 0x76, 0xdd, 0x22, 
+    0x71, 0x46, 0x5a, 0x22, 0xb3, 0x74, 0x5e, 0x22, 0x31, 0x64, 
+    0xa1, 0xdd, 0xce, 0x9b, 0xc9, 0x34, 0x49, 0xe9, 0x23, 0x17, 
+    0xa7, 0xe5, 0xe2, 0x03, 0x60, 0xb3, 0x65, 0x24, 0x24, 0xf8, 
+    0x79, 0xeb, 0xa9, 0xf0, 0x08, 0xe3, 0x05, 0xfa, 0x48, 0xe3, 
+    0x05, 0xf8, 0xb7, 0x1c, 0xfa, 0x07, 0xb7, 0x5f, 0x5f, 0x26, 
+    0x52, 0xde, 0x36, 0x18, 0x87, 0x3d, 0x4d, 0xc4, 0x84, 0x61, 
+    0x2e, 0x5b, 0xfa, 0x02, 0x43, 0xae, 0x1e, 0xa7, 0x0a, 0xa6, 
+    0x9e, 0xa7, 0x0a, 0xa4, 0x61, 0x58, 0xf5, 0x5b, 0xce, 0x40, 
+    0xf0, 0xe3, 0x30, 0xb1, 0x0d, 0xca, 0xdc, 0x63, 0x69, 0x1a, 
+    0xf1, 0x3c, 0x88, 0x5a, 0xd1, 0xd1, 0xde, 0x33, 0xa0, 0x71, 
+    0xac, 0x82, 0xb0, 0x71, 0x8c, 0x82, 0x4f, 0x8e, 0x73, 0x7d, 
+    0x15, 0x7e, 0x14, 0x61, 0x32, 0x62, 0x14, 0x08, 0x5b, 0x91, 
+    0x64, 0x3b, 0xdc, 0x7c, 0x64, 0x6c, 0x09, 0x92, 0xf0, 0xda, 
+    0x2d, 0x56, 0x64, 0xfc, 0x0d, 0x56, 0x66, 0xfc, 0xf2, 0xa9, 
+    0x9b, 0x03, 0x1e, 0x93, 0xf3, 0x4c, 0xd4, 0x2c, 0x98, 0xa2, 
+    0x8f, 0xfd, 0x9a, 0xd3, 0xc8, 0xe7, 0x9f, 0x9e, 0xb1, 0x14, 
+    0x6b, 0x98, 0x21, 0x14, 0x0b, 0x08, 0x31, 0x14, 0x0b, 0x08, 
+    0xce, 0xeb, 0xf4, 0xf7, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0xb2, 0x73, 0xc5, 0x06, 
+    0xe6, 0xce, 0xd2, 0xe7, 0xd5, 0x7d, 0xf1, 0x55, 0xcc, 0xd9, 
+    0xf2, 0x91, 0xeb, 0x6e, 0x04, 0x02, 0x80, 0x64, 0xf3, 0x52, 
+    0xc0, 0x64, 0xf3, 0x12, 0x3f, 0x9b, 0x0c, 0xed, 0x03, 0x29, 
+    0xfb, 0x5d, 0xa6, 0x30, 0xd7, 0x96, 0xb2, 0xa5, 0xfa, 0x9d, 
+    0x80, 0x2c, 0xef, 0xed, 0xee, 0x93, 0x9c, 0x89, 0xa6, 0xf7, 
+    0x95, 0x9b, 0xa6, 0xb7, 0x95, 0x8b, 0x59, 0x48, 0x6a, 0x74, 
+    0x89, 0x28, 0x26, 0xa3, 0x73, 0xd3, 0xf4, 0xd6, 0x55, 0xce, 
+    0x4b, 0xcb, 0x6a, 0x2e, 0x64, 0x64, 0x16, 0x4e, 0xb1, 0xd0, 
+    0xb6, 0xed, 0xba, 0xdc, 0xb6, 0xee, 0xba, 0xdc, 0x49, 0x11, 
+    0x45, 0x23, 0x4c, 0x0c, 0x8e, 0x70, 0x70, 0x4a, 0x53, 0x5e, 
+    0x52, 0xeb, 0x38, 0xf2, 0xb9, 0x4f, 0x91, 0xee, 0x4a, 0xff, 
+    0x4a, 0xa8, 0x8c, 0xee, 0x80, 0xea, 0x0c, 0xee, 0x80, 0xea, 
+    0xf3, 0x11, 0x7f, 0x15, 0xbd, 0x66, 0xe5, 0x4a, 0xf7, 0x6d, 
+    0xa2, 0xbe, 0xfe, 0x87, 0xc1, 0xc9, 0x24, 0xa4, 0x7e, 0xb3, 
+    0x68, 0x06, 0xda, 0xad, 0xa0, 0x17, 0x52, 0xa9, 0xa4, 0x06, 
+    0x52, 0xa9, 0x5b, 0xf9, 0xad, 0x56, 0xad, 0xe7, 0x3e, 0x5e, 
+    0x0b, 0x9c, 0xcf, 0x74, 0xb4, 0x32, 0x8e, 0x88, 0xb6, 0x76, 
+    0x89, 0xf5, 0x85, 0x7f, 0x0a, 0x0f, 0x14, 0xb2, 0x29, 0x5f, 
+    0x14, 0xb6, 0x29, 0x5f, 0xeb, 0x49, 0xd6, 0xa0, 0x1f, 0x48, 
+    0x4b, 0xac, 0x49, 0x5b, 0xcb, 0x12, 0x6d, 0xc2, 0xda, 0x96, 
+    0x7c, 0xda, 0xfb, 0xbe, 0xd6, 0xf8, 0xaa, 0xfe, 0x53, 0x7a, 
+    0xe2, 0xfe, 0x5b, 0xfa, 0xea, 0xfe, 0xa4, 0x05, 0x15, 0x01, 
+    0x9a, 0xc6, 0xf6, 0x49, 0x9c, 0x36, 0x7c, 0xe7, 0xc6, 0x33, 
+    0x73, 0xd3, 0xf4, 0x03, 0xe3, 0x48, 0x56, 0x5d, 0x55, 0x1f, 
+    0x54, 0x13, 0x8f, 0xbe, 0x54, 0x13, 0x87, 0x9e, 0xab, 0xec, 
+    0x78, 0x61, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x40, 0x6a, 0xf8, 0x06, 0xc9, 0xa9, 
+    0xdd, 0x4b, 0x2f, 0x50, 0x9c, 0x1a, 0xeb, 0xb6, 0x24, 0xa9, 
+    0x73, 0xa2, 0xb0, 0x68, 0xbd, 0x82, 0xbc, 0x2c, 0xbf, 0x82, 
+    0xbc, 0x2c, 0x40, 0x7d, 0x43, 0xd3, 0xac, 0xfb, 0x7d, 0x7c, 
+    0x66, 0x6e, 0xe5, 0x0c, 0x97, 0xc8, 0x05, 0x2b, 0xcd, 0x02, 
+    0xdf, 0xf0, 0xcc, 0x43, 0x82, 0xbc, 0x89, 0x4a, 0xe7, 0xb1, 
+    0x8d, 0x4a, 0xc7, 0xb1, 0x72, 0xb5, 0x38, 0x4e, 0x08, 0x40, 
+    0x79, 0x74, 0x29, 0x9f, 0xbc, 0xb2, 0x4f, 0x81, 0x63, 0x49, 
+    0x1f, 0xb6, 0x70, 0x5b, 0x88, 0xe2, 0x47, 0x8e, 0xe8, 0xe0, 
+    0x04, 0xa2, 0xec, 0xe0, 0x04, 0xa2, 0x13, 0x1f, 0xfb, 0x5d, 
+    0x5e, 0xe2, 0x0d, 0x6e, 0x8f, 0x54, 0x00, 0x39, 0x79, 0x88, 
+    0x5f, 0x96, 0xa3, 0x30, 0xaa, 0x8b, 0x11, 0x56, 0xba, 0xf5, 
+    0x20, 0xe6, 0x3a, 0x4e, 0x20, 0xf6, 0xba, 0x4f, 0xdf, 0x09, 
+    0x45, 0xb0, 0x61, 0x3a, 0x89, 0x85, 0x99, 0x15, 0x41, 0x4b, 
+    0xea, 0x2f, 0xa1, 0x35, 0xba, 0x23, 0xe2, 0xd8, 0x0a, 0x43, 
+    0xa9, 0xc6, 0x68, 0x21, 0xa9, 0xd4, 0x6a, 0x23, 0xa9, 0xd0, 
+    0x95, 0xdc, 0x56, 0x2f, 0x22, 0x82, 0x8d, 0xa1, 0x5f, 0xd8, 
+    0xc0, 0x9f, 0x61, 0x63, 0x1b, 0xe4, 0x7c, 0x40, 0xa1, 0x32, 
+    0xbf, 0x0f, 0x79, 0x56, 0x3f, 0x4e, 0x19, 0x22, 0x3f, 0x4e, 
+    0x19, 0x26, 0xc0, 0xb1, 0xe6, 0xd9, 0x73, 0x1c, 0x55, 0xa0, 
+    0x61, 0xdd, 0xe3, 0x4b, 0x97, 0xb1, 0xe1, 0x31, 0x26, 0xd5, 
+    0xfb, 0xaa, 0x86, 0x65, 0x41, 0xab, 0x82, 0xcc, 0xc2, 0x3b, 
+    0x82, 0xcc, 0x83, 0x3b, 0x7d, 0x33, 0x3c, 0xc4, 0x81, 0xa5, 
+    0xf5, 0xe1, 0x9a, 0x9d, 0xba, 0x62, 0x10, 0xf7, 0xcd, 0x1f, 
+    0x63, 0x15, 0x35, 0x79, 0x23, 0x09, 0xf8, 0x18, 0x69, 0x59, 
+    0xfc, 0x58, 0x63, 0x59, 0xfc, 0x58, 0x9c, 0xa6, 0x03, 0xa7, 
+    0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x04, 0x3b, 0x3b, 0xd4, 0x29, 0xf6, 0x00, 0x55, 
+    0x52, 0xb9, 0x1d, 0x70, 0x46, 0x8d, 0x66, 0xdd, 0x55, 0xab, 
+    0xa6, 0x7f, 0x50, 0xb1, 0xee, 0xff, 0x10, 0xa1, 0xee, 0xff, 
+    0xef, 0x5e, 0x11, 0x00, 0x76, 0x2a, 0x8b, 0x86, 0x12, 0x2c, 
+    0xb8, 0x77, 0x7a, 0xc5, 0x6d, 0x83, 0x79, 0x58, 0x39, 0x84, 
+    0x86, 0xc4, 0x99, 0x4e, 0x3c, 0xb4, 0xd4, 0x40, 0x3c, 0xb4, 
+    0x94, 0x40, 0xc3, 0x4b, 0x6b, 0xbf, 0x01, 0xe4, 0x14, 0x40, 
+    0x64, 0xa1, 0x03, 0x58, 0xa7, 0x78, 0x65, 0x9b, 0x20, 0xd0, 
+    0x0e, 0x64, 0xd5, 0xdf, 0x5f, 0x88, 0x65, 0xfe, 0x6b, 0xa5, 
+    0x65, 0xfe, 0x6f, 0xa5, 0x9a, 0x01, 0x90, 0x5a, 0x44, 0x37, 
+    0xeb, 0xcf, 0xf0, 0x23, 0x0a, 0x0c, 0x55, 0xce, 0x91, 0x22, 
+    0xf7, 0xdd, 0xc8, 0x4e, 0xa7, 0xfa, 0x08, 0xaf, 0x9f, 0xd5, 
+    0x48, 0xd0, 0xb7, 0xdd, 0x48, 0xdd, 0x48, 0x22, 0xb7, 0x22, 
+    0xf6, 0x4d, 0x6d, 0x15, 0x05, 0xdb, 0x34, 0xb2, 0xa6, 0x73, 
+    0xa1, 0x01, 0x0c, 0x6e, 0x35, 0xd6, 0x58, 0x18, 0x80, 0x9c, 
+    0x4a, 0x08, 0x08, 0xb4, 0x48, 0x08, 0x08, 0xb4, 0xb7, 0xf7, 
+    0xf7, 0x4b, 0xd8, 0x1a, 0xc3, 0x7b, 0xd2, 0x74, 0x64, 0x21, 
+    0x0b, 0xc6, 0xd6, 0xb0, 0x68, 0x2a, 0xf3, 0xab, 0x74, 0x17, 
+    0x59, 0x69, 0x7c, 0x12, 0x88, 0xeb, 0x7c, 0x12, 0x98, 0xeb, 
+    0x83, 0xed, 0x67, 0x14, 0xd8, 0x4e, 0xad, 0x85, 0x03, 0xda, 
+    0x33, 0x63, 0xf3, 0x94, 0xdb, 0x23, 0xe1, 0x5d, 0x5f, 0x21, 
+    0x7b, 0xfd, 0x27, 0x1d, 0xec, 0x95, 0x1f, 0x53, 0xf9, 0x95, 
+    0x1f, 0x43, 0x06, 0x6a, 0xe0, 0xbc, 0x41, 0x9e, 0x3d, 0x77, 
+    0xc5, 0x9e, 0x5d, 0xba, 0x2b, 0x53, 0x0d, 0xae, 0xda, 0x92, 
+    0x81, 0x9c, 0x8a, 0xd6, 0x4b, 0xcb, 0xc6, 0x99, 0x46, 0xac, 
+    0xc6, 0x9d, 0x47, 0xac, 0x39, 0x62, 0xb8, 0x53, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x82, 0x87, 0x45, 0x45, 0x86, 0xdc, 0xb9, 0xe2, 0xca, 0x9b, 
+    0x84, 0xbb, 0x09, 0x78, 0x1b, 0xcf, 0x21, 0x2d, 0x2f, 0x53, 
+    0x85, 0xbe, 0x3e, 0xfb, 0x81, 0xbd, 0x3e, 0xfb, 0x7e, 0x42, 
+    0xc1, 0x04, 0xbb, 0x16, 0xd2, 0x57, 0x65, 0xb9, 0xde, 0x74, 
+    0x57, 0x17, 0xea, 0x24, 0x58, 0xf8, 0xb5, 0x62, 0x82, 0x5b, 
+    0x05, 0xbd, 0x97, 0x47, 0x64, 0x21, 0x93, 0x53, 0xe0, 0x21, 
+    0x6c, 0xac, 0x1f, 0xde, 0xee, 0x00, 0x50, 0x14, 0x63, 0xc9, 
+    0x80, 0xb5, 0xb0, 0x6c, 0x69, 0xfb, 0xcb, 0x54, 0xb3, 0x52, 
+    0x28, 0xa1, 0xde, 0x66, 0x32, 0x71, 0xa0, 0x52, 0x22, 0x71, 
+    0xa0, 0x52, 0xdd, 0x8e, 0x5f, 0xad, 0xae, 0x05, 0xf3, 0x7e, 
+    0x6e, 0x99, 0x69, 0xfa, 0x86, 0x04, 0xe6, 0x7f, 0xe1, 0xd1, 
+    0xdc, 0x00, 0x97, 0xd8, 0x6d, 0x5f, 0xc5, 0x5f, 0x48, 0x5f, 
+    0x85, 0x5d, 0x48, 0x5f, 0x7a, 0xa2, 0xb7, 0xa0, 0xb9, 0x99, 
+    0xd9, 0x4d, 0xbe, 0x56, 0xd4, 0xca, 0x76, 0xad, 0xf8, 0x9e, 
+    0xc1, 0xdd, 0x99, 0x0f, 0x7d, 0xb7, 0x3b, 0xa6, 0x75, 0x3f, 
+    0x7e, 0xb7, 0x75, 0xbf, 0x7b, 0xb7, 0x8a, 0x40, 0x84, 0x48, 
+    0x6f, 0x69, 0x1f, 0x7b, 0x05, 0xa6, 0xe5, 0xe7, 0xff, 0x7d, 
+    0x1a, 0x65, 0x1b, 0x25, 0x26, 0xee, 0xce, 0xb0, 0x18, 0x5a, 
+    0xde, 0xb1, 0x39, 0xc5, 0xde, 0xb1, 0x1b, 0xcc, 0x21, 0x4e, 
+    0xe4, 0x33, 0xcd, 0xac, 0x03, 0xb1, 0xef, 0xa5, 0xb5, 0x90, 
+    0x7b, 0xfb, 0x12, 0xf4, 0x7b, 0x35, 0xf6, 0x33, 0x97, 0xb3, 
+    0x55, 0x29, 0x61, 0xb3, 0x55, 0x31, 0x61, 0xb1, 0x55, 0x31, 
+    0x9e, 0x4e, 0xaa, 0xce, 0xf5, 0x1b, 0xc8, 0xed, 0x32, 0x8b, 
+    0xa8, 0x53, 0x22, 0x96, 0x56, 0x3a, 0x40, 0x49, 0x7e, 0x5e, 
+    0xb5, 0x20, 0x0f, 0x18, 0x86, 0x50, 0x2a, 0x2a, 0x86, 0x50, 
+    0x2a, 0x3a, 0x79, 0xaf, 0xd5, 0xc5, 0x00, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x9d, 0xe1, 
+    0x7a, 0x57, 0x02, 0xc3, 0x96, 0x40, 0xd8, 0x31, 0x97, 0x0e, 
+    0x76, 0x69, 0xd4, 0xa8, 0x12, 0x56, 0x5b, 0xad, 0x30, 0x74, 
+    0x1c, 0x8a, 0x30, 0x74, 0x1d, 0x88, 0xcf, 0x8b, 0xe2, 0x77, 
+    0x77, 0xa3, 0x56, 0x23, 0x21, 0x03, 0xc4, 0xa9, 0x27, 0xdd, 
+    0x3d, 0x7e, 0x29, 0xa5, 0xca, 0xc6, 0x84, 0x4a, 0xda, 0x5a, 
+    0xa4, 0xe4, 0xd6, 0xca, 0xa4, 0xe4, 0xd6, 0xca, 0x5b, 0x1b, 
+    0x29, 0x35, 0xd3, 0x77, 0x5a, 0x14, 0x6a, 0x59, 0xdc, 0xaf, 
+    0x98, 0xda, 0x97, 0x10, 0x8b, 0xa1, 0x8c, 0xf0, 0xe1, 0x38, 
+    0xaf, 0xa4, 0x81, 0xa7, 0xce, 0xe4, 0x81, 0xa6, 0x8e, 0xe4, 
+    0x7e, 0x59, 0x71, 0x1b, 0x03, 0x43, 0x26, 0x3e, 0x29, 0xd6, 
+    0xff, 0x60, 0x8b, 0x94, 0x72, 0x6c, 0x6c, 0xbe, 0x62, 0xfc, 
+    0x51, 0xa6, 0x3d, 0x62, 0x48, 0xab, 0x2d, 0xfe, 0x48, 0xab, 
+    0x2d, 0xfe, 0xb7, 0x54, 0xd2, 0x01, 0x2d, 0xa6, 0xdd, 0xf0, 
+    0x2a, 0x9a, 0xe3, 0x81, 0x8e, 0xe1, 0xf4, 0x42, 0x21, 0x10, 
+    0x66, 0x6b, 0xdd, 0xb5, 0x8e, 0x4c, 0x0e, 0x51, 0xcf, 0x72, 
+    0x0c, 0x71, 0xce, 0x72, 0xf3, 0x8e, 0x31, 0x8d, 0x71, 0x03, 
+    0xba, 0xc5, 0xc6, 0x42, 0xfe, 0xd4, 0xcc, 0x04, 0xb4, 0xd3, 
+    0x12, 0x99, 0xc5, 0x99, 0x58, 0x45, 0x3f, 0xe8, 0x99, 0xd9, 
+    0x1e, 0x59, 0x99, 0xcd, 0x9e, 0xd9, 0x66, 0x32, 0x61, 0x26, 
+    0x55, 0x01, 0xf8, 0xf1, 0xe3, 0x4e, 0xf6, 0x6d, 0x4e, 0x27, 
+    0xdf, 0x75, 0xb9, 0x69, 0x06, 0x99, 0x70, 0x84, 0x4a, 0x6a, 
+    0x36, 0xa6, 0x43, 0x6c, 0x34, 0x84, 0x42, 0x6c, 0xcb, 0x7b, 
+    0xbd, 0x93, 0x9e, 0x8f, 0xcb, 0xd2, 0xf9, 0x3a, 0xe5, 0x16, 
+    0x75, 0x34, 0x7b, 0xd0, 0x46, 0x4e, 0xfc, 0x82, 0xea, 0xab, 
+    0x23, 0x01, 0x4a, 0x39, 0x06, 0x02, 0x4a, 0x39, 0x26, 0x02, 
+    0xb5, 0xc6, 0xd9, 0xfd, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x55, 0xfe, 0x07, 0xbc, 
+    0x1a, 0xd1, 0x1d, 0x62, 0xb7, 0xcb, 0xfb, 0xf2, 0x3a, 0xc4, 
+    0xa8, 0xb2, 0x90, 0x25, 0x12, 0x25, 0x9e, 0xb4, 0x92, 0x69, 
+    0x9e, 0xa5, 0x92, 0x69, 0x61, 0x5a, 0x6d, 0x96, 0x85, 0x2c, 
+    0xb7, 0x43, 0xbb, 0x58, 0x0f, 0xf6, 0x49, 0x10, 0x36, 0x28, 
+    0xc0, 0xde, 0xec, 0x29, 0x48, 0x46, 0xe7, 0x3f, 0x0c, 0x57, 
+    0xa6, 0x3b, 0x08, 0x57, 0xa6, 0x3b, 0xf7, 0xa8, 0x59, 0xc4, 
+    0x56, 0x88, 0x07, 0x77, 0x0d, 0x50, 0x41, 0x93, 0xa3, 0x6b, 
+    0x26, 0x85, 0x6d, 0xd8, 0xdd, 0x7b, 0x00, 0xd0, 0xa7, 0x19, 
+    0x8c, 0xa8, 0xc7, 0x89, 0x88, 0xa8, 0xc7, 0x89, 0x77, 0x57, 
+    0x38, 0x76, 0x7a, 0x99, 0xee, 0x29, 0x21, 0x04, 0x71, 0x8f, 
+    0x8a, 0x10, 0x3f, 0x6b, 0x7a, 0xe4, 0x8d, 0x74, 0x4f, 0x99, 
+    0xa1, 0x71, 0x37, 0xcc, 0x1d, 0x60, 0x37, 0xcc, 0x9d, 0x60, 
+    0xc8, 0x33, 0x62, 0x9f, 0x8c, 0x6c, 0xed, 0xb5, 0xa1, 0xb0, 
+    0x6a, 0x75, 0xb2, 0x79, 0xac, 0xba, 0x4a, 0x50, 0x9c, 0x08, 
+    0x34, 0x95, 0x21, 0x0c, 0x70, 0x99, 0xed, 0x8b, 0x70, 0x99, 
+    0xed, 0x8b, 0x8f, 0x66, 0x12, 0x74, 0xd1, 0x4f, 0x29, 0xd6, 
+    0x91, 0x18, 0xfb, 0x98, 0xc8, 0xfa, 0x08, 0x4c, 0xd9, 0x5f, 
+    0x17, 0x7e, 0x2d, 0x81, 0x9b, 0x9c, 0x9c, 0xa3, 0x13, 0x3d, 
+    0x9c, 0xa3, 0x13, 0x3d, 0x63, 0x5c, 0xec, 0xc2, 0x7c, 0x6f, 
+    0x6c, 0x95, 0xa4, 0xcf, 0xec, 0x02, 0x94, 0xa6, 0x86, 0xa3, 
+    0x3f, 0x69, 0x96, 0x3d, 0x1a, 0xdd, 0xd0, 0xf1, 0x10, 0x84, 
+    0xce, 0x21, 0x10, 0xc5, 0xce, 0x21, 0xef, 0x3a, 0x31, 0xde, 
+    0x3f, 0xda, 0x87, 0x1f, 0x1e, 0xb6, 0x69, 0x34, 0x95, 0x1c, 
+    0xd3, 0x67, 0xd6, 0xa0, 0x25, 0x98, 0x26, 0x1a, 0x4c, 0xb2, 
+    0x02, 0x2a, 0x1f, 0xae, 0x06, 0x2a, 0x1d, 0xaa, 0xf9, 0xd5, 
+    0xe2, 0x55, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x89, 0x46, 0x8b, 0x02, 0xaf, 0x78, 
+    0x3d, 0x9f, 0x52, 0xf3, 0xa6, 0x4a, 0xbd, 0xc3, 0x39, 0x7f, 
+    0xc6, 0x6c, 0xd2, 0x95, 0xc8, 0x66, 0xf6, 0xfd, 0xc8, 0x66, 
+    0xf6, 0xfd, 0x37, 0x99, 0x09, 0x02, 0xbe, 0x4b, 0xf3, 0xd7, 
+    0x43, 0x58, 0x80, 0xf3, 0x0a, 0xeb, 0xd9, 0x24, 0x0e, 0x40, 
+    0x2b, 0x68, 0x93, 0x12, 0x7f, 0xf6, 0x96, 0x19, 0x2f, 0xd5, 
+    0x96, 0x13, 0x2d, 0xf6, 0x69, 0xec, 0xd2, 0x09, 0xea, 0xb8, 
+    0x07, 0x8e, 0x3d, 0xea, 0x71, 0x6b, 0x94, 0x2e, 0xf6, 0x1c, 
+    0xaf, 0x90, 0xd5, 0xd7, 0x15, 0x78, 0x1c, 0x9d, 0xa5, 0x18, 
+    0x1e, 0xc9, 0xa5, 0x18, 0x1c, 0xcd, 0x5a, 0xe7, 0xe3, 0x32, 
+    0xaa, 0x55, 0x62, 0x05, 0x0d, 0x14, 0xa6, 0x2a, 0x55, 0xbb, 
+    0x99, 0x0e, 0x82, 0x86, 0xe7, 0x4a, 0xf7, 0x92, 0xe9, 0x4b, 
+    0xa7, 0x71, 0xa9, 0x4b, 0xb7, 0x90, 0xa9, 0x4b, 0x48, 0x6f, 
+    0x56, 0xb4, 0xe2, 0x3a, 0x02, 0x50, 0xb7, 0xed, 0x73, 0xf0, 
+    0xee, 0x20, 0x1f, 0xf7, 0x41, 0x9b, 0xbb, 0x16, 0x96, 0x81, 
+    0xe1, 0x5b, 0xe3, 0x82, 0x82, 0x5e, 0x43, 0x82, 0x83, 0x5e, 
+    0xbc, 0x7d, 0x7c, 0xa1, 0xb3, 0x34, 0xda, 0x07, 0x6c, 0xdf, 
+    0xc0, 0x69, 0x43, 0x40, 0xf8, 0x39, 0x54, 0xd2, 0x27, 0xaf, 
+    0x11, 0x38, 0x3c, 0x06, 0xe0, 0x18, 0xb4, 0xae, 0xc0, 0x18, 
+    0x34, 0xae, 0x3f, 0xe7, 0xcb, 0x51, 0xe7, 0xca, 0x17, 0x92, 
+    0xd6, 0x51, 0x00, 0x27, 0xef, 0x54, 0x52, 0x97, 0xc3, 0xac, 
+    0x98, 0x5b, 0xf0, 0x1f, 0xb9, 0x88, 0xfd, 0x76, 0xd1, 0xa3, 
+    0xdd, 0x56, 0xd1, 0xa3, 0x22, 0xa9, 0x2e, 0x5c, 0xe6, 0x3b, 
+    0x8a, 0xeb, 0x1e, 0x1b, 0x31, 0xf4, 0xd1, 0x84, 0x2b, 0x09, 
+    0xbf, 0x71, 0xab, 0x4c, 0x86, 0x04, 0x64, 0x1b, 0x96, 0x30, 
+    0xe5, 0x28, 0x86, 0x30, 0xe1, 0x28, 0x79, 0xcf, 0x1e, 0xd7, 
+    0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0xad, 0x70, 0x07, 0x15, 0x41, 0xe2, 0xda, 0xf8, 
+    0x5f, 0xfa, 0x29, 0x0b, 0x74, 0xd5, 0x70, 0x2e, 0x25, 0xac, 
+    0xc1, 0xbe, 0x43, 0x9b, 0xc0, 0x84, 0x47, 0x99, 0xc0, 0x84, 
+    0xb8, 0x66, 0x3f, 0x7b, 0x86, 0xe6, 0x84, 0x37, 0x13, 0xf4, 
+    0x73, 0x2b, 0x32, 0x37, 0xa9, 0x86, 0x87, 0x2f, 0x4e, 0xd1, 
+    0x53, 0x3d, 0xc6, 0x78, 0x53, 0xb7, 0x77, 0x7a, 0x53, 0x37, 
+    0x77, 0x78, 0xac, 0xc8, 0x88, 0x87, 0x2c, 0x1e, 0xe1, 0xfa, 
+    0xb6, 0xbb, 0x93, 0x32, 0x33, 0xe2, 0x1a, 0xe0, 0x14, 0x80, 
+    0xe5, 0xa4, 0x1d, 0x92, 0x2a, 0x11, 0x5c, 0xba, 0x49, 0x29, 
+    0x5c, 0xb2, 0x49, 0x29, 0xa3, 0x4d, 0xb6, 0xd6, 0x04, 0xd3, 
+    0x44, 0x0e, 0x03, 0x98, 0x01, 0x01, 0xec, 0xf0, 0xf4, 0xb8, 
+    0xc3, 0x81, 0x00, 0x66, 0xbc, 0x34, 0x46, 0xcf, 0x51, 0x01, 
+    0x17, 0x0f, 0xd1, 0x21, 0x16, 0x0f, 0x2e, 0xde, 0xe9, 0xf0, 
+    0x8e, 0x46, 0x50, 0x4f, 0x38, 0x65, 0x3c, 0xf8, 0x30, 0xc5, 
+    0x6d, 0x04, 0x18, 0x23, 0x95, 0xb4, 0xb8, 0x9d, 0x9e, 0x9c, 
+    0x7a, 0xad, 0x94, 0x5b, 0x7a, 0xad, 0x94, 0x99, 0x85, 0x52, 
+    0x6b, 0x66, 0x5d, 0x7d, 0xbe, 0x84, 0xdf, 0xa9, 0xa8, 0xef, 
+    0xd6, 0x05, 0x06, 0x34, 0xcc, 0x56, 0x9f, 0x5c, 0x76, 0x96, 
+    0x25, 0x99, 0x3e, 0x5d, 0xb7, 0x94, 0x3e, 0x5c, 0x37, 0x9c, 
+    0xc1, 0xa3, 0xc8, 0x63, 0x84, 0x7c, 0x30, 0x1e, 0x2d, 0x81, 
+    0x43, 0x76, 0xb9, 0xdf, 0x6a, 0x4b, 0x53, 0xd7, 0xb4, 0x83, 
+    0x11, 0xfb, 0x1b, 0xe5, 0x0b, 0xe3, 0xe3, 0x61, 0x0b, 0xe3, 
+    0x73, 0x65, 0xf4, 0x1c, 0x8c, 0x9a, 0x42, 0xb0, 0x81, 0xf5, 
+    0x3f, 0xd6, 0x5e, 0x2b, 0x09, 0xd1, 0x68, 0x78, 0xf6, 0x7c, 
+    0x34, 0x97, 0xb3, 0x72, 0xf5, 0x7b, 0x82, 0xb8, 0x7c, 0x1d, 
+    0x82, 0xb8, 0x7c, 0x5d, 0x7d, 0x47, 0x83, 0xa2, 0x00, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x42, 0x78, 0x38, 0x30, 0xe9, 0x08, 0x72, 0x8b, 0x90, 0xd1, 
+    0x9b, 0xa8, 0x8d, 0xe8, 0x21, 0xd0, 0x7a, 0xf5, 0x97, 0xe7, 
+    0xfc, 0xf1, 0x37, 0x39, 0xfc, 0xf1, 0x37, 0x39, 0x03, 0x0e, 
+    0xc8, 0xc6, 0xa8, 0x3a, 0x07, 0x1f, 0xc9, 0x89, 0x53, 0xa3, 
+    0x67, 0x07, 0xe8, 0xd0, 0x18, 0x6d, 0x74, 0x14, 0xcf, 0xbd, 
+    0xd8, 0x4a, 0x27, 0x2e, 0xd8, 0x0e, 0x27, 0x2e, 0xd8, 0x0e, 
+    0xd8, 0xd1, 0x27, 0xf1, 0x20, 0x4d, 0x61, 0xe8, 0xb8, 0x12, 
+    0x70, 0xe7, 0x17, 0x55, 0x50, 0x4b, 0x5c, 0x9c, 0x4e, 0xd7, 
+    0xe5, 0x82, 0x8e, 0x6c, 0xce, 0x22, 0x5e, 0x13, 0xc4, 0x02, 
+    0x5e, 0x53, 0x3b, 0xfd, 0xa1, 0xac, 0x83, 0xaf, 0x89, 0xf6, 
+    0x1d, 0x2f, 0x00, 0x69, 0xe9, 0xc2, 0x12, 0xbe, 0xca, 0x32, 
+    0xbc, 0xd8, 0xcc, 0xb0, 0x18, 0x2f, 0x64, 0xb3, 0x19, 0xa5, 
+    0x6c, 0xb2, 0x19, 0xa5, 0x93, 0x4d, 0xe6, 0x5a, 0xbf, 0x77, 
+    0xce, 0x27, 0xb7, 0xd3, 0x92, 0xf1, 0x5a, 0x95, 0x4e, 0x05, 
+    0x60, 0x79, 0x74, 0xf9, 0x90, 0xf8, 0xb5, 0x6a, 0x16, 0xd9, 
+    0x35, 0x65, 0x16, 0xd9, 0xb5, 0x65, 0xe9, 0x26, 0x4a, 0x9a, 
+    0xa8, 0xac, 0x85, 0xb5, 0xe3, 0x29, 0x9a, 0xb9, 0xbc, 0x19, 
+    0xa6, 0xe1, 0x7e, 0xd0, 0xe6, 0x76, 0xad, 0x7b, 0xeb, 0x09, 
+    0x5d, 0x51, 0xe3, 0x0e, 0xdd, 0x51, 0xe3, 0x0c, 0x22, 0xae, 
+    0x1c, 0xf3, 0x3a, 0x2a, 0xb9, 0x43, 0xfd, 0x01, 0x22, 0x4a, 
+    0x5c, 0xc7, 0xff, 0x62, 0xe6, 0x8c, 0xdd, 0x14, 0x57, 0x22, 
+    0x1c, 0xff, 0x74, 0x07, 0x0d, 0xf1, 0x74, 0x27, 0x1d, 0xf7, 
+    0x8b, 0xd8, 0xe2, 0x08, 0x4a, 0x4e, 0x03, 0xa9, 0xa5, 0xff, 
+    0x80, 0xfa, 0x45, 0x34, 0x7e, 0xde, 0x4c, 0x65, 0x0e, 0x86, 
+    0xae, 0xc1, 0xe9, 0x00, 0x61, 0xc0, 0x42, 0xc2, 0xe1, 0xe0, 
+    0x42, 0xc2, 0x1e, 0x1f, 0xbd, 0x3d, 0x01, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x2d, 0xcf, 
+    0xaf, 0x90, 0x1a, 0xc5, 0x29, 0x00, 0x03, 0x62, 0xfb, 0xdf, 
+    0x4d, 0x4b, 0xdf, 0x21, 0x43, 0xe1, 0x2b, 0x0a, 0x86, 0x49, 
+    0xfe, 0x8c, 0x87, 0x49, 0xfe, 0x8c, 0x78, 0xb6, 0x01, 0x73, 
+    0xf6, 0x3f, 0x6f, 0xd2, 0xe2, 0xf7, 0x78, 0x40, 0xb9, 0xd6, 
+    0xe0, 0xb4, 0x21, 0x29, 0x0e, 0x8c, 0x83, 0xcc, 0xcf, 0xb2, 
+    0xa3, 0xd3, 0x4f, 0xb8, 0xa3, 0xd3, 0x4f, 0xb8, 0x5c, 0x2c, 
+    0xb0, 0x47, 0x42, 0xaf, 0x60, 0x7a, 0x4a, 0x01, 0xf7, 0xea, 
+    0xdb, 0x4f, 0xf9, 0xd6, 0xd9, 0x75, 0xe1, 0x68, 0xcc, 0x03, 
+    0xf5, 0x67, 0xb8, 0x95, 0x34, 0x67, 0xd8, 0x15, 0xf4, 0x67, 
+    0x27, 0xea, 0x0b, 0x98, 0x5c, 0x21, 0x3f, 0x10, 0x7c, 0xf6, 
+    0x39, 0x60, 0x9c, 0x3c, 0xb2, 0x6b, 0x01, 0x2f, 0x0b, 0x8b, 
+    0xd4, 0x24, 0xbf, 0x0b, 0x96, 0x0c, 0xab, 0x03, 0x96, 0x0c, 
+    0xbb, 0x8b, 0x69, 0xf3, 0x44, 0x74, 0xff, 0xec, 0xa7, 0x4e, 
+    0x06, 0x09, 0x49, 0x8a, 0x3f, 0xc6, 0x0f, 0x2e, 0x55, 0x98, 
+    0xd2, 0x44, 0x04, 0xae, 0x2e, 0xa0, 0xd4, 0x86, 0x6e, 0xa8, 
+    0x54, 0x8e, 0x6e, 0xa8, 0xab, 0x71, 0x91, 0x57, 0xeb, 0x8d, 
+    0xe3, 0x92, 0x9d, 0x42, 0xa5, 0x88, 0x53, 0xd1, 0xfe, 0x8b, 
+    0x98, 0xef, 0xe3, 0x0b, 0xf6, 0x41, 0x1f, 0xed, 0x50, 0x4b, 
+    0x9f, 0x5f, 0x50, 0x4b, 0x9f, 0x4d, 0xaf, 0xb4, 0x60, 0xb2, 
+    0x95, 0xa2, 0x74, 0x96, 0xa2, 0x13, 0xc0, 0x3a, 0x16, 0x37, 
+    0x67, 0xb5, 0xa6, 0xde, 0x74, 0x1e, 0x26, 0xba, 0x8d, 0x3c, 
+    0xa4, 0x5f, 0x9f, 0x34, 0xa4, 0x5f, 0x9f, 0x34, 0x5b, 0xa0, 
+    0x60, 0xcb, 0xd6, 0xd7, 0x42, 0xf8, 0x12, 0x3c, 0x4b, 0xe4, 
+    0xca, 0xf0, 0xd2, 0x55, 0x6f, 0xfe, 0xa5, 0x6b, 0x64, 0xfd, 
+    0xc3, 0x1b, 0xe9, 0x77, 0xce, 0x1b, 0xe9, 0x77, 0xce, 0x1b, 
+    0x16, 0x88, 0x31, 0xe4, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x5e, 0xcd, 0xe5, 0x27, 
+    0x6b, 0x60, 0xe6, 0x30, 0xc4, 0xd6, 0xbb, 0xd8, 0xbe, 0x1d, 
+    0x9b, 0xa8, 0xde, 0x19, 0xec, 0x31, 0xfe, 0x9c, 0xe4, 0xa3, 
+    0xfe, 0x1c, 0xe4, 0xa0, 0x01, 0xe3, 0x1b, 0x5f, 0x4e, 0x58, 
+    0xd8, 0x86, 0x9a, 0xa5, 0xc2, 0x00, 0x86, 0xee, 0x79, 0x65, 
+    0x24, 0xa4, 0x73, 0xcd, 0xad, 0xba, 0x27, 0x03, 0x06, 0x39, 
+    0x6b, 0x81, 0x06, 0x39, 0x63, 0x81, 0xf9, 0xc6, 0x9c, 0x7e, 
+    0x9d, 0xc6, 0xb2, 0xe3, 0xa3, 0xcf, 0xc4, 0x58, 0xd8, 0x61, 
+    0x76, 0xaa, 0x78, 0x03, 0x07, 0xef, 0xa1, 0x9b, 0xff, 0xf0, 
+    0xb1, 0x2b, 0x5f, 0xd8, 0xb1, 0x2b, 0x5f, 0xc8, 0x4e, 0xd4, 
+    0xa0, 0x37, 0x77, 0x6a, 0x53, 0x55, 0x2a, 0xbb, 0xf6, 0x31, 
+    0x0e, 0x70, 0x60, 0x94, 0x58, 0x33, 0x4d, 0x49, 0xef, 0xea, 
+    0x71, 0xdf, 0x64, 0x10, 0x12, 0x4b, 0x6e, 0x10, 0x10, 0x4b, 
+    0x91, 0xef, 0xef, 0xb4, 0xb7, 0xdc, 0xd2, 0x98, 0x57, 0x81, 
+    0x3c, 0xd2, 0xe2, 0x43, 0xb8, 0xa2, 0xad, 0xa4, 0xf5, 0x08, 
+    0x19, 0xb7, 0xb2, 0x6d, 0x6b, 0xbc, 0xbb, 0x0e, 0x69, 0xb4, 
+    0xbb, 0x0e, 0x96, 0x4b, 0x44, 0xf1, 0x45, 0x3c, 0x05, 0x69, 
+    0x15, 0x18, 0xf9, 0xec, 0x04, 0x36, 0x72, 0x17, 0x43, 0x4d, 
+    0x2c, 0x88, 0x54, 0x74, 0xdb, 0xf3, 0x60, 0xf4, 0xbd, 0xba, 
+    0x40, 0x74, 0xbd, 0xaa, 0xbf, 0x8b, 0x42, 0x55, 0xff, 0xf8, 
+    0x65, 0xd2, 0x7b, 0x20, 0x2d, 0x3b, 0xa6, 0x4b, 0x66, 0x56, 
+    0xe3, 0xe3, 0x23, 0x87, 0xe0, 0x30, 0x2d, 0xbe, 0x7c, 0x61, 
+    0x2d, 0xcb, 0x7c, 0x61, 0x2d, 0x8b, 0x83, 0x9e, 0xd2, 0x74, 
+    0xb6, 0xc4, 0xb4, 0xf7, 0xf0, 0x9c, 0xe5, 0x40, 0x60, 0x9c, 
+    0x9c, 0x1b, 0xee, 0xe4, 0x8f, 0x88, 0x10, 0xe2, 0x91, 0x31, 
+    0x50, 0xd5, 0x9c, 0x50, 0x50, 0xd5, 0x9c, 0x10, 0xaf, 0x2a, 
+    0x63, 0xef, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0xd0, 0x71, 0xc5, 0x4a, 0x13, 0x27, 
+    0xe6, 0x0e, 0x3d, 0x53, 0x17, 0x33, 0x86, 0xf7, 0x72, 0xd8, 
+    0x96, 0x12, 0xc1, 0x58, 0x52, 0x5a, 0xd1, 0x98, 0x52, 0x5b, 
+    0xd1, 0x98, 0xad, 0xa4, 0x2e, 0x67, 0xf3, 0xcd, 0x40, 0x44, 
+    0xbd, 0x28, 0x21, 0x89, 0xa6, 0xb4, 0x9a, 0xb8, 0x30, 0xae, 
+    0xe8, 0xee, 0x2d, 0x39, 0xe2, 0x2e, 0x6f, 0x7c, 0xf1, 0x5c, 
+    0x6f, 0x78, 0xe3, 0x6e, 0x90, 0x87, 0x1c, 0x91, 0xc9, 0x8d, 
+    0xc2, 0x64, 0x2b, 0x03, 0x86, 0x5c, 0x94, 0xa8, 0xe6, 0x7a, 
+    0xf7, 0x40, 0x86, 0xbe, 0xbb, 0x61, 0x1d, 0xe0, 0x99, 0xf9, 
+    0xbd, 0xb0, 0x99, 0xf9, 0xbd, 0xb0, 0x66, 0x06, 0x42, 0x4f, 
+    0xc6, 0xb5, 0xef, 0xf2, 0xcb, 0x0c, 0xae, 0x1b, 0x09, 0x8a, 
+    0x03, 0xcd, 0xb3, 0x7b, 0x22, 0x02, 0x76, 0xe8, 0xab, 0x32, 
+    0x74, 0xcc, 0x22, 0x26, 0x76, 0xcc, 0x22, 0x22, 0x89, 0x33, 
+    0xdd, 0xdd, 0xcb, 0xca, 0xe4, 0x24, 0x0f, 0x9b, 0x44, 0xb0, 
+    0x0c, 0xec, 0x1d, 0x24, 0xe5, 0xc1, 0x47, 0xb3, 0xbc, 0x60, 
+    0xad, 0x51, 0xe2, 0xe2, 0x2d, 0x5d, 0xe0, 0xe2, 0x2d, 0x55, 
+    0x1f, 0x1d, 0xd2, 0xaa, 0x78, 0x65, 0x42, 0x20, 0xd9, 0xc6, 
+    0x7f, 0x58, 0x02, 0x8b, 0x8b, 0x54, 0x64, 0xea, 0xc5, 0x56, 
+    0x16, 0x04, 0x75, 0xf0, 0x74, 0x9c, 0xcc, 0x80, 0x54, 0x8e, 
+    0xc5, 0xf0, 0xab, 0x71, 0x3a, 0x0f, 0x31, 0x9e, 0x75, 0x1e, 
+    0x41, 0x8f, 0x48, 0x84, 0x25, 0x09, 0xd0, 0x30, 0xc0, 0x2f, 
+    0xda, 0xba, 0x0e, 0x33, 0xb8, 0xea, 0x46, 0x2d, 0xf4, 0x73, 
+    0x46, 0x2f, 0xf4, 0xf3, 0xb9, 0xd0, 0x0b, 0x0c, 0x2a, 0x2c, 
+    0xab, 0x38, 0x2c, 0xa1, 0x4e, 0xac, 0x04, 0x56, 0xf1, 0x8a, 
+    0xca, 0xc3, 0x14, 0xb6, 0x15, 0x83, 0x73, 0x6d, 0xb7, 0x91, 
+    0x18, 0x47, 0x97, 0x93, 0x10, 0x47, 0x68, 0x6c, 0xef, 0xb8, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x5d, 0x80, 0xb1, 0x27, 0xb4, 0xb2, 0x4b, 0x0c, 
+    0x6a, 0xaa, 0x6d, 0x8a, 0xf7, 0xae, 0xb4, 0x61, 0xb2, 0x81, 
+    0x47, 0x95, 0xf4, 0x8f, 0xc0, 0x83, 0xf4, 0x8b, 0xc0, 0x87, 
+    0x0b, 0x74, 0x3f, 0x78, 0x8b, 0xc9, 0x6d, 0x3d, 0x54, 0x1e, 
+    0x54, 0xdd, 0x68, 0x2b, 0xf7, 0xb2, 0x86, 0x79, 0xdb, 0x0a, 
+    0xdd, 0x1b, 0x4a, 0xb4, 0xe0, 0x06, 0xa7, 0x14, 0xe0, 0x0e, 
+    0xa7, 0x14, 0x1f, 0xf1, 0x58, 0xeb, 0x66, 0xf8, 0x4f, 0x13, 
+    0x43, 0x34, 0x4c, 0x27, 0x1d, 0xee, 0xe6, 0x19, 0x14, 0xc8, 
+    0x9d, 0x41, 0x68, 0x71, 0xba, 0xe5, 0x7f, 0x15, 0xac, 0xa9, 
+    0x7d, 0x35, 0xae, 0xe9, 0x82, 0xca, 0x51, 0x16, 0xe0, 0xaf, 
+    0x12, 0x83, 0xac, 0x0c, 0x63, 0x31, 0x0f, 0xb9, 0x00, 0x51, 
+    0x0e, 0x27, 0xac, 0x84, 0x12, 0x1e, 0x02, 0x4e, 0x0c, 0x2f, 
+    0x61, 0xce, 0x0c, 0x2f, 0x61, 0x4e, 0xf3, 0xd0, 0x9e, 0xb1, 
+    0x7c, 0x5b, 0xcb, 0xde, 0x06, 0x2c, 0x19, 0xeb, 0x19, 0x2e, 
+    0x71, 0x96, 0xc4, 0xac, 0x44, 0x03, 0x5d, 0x1b, 0xb6, 0xae, 
+    0x45, 0xbe, 0x97, 0x8f, 0x45, 0x3e, 0x97, 0x8f, 0xba, 0xc1, 
+    0x68, 0x70, 0x46, 0x67, 0xfa, 0x4b, 0x16, 0xd1, 0x80, 0x61, 
+    0x87, 0xdd, 0x7c, 0x75, 0x0b, 0xe0, 0x48, 0xf4, 0xa6, 0x0b, 
+    0xe1, 0x02, 0x0b, 0x8a, 0x61, 0x40, 0x8f, 0x8b, 0x61, 0x40, 
+    0x70, 0x74, 0x9e, 0xbf, 0x7b, 0xfb, 0xcb, 0x91, 0xf4, 0x67, 
+    0x6c, 0x57, 0xf6, 0x16, 0x4f, 0x75, 0x42, 0x28, 0xbf, 0x0e, 
+    0x62, 0x69, 0x70, 0xba, 0x4a, 0x2a, 0x60, 0xa0, 0x4a, 0x28, 
+    0x60, 0xa2, 0xb5, 0xd7, 0x9f, 0x5d, 0xde, 0x3d, 0x13, 0x11, 
+    0xb6, 0xad, 0x07, 0xf7, 0x65, 0xbf, 0xb0, 0xb6, 0xd6, 0x10, 
+    0x28, 0xf7, 0x9b, 0x62, 0x0b, 0x48, 0x53, 0xb4, 0x2b, 0x33, 
+    0xd3, 0xb4, 0x2b, 0x73, 0x2c, 0x4b, 0xd4, 0x8c, 0x00, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0xbb, 0x06, 0xf8, 0xf4, 0xe2, 0x7f, 0x58, 0xb0, 0x4c, 0x95, 
+    0xc4, 0x15, 0x18, 0xd4, 0x3c, 0xd2, 0x2f, 0xe7, 0x0d, 0x13, 
+    0x19, 0x67, 0x0f, 0x50, 0x39, 0x67, 0x0c, 0x50, 0xc6, 0x98, 
+    0xf3, 0xaf, 0xca, 0x03, 0x23, 0x31, 0x22, 0xf6, 0xe4, 0xc6, 
+    0x43, 0x06, 0xd3, 0x04, 0x83, 0xf9, 0x2b, 0xe0, 0x5a, 0x80, 
+    0x4c, 0x36, 0x12, 0xf0, 0xc1, 0x77, 0x12, 0xd1, 0xc9, 0x77, 
+    0xed, 0x2e, 0x36, 0x88, 0xe9, 0x74, 0x24, 0xbb, 0x6a, 0x18, 
+    0x27, 0x66, 0xfb, 0xab, 0x3c, 0x45, 0xef, 0xba, 0x9c, 0x13, 
+    0xe7, 0xe3, 0xe2, 0x8e, 0x62, 0x97, 0x76, 0x03, 0x62, 0xb3, 
+    0xf6, 0x03, 0x9d, 0x4c, 0x09, 0xfc, 0x3b, 0x18, 0x69, 0x09, 
+    0xc9, 0xb5, 0x86, 0x0b, 0xb5, 0xc5, 0xe9, 0x2f, 0xd8, 0x24, 
+    0x62, 0x33, 0x91, 0x04, 0x64, 0xf8, 0x90, 0x01, 0x14, 0xfa, 
+    0x90, 0x01, 0x20, 0xfa, 0x6f, 0xfe, 0xdf, 0x05, 0x8a, 0x3f, 
+    0x8b, 0x7e, 0xad, 0x86, 0xaf, 0x6e, 0x24, 0x8a, 0x72, 0xc7, 
+    0x72, 0xea, 0x7b, 0xf6, 0x20, 0x25, 0x44, 0xb7, 0xa2, 0xb0, 
+    0x76, 0xc7, 0xa2, 0xa0, 0x76, 0xd7, 0x5d, 0x5f, 0x89, 0x28, 
+    0x3e, 0xa2, 0x76, 0x94, 0x65, 0x98, 0x57, 0xe7, 0x89, 0xb6, 
+    0x26, 0x29, 0x9c, 0xfb, 0xb6, 0xa6, 0xa2, 0x7e, 0xfa, 0xe2, 
+    0xb0, 0xf4, 0xe9, 0xe6, 0xb0, 0xf4, 0xe9, 0xe6, 0x4f, 0x0b, 
+    0x16, 0x19, 0x4d, 0xa5, 0x55, 0xd0, 0x1f, 0x4d, 0x7b, 0x8d, 
+    0xda, 0x14, 0xd1, 0xa6, 0x68, 0x62, 0xbb, 0xe5, 0x22, 0xba, 
+    0x2c, 0xc5, 0xc8, 0x7e, 0x7c, 0x49, 0xc8, 0x7a, 0x7c, 0x4d, 
+    0x37, 0x85, 0x83, 0xb2, 0xc3, 0x5e, 0x01, 0xcd, 0x9c, 0x11, 
+    0xf3, 0x2d, 0x7a, 0xee, 0x11, 0x31, 0xdc, 0xab, 0x44, 0xc4, 
+    0x51, 0x14, 0xb3, 0x77, 0x48, 0xb4, 0x45, 0x01, 0x58, 0xb4, 
+    0x45, 0x41, 0xa7, 0x4b, 0xba, 0xbe, 0x01, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0xbe, 0xc5, 
+    0x05, 0xaf, 0x59, 0x37, 0xee, 0x47, 0xeb, 0x14, 0xa1, 0x64, 
+    0xd2, 0x71, 0x6a, 0x0a, 0x66, 0x20, 0x65, 0xdf, 0xc6, 0x29, 
+    0x75, 0x22, 0x46, 0x21, 0x75, 0x26, 0xb9, 0xde, 0x8a, 0xd9, 
+    0x70, 0x38, 0x0c, 0x91, 0x5f, 0x36, 0x4e, 0xfd, 0x6f, 0x1f, 
+    0x8b, 0x2d, 0x97, 0xe7, 0x26, 0x1e, 0x06, 0x4f, 0x7c, 0xbe, 
+    0x02, 0xe7, 0xb5, 0xae, 0x82, 0xe7, 0xb4, 0xbe, 0x7d, 0x18, 
+    0x4b, 0x41, 0xed, 0xf3, 0x96, 0x19, 0xef, 0x3d, 0x82, 0xea, 
+    0xd6, 0x41, 0xb4, 0x36, 0xd4, 0x62, 0x5b, 0xf6, 0x85, 0xaf, 
+    0x52, 0xf8, 0xae, 0x61, 0x5e, 0xa7, 0xac, 0x61, 0x5e, 0xa7, 
+    0x53, 0x9e, 0xa1, 0x58, 0x63, 0x0e, 0x62, 0xd3, 0x65, 0x9c, 
+    0x9e, 0x88, 0x23, 0xdd, 0xa4, 0xc0, 0x8c, 0x63, 0xdd, 0xd9, 
+    0x45, 0xde, 0x60, 0x0f, 0xd6, 0x46, 0x02, 0xa3, 0x54, 0x4e, 
+    0x02, 0xa3, 0xab, 0xb1, 0xfd, 0x5c, 0xef, 0x16, 0x95, 0xbd, 
+    0x36, 0x9d, 0x3a, 0xd2, 0x62, 0x0c, 0x58, 0x20, 0x86, 0x9f, 
+    0x1d, 0xb8, 0xbf, 0x75, 0x78, 0xa6, 0x17, 0x7c, 0x59, 0x21, 
+    0x97, 0x7c, 0x59, 0x25, 0x68, 0x83, 0xa6, 0xda, 0x5f, 0xd0, 
+    0x64, 0xf4, 0x18, 0x66, 0x4e, 0xbf, 0x5b, 0xfc, 0x3f, 0x39, 
+    0xbe, 0xf4, 0x88, 0xe1, 0xe4, 0x44, 0x5a, 0xd6, 0xf7, 0x67, 
+    0x1a, 0x12, 0xf7, 0x44, 0x1a, 0x12, 0x08, 0xbb, 0xe5, 0xed, 
+    0xbc, 0xf8, 0xe7, 0x64, 0xf8, 0xf0, 0xa5, 0xc5, 0x00, 0xc5, 
+    0xf5, 0x39, 0x05, 0x8a, 0xc1, 0x6d, 0x29, 0xfd, 0x8e, 0x4d, 
+    0x68, 0x94, 0xc7, 0x6d, 0x28, 0x94, 0xc7, 0x6d, 0xd7, 0x6b, 
+    0x38, 0x92, 0x5a, 0xad, 0x07, 0xc2, 0x9b, 0x86, 0xe7, 0x02, 
+    0x5f, 0x10, 0xef, 0x0d, 0xe7, 0xf2, 0x98, 0x78, 0x3d, 0xe8, 
+    0xe4, 0x0f, 0x65, 0xc8, 0xd0, 0x7b, 0x65, 0xc8, 0xd0, 0x5b, 
+    0x9a, 0x37, 0x2f, 0xa4, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0xd0, 0xce, 0xb3, 0x2b, 
+    0x6c, 0xd6, 0x8b, 0x4e, 0xf2, 0xab, 0xeb, 0x05, 0xb0, 0xe7, 
+    0xcc, 0xd2, 0x15, 0xfe, 0x97, 0x4d, 0x86, 0xbe, 0xdf, 0xcd, 
+    0x86, 0xbe, 0xcf, 0xcd, 0x79, 0x41, 0x30, 0x32, 0x9c, 0xa9, 
+    0x61, 0x00, 0x3e, 0x63, 0xb4, 0x5b, 0x38, 0xf2, 0x36, 0x88, 
+    0x07, 0xb6, 0x64, 0x45, 0x92, 0xe4, 0x4d, 0xa2, 0x00, 0xb7, 
+    0x4d, 0x2c, 0x00, 0xb7, 0x4d, 0xac, 0xff, 0x48, 0xb2, 0x53, 
+    0xcf, 0x72, 0x93, 0xb4, 0x91, 0xbc, 0x1b, 0x03, 0x01, 0x7a, 
+    0xf9, 0x37, 0x06, 0xcf, 0x5c, 0x8e, 0x2f, 0xc9, 0x76, 0xb3, 
+    0x18, 0x7a, 0x76, 0xfc, 0x18, 0xba, 0x76, 0xbe, 0xe7, 0x05, 
+    0x89, 0x41, 0x6a, 0x72, 0x3f, 0xb8, 0x16, 0xfd, 0xf3, 0xdc, 
+    0xe4, 0x97, 0xbc, 0xcc, 0x12, 0x9b, 0x07, 0x02, 0xaf, 0xd9, 
+    0xdb, 0x92, 0x8d, 0x59, 0xc7, 0x06, 0x8d, 0xd9, 0xc7, 0x06, 
+    0x72, 0x26, 0x38, 0xf9, 0xcd, 0x73, 0xe4, 0x35, 0x7a, 0x80, 
+    0xf9, 0xfb, 0x09, 0x53, 0xe6, 0x44, 0x54, 0x9c, 0x24, 0x13, 
+    0xc8, 0xe3, 0x82, 0xe1, 0x48, 0xca, 0x94, 0x63, 0x48, 0xca, 
+    0x84, 0x63, 0xb7, 0x35, 0x7b, 0x9c, 0x41, 0xbb, 0x1b, 0xf7, 
+    0x2c, 0xfe, 0x1e, 0x13, 0x65, 0x39, 0xde, 0x26, 0x46, 0x50, 
+    0xf2, 0xa1, 0x00, 0x0b, 0xd5, 0x79, 0x54, 0x43, 0x15, 0x2a, 
+    0x44, 0x43, 0x15, 0x6b, 0xbb, 0xbc, 0xea, 0x94, 0xa3, 0x44, 
+    0xb6, 0xc5, 0x53, 0xaa, 0xcd, 0xa0, 0xe3, 0x06, 0xeb, 0x20, 
+    0x73, 0xc4, 0x45, 0x0d, 0x1a, 0xba, 0xda, 0x20, 0xd3, 0xe2, 
+    0xba, 0x64, 0xd3, 0xe2, 0xba, 0x20, 0x2c, 0x1d, 0x45, 0xdf, 
+    0x34, 0x47, 0x05, 0x41, 0x1c, 0xb3, 0x04, 0x02, 0x19, 0xa1, 
+    0xa4, 0x04, 0x7f, 0x75, 0xc3, 0xc7, 0xf9, 0xaf, 0xac, 0x8f, 
+    0xd9, 0xf3, 0x4f, 0xb7, 0xd9, 0xf3, 0x4f, 0x97, 0x26, 0x0c, 
+    0xb0, 0x68, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x13, 0xb9, 0xd8, 0x0b, 0xa9, 0xaf, 
+    0x83, 0x19, 0xb5, 0xea, 0x75, 0x81, 0xb1, 0xd3, 0x52, 0x95, 
+    0x60, 0x4b, 0x18, 0xde, 0x74, 0x9f, 0x98, 0x9d, 0x74, 0xdf, 
+    0x18, 0x9d, 0x8b, 0x20, 0xe7, 0x62, 0x82, 0x65, 0xb3, 0x18, 
+    0x0c, 0xa4, 0x3b, 0xfb, 0x39, 0xff, 0xe7, 0x47, 0x7b, 0x1d, 
+    0xb6, 0xfe, 0x7b, 0xb3, 0xa6, 0xda, 0x53, 0xb5, 0x66, 0xdb, 
+    0x5b, 0xb5, 0x66, 0xdb, 0xa4, 0x4a, 0x99, 0x24, 0x01, 0x3f, 
+    0xf3, 0xc0, 0x60, 0x70, 0x14, 0xd2, 0xe0, 0xf8, 0xc4, 0x31, 
+    0xc9, 0x20, 0xc1, 0x69, 0x35, 0x30, 0x56, 0xc9, 0x84, 0x28, 
+    0x4d, 0x4a, 0x8c, 0x28, 0x4d, 0x4a, 0x73, 0xd7, 0xb2, 0xb5, 
+    0x3f, 0x6e, 0x45, 0xd9, 0xa0, 0x54, 0x23, 0x06, 0x23, 0xcd, 
+    0x8f, 0x76, 0x39, 0x7d, 0xa7, 0x4d, 0x25, 0x90, 0x1e, 0x6b, 
+    0x21, 0xdf, 0xf3, 0x29, 0x21, 0xd7, 0xfb, 0x69, 0xde, 0x28, 
+    0x04, 0x96, 0x87, 0x17, 0xd5, 0x68, 0x9c, 0x09, 0x04, 0xce, 
+    0x1c, 0xc2, 0x2e, 0xed, 0xe5, 0xe2, 0xf5, 0x61, 0x52, 0x34, 
+    0x4f, 0x27, 0x5a, 0x66, 0x43, 0x21, 0x7a, 0x66, 0x43, 0x21, 
+    0x85, 0x99, 0xbc, 0xde, 0xb3, 0x1a, 0x43, 0x87, 0x0a, 0xd3, 
+    0xda, 0xc4, 0xa9, 0xfd, 0x54, 0xd3, 0x60, 0xa0, 0x2c, 0x41, 
+    0x75, 0x74, 0x20, 0x25, 0x68, 0x68, 0x2c, 0x03, 0x68, 0x60, 
+    0x2c, 0x03, 0x97, 0x9f, 0xd3, 0xfc, 0xec, 0x2a, 0x7c, 0x2e, 
+    0x7b, 0x70, 0x1b, 0x98, 0xc6, 0xa7, 0xd2, 0x32, 0xbd, 0x35, 
+    0xff, 0x14, 0x27, 0x7d, 0x1a, 0x50, 0x8e, 0x61, 0x3e, 0x50, 
+    0x8f, 0x61, 0x3e, 0x50, 0x70, 0x9e, 0xc1, 0xaf, 0x06, 0x1f, 
+    0xe1, 0xf8, 0x97, 0xab, 0x39, 0x0c, 0x51, 0x48, 0x88, 0x2a, 
+    0xa4, 0xdf, 0x5c, 0x42, 0x5e, 0xee, 0x8a, 0xec, 0x5c, 0xef, 
+    0xcc, 0xdc, 0x5c, 0xef, 0xcc, 0xdc, 0xa3, 0x10, 0x33, 0x23, 
+    0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00
+};
+
diff --git a/neureka/pointwise/Makefile b/neureka/pointwise/Makefile
new file mode 100644
index 0000000..ad8dc3b
--- /dev/null
+++ b/neureka/pointwise/Makefile
@@ -0,0 +1,3 @@
+include ../app/Makefile
+
+STIM_DIR := ../pointwise/
\ No newline at end of file
diff --git a/neureka/pointwise/inc/bias.h b/neureka/pointwise/inc/bias.h
new file mode 100644
index 0000000..f28d0f5
--- /dev/null
+++ b/neureka/pointwise/inc/bias.h
@@ -0,0 +1,9 @@
+#ifndef __BIAS_H__
+#define __BIAS_H__
+
+#include <pmsis.h>
+
+#define BIAS_SIZE (39)
+extern int32_t bias[BIAS_SIZE];
+
+#endif  // __BIAS_H__
diff --git a/neureka/pointwise/inc/input.h b/neureka/pointwise/inc/input.h
new file mode 100644
index 0000000..985f9a9
--- /dev/null
+++ b/neureka/pointwise/inc/input.h
@@ -0,0 +1,9 @@
+#ifndef __INPUT_H__
+#define __INPUT_H__
+
+#include <pmsis.h>
+
+#define INPUT_SIZE (8995)
+extern uint8_t input[INPUT_SIZE];
+
+#endif  // __INPUT_H__
diff --git a/neureka/pointwise/inc/layer_conf.h b/neureka/pointwise/inc/layer_conf.h
new file mode 100644
index 0000000..e5e4812
--- /dev/null
+++ b/neureka/pointwise/inc/layer_conf.h
@@ -0,0 +1,42 @@
+#ifndef __LAYER_CONF_H__
+#define __LAYER_CONF_H__
+
+#define TEST_NAME "test"
+#define INPUT_HEIGHT (7)
+#define INPUT_WIDTH (5)
+#define INPUT_CHANNEL (257)
+#define INPUT_SIGNED (0)
+#define INPUT_BITS (8)
+
+#define OUTPUT_HEIGHT (7)
+#define OUTPUT_WIDTH (5)
+#define OUTPUT_CHANNEL (39)
+#define OUTPUT_BITS (8)
+
+#define WEIGHT_HEIGHT (1)
+#define WEIGHT_WIDTH (1)
+#define WEIGHT_CHANNEL_IN (257)
+#define WEIGHT_CHANNEL_OUT (39)
+#define WEIGHT_BITS (8)
+#define WEIGHT_OFFSET (-128)
+
+#define SCALE_BITS (8)
+
+#define BIAS_BITS (32)
+
+#define PADDING_TOP (0)
+#define PADDING_BOTTOM (0)
+#define PADDING_LEFT (0)
+#define PADDING_RIGHT (0)
+#define PADDING_VALUE (0)
+
+#define STRIDE_HEIGHT (1)
+#define STRIDE_WIDTH (1)
+
+#define GROUPS (1)
+#define OUTSHIFT (12)
+#define HAS_NORM_QUANT (1)
+#define HAS_BIAS (1)
+#define HAS_RELU (1)
+
+#endif  // __LAYER_CONF_H__
diff --git a/neureka/pointwise/inc/output.h b/neureka/pointwise/inc/output.h
new file mode 100644
index 0000000..68c6c11
--- /dev/null
+++ b/neureka/pointwise/inc/output.h
@@ -0,0 +1,14 @@
+#ifndef __OUTPUT_H__
+#define __OUTPUT_H__
+
+#include <pmsis.h>
+
+#define OUTPUT_SIZE (1365)
+extern uint8_t output[OUTPUT_SIZE];
+
+#define GOLDEN_OUTPUT_SIZE (1365)
+extern uint8_t golden_output[GOLDEN_OUTPUT_SIZE];
+
+int check_output();
+
+#endif  // __OUTPUT_H__
diff --git a/neureka/pointwise/inc/scale.h b/neureka/pointwise/inc/scale.h
new file mode 100644
index 0000000..67dea26
--- /dev/null
+++ b/neureka/pointwise/inc/scale.h
@@ -0,0 +1,9 @@
+#ifndef __SCALE_H__
+#define __SCALE_H__
+
+#include <pmsis.h>
+
+#define SCALE_SIZE (39)
+extern uint8_t scale[SCALE_SIZE];
+
+#endif  // __SCALE_H__
diff --git a/neureka/pointwise/inc/weight.h b/neureka/pointwise/inc/weight.h
new file mode 100644
index 0000000..ba39fae
--- /dev/null
+++ b/neureka/pointwise/inc/weight.h
@@ -0,0 +1,9 @@
+#ifndef __WEIGHT_H__
+#define __WEIGHT_H__
+
+#include <pmsis.h>
+
+#define WEIGHT_SIZE (11232)
+extern uint8_t weight[WEIGHT_SIZE];
+
+#endif  // __WEIGHT_H__
diff --git a/neureka/pointwise/src/bias.c b/neureka/pointwise/src/bias.c
new file mode 100644
index 0000000..55b2013
--- /dev/null
+++ b/neureka/pointwise/src/bias.c
@@ -0,0 +1,10 @@
+#include "bias.h"
+
+#define BIAS_SIZE (39)
+PI_L1 int32_t bias[BIAS_SIZE] = {
+    0x36dfa, -0xec5f, 0x3ee6f, -0x23173, -0xb8b1, -0x236a3, -0x3fe84, 0x1285e, 0x9a43, -0xfb97, 
+    0x396a7, -0x16b65, 0x5527, 0xb8f, 0xe316, 0x3c2e1, 0x2bf5e, -0x1adf0, -0x486, 0x1ed7d, 
+    0x3a88e, 0x8e46, -0x3e8ef, 0xcbee, 0x65e5, 0x2426b, 0x298f4, 0x205dc, 0x2598, -0x2151c, 
+    0x390ef, -0x2318f, 0x1bc1, 0xe222, 0x1a09e, -0x1f3fd, 0x3d956, -0x12e4d, 0x33ee0
+};
+
diff --git a/neureka/pointwise/src/input.c b/neureka/pointwise/src/input.c
new file mode 100644
index 0000000..b6afde2
--- /dev/null
+++ b/neureka/pointwise/src/input.c
@@ -0,0 +1,906 @@
+#include "input.h"
+
+#define INPUT_SIZE (8995)
+PI_L1 uint8_t input[INPUT_SIZE] = {
+    0xc2, 0xe1, 0x4d, 0x2d, 0x90, 0xf3, 0x32, 0xc8, 0x6d, 0x7a, 
+    0x72, 0x92, 0xc8, 0x2d, 0x28, 0x1b, 0x48, 0xe5, 0x2c, 0x55, 
+    0x62, 0x70, 0xd0, 0x6e, 0x3c, 0x2b, 0x77, 0x12, 0xdf, 0x21, 
+    0x02, 0x62, 0xc1, 0xae, 0x33, 0x74, 0xd1, 0xbd, 0x81, 0x35, 
+    0xb8, 0xf3, 0x0b, 0xb3, 0x5f, 0x10, 0x9c, 0xe5, 0x2d, 0x8d, 
+    0xa2, 0x92, 0x94, 0xbb, 0x8e, 0xda, 0x96, 0xf8, 0xe2, 0x45, 
+    0x7f, 0xb6, 0x04, 0x30, 0x4b, 0x00, 0x06, 0x95, 0x77, 0x44, 
+    0x1f, 0x66, 0xaf, 0x5d, 0x95, 0xb9, 0x4e, 0xcd, 0xde, 0x94, 
+    0x7f, 0x2e, 0x93, 0x3b, 0xec, 0x87, 0x51, 0xd8, 0x96, 0x02, 
+    0x91, 0x96, 0x18, 0x23, 0x4c, 0xd1, 0x04, 0x66, 0xde, 0xf6, 
+    0xc6, 0x32, 0xec, 0x4c, 0x59, 0x75, 0xd6, 0x30, 0x17, 0xc5, 
+    0xe6, 0xc4, 0xe7, 0x8e, 0x65, 0x02, 0x1d, 0x3a, 0xd9, 0x3c, 
+    0x8f, 0xe2, 0xf4, 0xcc, 0x88, 0xc6, 0xea, 0xc3, 0xaf, 0x4b, 
+    0xb6, 0x48, 0x01, 0xaf, 0x0f, 0xfa, 0xdc, 0x74, 0x1c, 0x29, 
+    0x31, 0xf8, 0x9c, 0x54, 0xa4, 0xf0, 0xb5, 0xbe, 0xe4, 0x43, 
+    0xaf, 0x5a, 0x62, 0x55, 0x69, 0x89, 0x53, 0x0c, 0xc7, 0x1a, 
+    0x70, 0xb0, 0x1f, 0xe7, 0x91, 0x60, 0x08, 0x51, 0xd8, 0xc6, 
+    0x8e, 0x37, 0x48, 0x64, 0x57, 0x90, 0xef, 0x85, 0xb6, 0x1d, 
+    0xc9, 0x95, 0x47, 0x9f, 0xd8, 0xe3, 0xd0, 0x77, 0x1a, 0x59, 
+    0x1c, 0x4f, 0x1e, 0x65, 0x08, 0xba, 0x3a, 0xc4, 0xd0, 0x29, 
+    0x98, 0x72, 0x2f, 0x54, 0x3f, 0x2d, 0xd0, 0x00, 0xbd, 0x4d, 
+    0xad, 0x8a, 0x99, 0xe8, 0x75, 0x42, 0x32, 0xc7, 0x01, 0xf3, 
+    0xf9, 0xbe, 0x66, 0x2a, 0x77, 0x93, 0x90, 0x69, 0x55, 0x34, 
+    0x6e, 0x48, 0x1a, 0x9b, 0xc6, 0xbb, 0x35, 0x5d, 0x64, 0xd8, 
+    0xe0, 0x86, 0x1a, 0x2f, 0xa1, 0x5f, 0xd9, 0x07, 0xb8, 0x42, 
+    0x49, 0xac, 0x6c, 0x0f, 0xdd, 0x27, 0x69, 0x36, 0x43, 0x14, 
+    0x99, 0x7f, 0x01, 0xdd, 0xad, 0x30, 0xf2, 0x46, 0xa4, 0x5b, 
+    0x74, 0xed, 0xac, 0x90, 0xbb, 0x75, 0xd1, 0x64, 0xe1, 0x5d, 
+    0x3d, 0xf1, 0x29, 0x6d, 0x44, 0xb2, 0x4a, 0x39, 0x8c, 0x9b, 
+    0x25, 0xa2, 0x89, 0x5c, 0x14, 0xba, 0xd9, 0xa0, 0x0b, 0x22, 
+    0x84, 0xfb, 0xd5, 0xe9, 0xe2, 0x42, 0x0c, 0xe8, 0x5d, 0x0a, 
+    0x99, 0xf7, 0x61, 0x35, 0xe5, 0x7b, 0x8c, 0x2c, 0x38, 0x1a, 
+    0xca, 0x94, 0x66, 0xae, 0x45, 0xa9, 0x4c, 0x52, 0x18, 0x1f, 
+    0x08, 0xf8, 0xfd, 0x9c, 0x03, 0xf2, 0xfb, 0xfe, 0xa9, 0x55, 
+    0xae, 0x69, 0x2c, 0x93, 0xd7, 0x8d, 0x2f, 0x63, 0xf7, 0x17, 
+    0x37, 0xe3, 0xeb, 0x0f, 0x2e, 0x2b, 0x33, 0x53, 0xcc, 0xfb, 
+    0x4d, 0x4f, 0x7f, 0x85, 0xcf, 0x16, 0x34, 0x88, 0x2d, 0x44, 
+    0x58, 0xda, 0x76, 0xca, 0xc3, 0xa8, 0x45, 0xc7, 0x9a, 0x25, 
+    0xd8, 0x59, 0x83, 0x5b, 0x21, 0xcd, 0x00, 0x2b, 0xcc, 0xa7, 
+    0x24, 0x45, 0x98, 0x02, 0x64, 0x7d, 0x18, 0x19, 0xe0, 0x03, 
+    0x0d, 0xc9, 0x1a, 0x5b, 0x29, 0x61, 0x46, 0xb4, 0x54, 0xe4, 
+    0x20, 0xb0, 0xcd, 0xa3, 0xf6, 0xb9, 0x43, 0x91, 0xcc, 0xc8, 
+    0x2b, 0xea, 0x49, 0xce, 0x83, 0x12, 0xb6, 0xc2, 0x38, 0x36, 
+    0x0b, 0x71, 0x22, 0x72, 0x78, 0x77, 0x01, 0xf0, 0xd3, 0x74, 
+    0x53, 0x70, 0x19, 0xf8, 0xac, 0x70, 0xf1, 0xd0, 0xdc, 0xea, 
+    0x71, 0x3b, 0xa8, 0xfc, 0x12, 0x3b, 0x06, 0x5c, 0x7d, 0x21, 
+    0x6e, 0x1b, 0xbf, 0xea, 0x79, 0x84, 0x19, 0x42, 0xf3, 0xbd, 
+    0xc1, 0x0e, 0xf8, 0x31, 0x3c, 0xfc, 0x34, 0x7f, 0xf6, 0x34, 
+    0xad, 0x0d, 0x01, 0x19, 0xa0, 0x61, 0x1e, 0xe8, 0xc6, 0x56, 
+    0x42, 0xf5, 0xb2, 0xa4, 0x90, 0x6e, 0x59, 0x3f, 0x89, 0x92, 
+    0x8d, 0x5d, 0x17, 0x69, 0x41, 0x19, 0x17, 0xca, 0xbd, 0x5b, 
+    0xb1, 0x4f, 0xd7, 0xcb, 0xe9, 0x68, 0xc3, 0x38, 0x5a, 0x5c, 
+    0x7f, 0x0b, 0xe9, 0xf3, 0x6e, 0x20, 0xcd, 0xca, 0x2d, 0x95, 
+    0xec, 0x85, 0xf0, 0xcd, 0xd4, 0xbd, 0x1f, 0xe5, 0xda, 0xf2, 
+    0x1b, 0xca, 0x43, 0x2e, 0x15, 0x06, 0xba, 0x29, 0x8e, 0xc0, 
+    0x99, 0x5e, 0x0f, 0x4f, 0x47, 0x41, 0xac, 0x69, 0xb4, 0xcd, 
+    0x00, 0x10, 0x21, 0xfc, 0x6b, 0xe2, 0x3e, 0x9b, 0xc6, 0xaa, 
+    0x5d, 0x8c, 0x28, 0x54, 0x36, 0xdc, 0x6a, 0xbc, 0xa6, 0x8d, 
+    0xf2, 0x07, 0x3b, 0x1c, 0x33, 0x05, 0x82, 0xd4, 0xaa, 0x5c, 
+    0xd4, 0xc6, 0x8d, 0x2d, 0x1e, 0x2c, 0xe4, 0xed, 0x12, 0x75, 
+    0x14, 0xbb, 0xf6, 0x04, 0x21, 0xc6, 0xdd, 0xc6, 0x52, 0xd5, 
+    0x71, 0x8b, 0x83, 0x0b, 0x89, 0x41, 0xe3, 0x36, 0x12, 0xef, 
+    0xdd, 0xce, 0x03, 0xdd, 0x7e, 0xef, 0x93, 0x22, 0x63, 0x25, 
+    0xb7, 0x1f, 0x4c, 0x17, 0x4e, 0x2e, 0x14, 0x1f, 0xc3, 0x6e, 
+    0xd7, 0x3c, 0x69, 0x15, 0xb7, 0xe8, 0x62, 0x2c, 0x7a, 0xbf, 
+    0x24, 0xe4, 0xa5, 0x50, 0xef, 0xa0, 0x53, 0x7c, 0x49, 0x20, 
+    0x37, 0x46, 0x27, 0x5f, 0x81, 0x1f, 0xb9, 0x08, 0x3b, 0x8f, 
+    0xcc, 0x13, 0xf6, 0x4c, 0x2f, 0x17, 0x68, 0x04, 0x7b, 0xd6, 
+    0x93, 0x1a, 0x50, 0x38, 0x2e, 0xda, 0x0a, 0x21, 0x4e, 0xf1, 
+    0x03, 0xea, 0xeb, 0x59, 0x63, 0xef, 0xf9, 0xf2, 0xcf, 0xa0, 
+    0x43, 0xc3, 0x6b, 0x9e, 0x07, 0x70, 0x05, 0x7d, 0xa5, 0x88, 
+    0x81, 0x45, 0xf1, 0x77, 0x5b, 0xd3, 0xc7, 0x3c, 0x0e, 0x3c, 
+    0xc5, 0x12, 0xc6, 0x37, 0xa9, 0xc7, 0x97, 0x84, 0xcf, 0x77, 
+    0xd1, 0x25, 0x98, 0x69, 0xf8, 0x1d, 0x8e, 0x97, 0xea, 0x8b, 
+    0xdb, 0xba, 0x02, 0x66, 0xee, 0x60, 0xea, 0x5b, 0x00, 0x7e, 
+    0xef, 0xed, 0x99, 0x1d, 0x97, 0xc5, 0x78, 0x53, 0x95, 0x7e, 
+    0xc6, 0xb2, 0x99, 0x51, 0xfa, 0xac, 0x55, 0xbe, 0x59, 0x1a, 
+    0xba, 0x87, 0x65, 0x19, 0x20, 0xe1, 0x3a, 0xa8, 0x52, 0xd1, 
+    0x3d, 0x77, 0xb9, 0x68, 0xdc, 0x63, 0x5a, 0xc0, 0xbc, 0x75, 
+    0x94, 0xc6, 0xa0, 0xc7, 0x31, 0xc2, 0x00, 0x63, 0x55, 0x74, 
+    0x5e, 0x79, 0x12, 0xdb, 0xc0, 0xbf, 0xd6, 0xbe, 0x5a, 0xa9, 
+    0xa9, 0xbd, 0xf2, 0xf8, 0x76, 0x1e, 0xfc, 0xc8, 0x77, 0x3b, 
+    0xb1, 0xc0, 0x4c, 0xa0, 0xa2, 0x92, 0xb5, 0x37, 0x63, 0x38, 
+    0x64, 0xf5, 0x06, 0xee, 0x53, 0x08, 0xc2, 0x70, 0xdc, 0x34, 
+    0x60, 0xa8, 0x23, 0x78, 0xb3, 0xed, 0xe6, 0x1d, 0x6a, 0x7e, 
+    0xbf, 0x88, 0x4c, 0x01, 0x9f, 0x52, 0x1a, 0xac, 0xa6, 0x57, 
+    0x6f, 0xb7, 0xae, 0x02, 0x4d, 0x96, 0x8b, 0xfe, 0x92, 0x09, 
+    0x01, 0xb5, 0xa2, 0x9e, 0x45, 0x09, 0x61, 0xbc, 0xa7, 0x70, 
+    0x4e, 0x5a, 0x0d, 0x88, 0xc2, 0xe0, 0xdd, 0x81, 0xf7, 0x6b, 
+    0xa8, 0x2d, 0xc7, 0xd9, 0x32, 0x8b, 0x29, 0x35, 0x65, 0x10, 
+    0xb2, 0x4a, 0xe9, 0xb0, 0x58, 0xd7, 0x78, 0xae, 0x66, 0x6c, 
+    0xd0, 0x6f, 0x4c, 0xfe, 0x5c, 0x09, 0xee, 0xb1, 0x99, 0xdd, 
+    0x67, 0xa2, 0x01, 0x87, 0xe8, 0xe1, 0xd5, 0x8e, 0xf3, 0x4c, 
+    0x68, 0x67, 0x4b, 0xb8, 0x73, 0x9a, 0xa9, 0x5c, 0x8a, 0x37, 
+    0x6b, 0xb9, 0xf0, 0x42, 0xb5, 0xbe, 0x69, 0x83, 0x60, 0x8c, 
+    0xdb, 0x33, 0xf0, 0x74, 0x57, 0x3b, 0x81, 0xee, 0xb2, 0x2b, 
+    0xf2, 0xf3, 0xad, 0xe0, 0x68, 0x94, 0xb4, 0xaa, 0x33, 0x43, 
+    0xe0, 0x3e, 0xbe, 0x7d, 0x58, 0xb9, 0xa2, 0x44, 0x80, 0x6f, 
+    0x63, 0xbe, 0xf9, 0xc3, 0x5a, 0x84, 0xfe, 0xbd, 0xbe, 0xcb, 
+    0xad, 0x3a, 0x77, 0x6c, 0x36, 0xc9, 0x94, 0x61, 0x17, 0xd8, 
+    0xa2, 0x5d, 0xe0, 0x60, 0xfa, 0xa5, 0xc3, 0x68, 0x73, 0xb8, 
+    0x81, 0xbb, 0x00, 0x6a, 0xc4, 0xdd, 0xb9, 0x10, 0x25, 0x75, 
+    0x93, 0xb6, 0x51, 0xca, 0xbf, 0x75, 0xca, 0x64, 0xdf, 0x67, 
+    0xe4, 0x62, 0xd8, 0x46, 0xed, 0x2a, 0xc2, 0xd7, 0x91, 0x99, 
+    0x27, 0x7a, 0xfc, 0x21, 0xaf, 0x3c, 0x7d, 0xc9, 0x2a, 0xfd, 
+    0x40, 0x6a, 0xca, 0xb4, 0x90, 0x78, 0x3c, 0xe7, 0xae, 0x62, 
+    0x04, 0x98, 0x7b, 0x01, 0xe8, 0xed, 0x6f, 0xe3, 0x3f, 0xed, 
+    0xc9, 0xee, 0x98, 0x66, 0x3a, 0xde, 0x89, 0xd6, 0xb8, 0x25, 
+    0xf0, 0x09, 0xd1, 0xa4, 0x77, 0xec, 0xa3, 0xd6, 0xdb, 0x15, 
+    0xe5, 0x3b, 0xdd, 0x36, 0x96, 0x69, 0x26, 0x99, 0x86, 0xcf, 
+    0x32, 0x6e, 0x4b, 0x57, 0x00, 0x52, 0x12, 0xa4, 0xf8, 0x37, 
+    0xf5, 0x27, 0x70, 0xc7, 0x4e, 0xb8, 0x09, 0xe5, 0x18, 0xaf, 
+    0x23, 0x59, 0xa6, 0xbe, 0xa5, 0x25, 0x37, 0xb1, 0xf0, 0x60, 
+    0x95, 0xcf, 0x6a, 0x8c, 0x96, 0xd8, 0x1c, 0x02, 0xd5, 0x5d, 
+    0x32, 0x44, 0x2b, 0x75, 0x52, 0xcf, 0x08, 0x74, 0x2e, 0x1c, 
+    0xfa, 0x91, 0x7e, 0xd5, 0x36, 0xe7, 0xfe, 0x85, 0x68, 0x09, 
+    0x45, 0xdb, 0x7e, 0x98, 0x2a, 0x7b, 0x6f, 0xe8, 0x04, 0x6c, 
+    0xf1, 0xb0, 0xda, 0x5d, 0x9c, 0xb1, 0x81, 0xe7, 0x1d, 0x16, 
+    0x33, 0xb6, 0xb8, 0x10, 0xf0, 0xfd, 0x25, 0x1f, 0xb8, 0xf1, 
+    0xea, 0x60, 0x58, 0xe5, 0x3a, 0xf0, 0x6a, 0xbd, 0x05, 0xc1, 
+    0xe1, 0x5c, 0x8e, 0x16, 0xdb, 0x61, 0xb8, 0x07, 0x97, 0x1d, 
+    0x84, 0x76, 0x44, 0x43, 0x55, 0xc5, 0x2f, 0x07, 0x39, 0x76, 
+    0xfb, 0x7d, 0x65, 0xb2, 0xdc, 0xdc, 0x9e, 0xed, 0x17, 0x38, 
+    0xdf, 0x1b, 0xb9, 0x27, 0xe1, 0xbb, 0x7d, 0x17, 0x23, 0xd1, 
+    0x9b, 0x49, 0x20, 0xb2, 0x3b, 0x86, 0x9e, 0xae, 0x86, 0xfd, 
+    0xc1, 0xbf, 0x4d, 0x31, 0x35, 0xc7, 0xd1, 0x49, 0xe7, 0x8d, 
+    0x83, 0xd2, 0xcc, 0x75, 0x3a, 0x25, 0x69, 0x82, 0x68, 0x72, 
+    0x46, 0x3f, 0x6d, 0xa4, 0xfa, 0x73, 0xa5, 0x30, 0x5c, 0x2f, 
+    0xd8, 0x61, 0xa0, 0x1e, 0x75, 0xcc, 0x8f, 0x56, 0x8c, 0x13, 
+    0xc6, 0x81, 0x69, 0x3b, 0x18, 0x86, 0x04, 0x08, 0x6b, 0xdd, 
+    0x89, 0x08, 0x6a, 0x09, 0x7f, 0xd6, 0x96, 0x86, 0x60, 0x1e, 
+    0x1a, 0x5b, 0x5d, 0xc3, 0x6a, 0x40, 0xc8, 0x4c, 0x49, 0x3b, 
+    0xed, 0xfa, 0xe0, 0x59, 0x5a, 0x85, 0x62, 0xa2, 0xce, 0x1b, 
+    0x02, 0xa0, 0x8f, 0x3b, 0xdb, 0x54, 0xe4, 0x83, 0x99, 0xb5, 
+    0xdd, 0xb0, 0xb0, 0x0b, 0xb0, 0xfa, 0x63, 0x5c, 0xab, 0x9b, 
+    0xe7, 0x21, 0xa4, 0x31, 0xa9, 0x59, 0x62, 0x07, 0x6f, 0x9c, 
+    0x01, 0x4a, 0x33, 0xe3, 0xa4, 0xba, 0xae, 0x8a, 0xf7, 0x4c, 
+    0x3f, 0x06, 0xa1, 0x61, 0x2b, 0xa5, 0xc3, 0xb1, 0xe0, 0xe6, 
+    0x42, 0x40, 0x21, 0x3b, 0x00, 0x3f, 0xbd, 0x2e, 0xfd, 0xfc, 
+    0xa8, 0x93, 0x2c, 0x9f, 0x33, 0x1b, 0x1b, 0xe0, 0x59, 0xa5, 
+    0xf6, 0xf4, 0x6b, 0x79, 0x07, 0x95, 0x55, 0x0e, 0x4c, 0x2c, 
+    0x9c, 0x65, 0xca, 0xb5, 0x0b, 0x5e, 0x8a, 0xaf, 0xf9, 0x48, 
+    0xad, 0xee, 0x75, 0x5d, 0xf4, 0x18, 0xf2, 0x24, 0x80, 0xd1, 
+    0x84, 0xd5, 0x72, 0x1e, 0xe9, 0x1a, 0xbc, 0x08, 0x44, 0xbd, 
+    0x4c, 0xf3, 0xa9, 0xaa, 0x83, 0xe2, 0xb8, 0x75, 0x89, 0x00, 
+    0x22, 0x84, 0x76, 0x0b, 0xf9, 0x11, 0xd9, 0x23, 0xd6, 0xc3, 
+    0x02, 0x5d, 0x80, 0xc9, 0x0d, 0xa0, 0x01, 0xf1, 0xfe, 0x5c, 
+    0x70, 0xad, 0x47, 0x58, 0xb4, 0x0b, 0x6e, 0xe6, 0xd6, 0xbc, 
+    0x21, 0xfd, 0xd6, 0x2d, 0x8c, 0x98, 0x56, 0x57, 0x6a, 0xaa, 
+    0x3f, 0xf2, 0x72, 0x68, 0x97, 0x8a, 0x38, 0x8b, 0xd4, 0xf6, 
+    0xe7, 0x97, 0x77, 0xf5, 0xfe, 0x3f, 0x74, 0x1b, 0xfa, 0xbc, 
+    0x43, 0xc9, 0xc5, 0x46, 0x3c, 0xa1, 0x1f, 0x20, 0x9a, 0xe0, 
+    0xcb, 0xa4, 0x92, 0x3e, 0xd4, 0xf7, 0xdb, 0x72, 0x17, 0x6a, 
+    0x7c, 0xb5, 0xed, 0x96, 0xf8, 0x08, 0xf9, 0x95, 0x1f, 0xa4, 
+    0xb1, 0xeb, 0x43, 0x7b, 0x54, 0x8b, 0x81, 0x03, 0xe9, 0xf0, 
+    0x27, 0x82, 0x12, 0x38, 0x2f, 0xdf, 0x2b, 0x84, 0x7f, 0x4b, 
+    0x21, 0x05, 0x73, 0x1c, 0x14, 0xfb, 0xd2, 0x8e, 0xfe, 0x26, 
+    0x56, 0xed, 0x69, 0x91, 0x04, 0x53, 0xea, 0x98, 0xde, 0x6a, 
+    0xd1, 0x36, 0xb5, 0x8e, 0x63, 0x85, 0xf1, 0x3c, 0xcb, 0xe5, 
+    0x10, 0x4e, 0xd3, 0x36, 0xb9, 0xab, 0x03, 0x39, 0x09, 0x8e, 
+    0xf5, 0xd3, 0xf7, 0x35, 0x70, 0x3b, 0x0b, 0x98, 0xb7, 0xb6, 
+    0x58, 0x31, 0x20, 0xed, 0x10, 0x7a, 0x63, 0x9a, 0x3c, 0x3d, 
+    0x4e, 0x34, 0x40, 0x1d, 0xdb, 0x43, 0x2f, 0x6f, 0x74, 0x51, 
+    0xc3, 0x77, 0x76, 0xe8, 0x5d, 0x0d, 0x0f, 0x22, 0xd7, 0x3f, 
+    0xfb, 0xf5, 0xce, 0x6a, 0x60, 0x81, 0x42, 0x64, 0x81, 0x45, 
+    0x24, 0xeb, 0x81, 0xf8, 0xe5, 0x13, 0xd7, 0xec, 0x07, 0x1d, 
+    0x5c, 0x42, 0xe2, 0x9b, 0x98, 0x9e, 0x57, 0x0f, 0xed, 0xd8, 
+    0x26, 0x91, 0x8c, 0x50, 0x69, 0x5b, 0x3a, 0x0a, 0x8e, 0x23, 
+    0x20, 0xaa, 0x3e, 0x3d, 0xd5, 0x83, 0x87, 0x8c, 0xe2, 0xe7, 
+    0xbc, 0x24, 0x90, 0x4e, 0x1a, 0x69, 0x18, 0x0a, 0x7e, 0x68, 
+    0x1c, 0x6f, 0xd2, 0x26, 0x76, 0x37, 0xe5, 0x1d, 0xe1, 0x96, 
+    0x42, 0x74, 0x72, 0x16, 0x8b, 0xa3, 0x90, 0xa1, 0x91, 0x48, 
+    0x4b, 0x56, 0x09, 0x2e, 0xea, 0xa3, 0x08, 0xd9, 0x2f, 0x04, 
+    0x51, 0x60, 0x65, 0x3b, 0x4b, 0x95, 0xba, 0xe8, 0x87, 0x16, 
+    0xe5, 0x3f, 0xb8, 0x42, 0x08, 0xdb, 0x82, 0x7b, 0xe7, 0xf6, 
+    0xb7, 0x6d, 0xc2, 0x82, 0xb6, 0x4f, 0xac, 0xcc, 0x70, 0x2f, 
+    0x56, 0xbc, 0xd4, 0x79, 0x96, 0xca, 0x02, 0x38, 0x40, 0xd3, 
+    0x0d, 0x8b, 0xd9, 0x53, 0x4d, 0xd7, 0xe6, 0x96, 0x70, 0x71, 
+    0x6a, 0xf5, 0xd3, 0xdb, 0xa7, 0x42, 0xe7, 0x6c, 0x04, 0xf1, 
+    0x8c, 0xe4, 0xf4, 0x57, 0x32, 0xf6, 0x94, 0x80, 0x5e, 0xa3, 
+    0xe2, 0x2e, 0x59, 0xb3, 0xf9, 0x4e, 0x40, 0x9a, 0x72, 0x7d, 
+    0xa5, 0xc8, 0x83, 0xc1, 0xae, 0x84, 0x41, 0x3e, 0x9b, 0xa6, 
+    0x9b, 0x5b, 0xf9, 0xc1, 0xe4, 0xad, 0x0f, 0xc7, 0x3a, 0x62, 
+    0xd0, 0x0b, 0x43, 0xd0, 0x59, 0x91, 0x25, 0x26, 0x91, 0x0a, 
+    0xe1, 0x44, 0xdf, 0xdf, 0xbb, 0xca, 0xec, 0x04, 0x8a, 0xb7, 
+    0x21, 0x18, 0x93, 0x5a, 0xac, 0x03, 0xe4, 0x6a, 0xf5, 0x2b, 
+    0x2b, 0xb5, 0xc9, 0x07, 0x62, 0x44, 0xf6, 0x2e, 0xc5, 0x03, 
+    0xb7, 0xb1, 0xd3, 0x18, 0xe9, 0xd3, 0x93, 0xd1, 0x52, 0x8c, 
+    0x36, 0xd1, 0x89, 0xfd, 0xca, 0x71, 0x42, 0x36, 0xd0, 0x3b, 
+    0x7b, 0x9b, 0xcb, 0x22, 0xf8, 0x52, 0xcc, 0x56, 0xb8, 0xfe, 
+    0x0d, 0xc2, 0x49, 0x0a, 0x4d, 0xc6, 0x66, 0x52, 0xa4, 0x31, 
+    0x3c, 0x79, 0x9c, 0xd4, 0x9b, 0x84, 0x13, 0xec, 0x8f, 0xa1, 
+    0x70, 0xd2, 0xf7, 0x73, 0x71, 0x47, 0x58, 0xef, 0x29, 0x40, 
+    0x48, 0x76, 0x11, 0x49, 0x9f, 0xf9, 0x6f, 0x93, 0xcd, 0xa4, 
+    0x3e, 0xea, 0xf3, 0xfd, 0x14, 0xa0, 0xce, 0x0e, 0x29, 0x53, 
+    0xfb, 0xbc, 0xfe, 0xcf, 0x1f, 0x62, 0x26, 0x65, 0xd1, 0x57, 
+    0x6e, 0x6f, 0xac, 0x29, 0x83, 0x44, 0x2e, 0x56, 0x48, 0xba, 
+    0xca, 0x1f, 0x60, 0x54, 0xd6, 0x50, 0xf9, 0xc8, 0xdf, 0xd0, 
+    0xa4, 0xe1, 0x49, 0x0e, 0x30, 0x1b, 0x7b, 0xb5, 0x78, 0x17, 
+    0xa4, 0x94, 0x0a, 0x5f, 0xd3, 0x12, 0x4a, 0x0c, 0xb1, 0x35, 
+    0xdd, 0x1c, 0x54, 0x83, 0xc3, 0x85, 0x06, 0x4b, 0x18, 0x46, 
+    0x20, 0xf4, 0x24, 0x77, 0x22, 0x34, 0x58, 0x93, 0xf2, 0xd8, 
+    0x03, 0xac, 0x64, 0x08, 0x88, 0xb8, 0x9a, 0x21, 0x94, 0xda, 
+    0xe1, 0x64, 0xc8, 0x2b, 0x30, 0xa8, 0xc4, 0xb9, 0x54, 0x6e, 
+    0xef, 0x5b, 0x5d, 0x62, 0x6e, 0xda, 0x76, 0xf6, 0x59, 0x54, 
+    0x1f, 0x62, 0x1c, 0xcf, 0xef, 0x84, 0x8e, 0x04, 0x6a, 0x62, 
+    0x28, 0x39, 0x24, 0xae, 0xa9, 0x33, 0x48, 0x69, 0x0f, 0x63, 
+    0x65, 0x01, 0x2a, 0xb6, 0x15, 0xf0, 0x73, 0x09, 0xce, 0x81, 
+    0xfb, 0x3d, 0x81, 0x2e, 0x46, 0x35, 0x33, 0xf9, 0x59, 0x99, 
+    0xd5, 0x33, 0x42, 0x08, 0xbd, 0xfe, 0x59, 0xe1, 0x98, 0x7d, 
+    0xdd, 0x46, 0x44, 0x6e, 0x6d, 0x08, 0xf2, 0xfa, 0xdd, 0xce, 
+    0xb1, 0x94, 0x09, 0x6c, 0x21, 0xb2, 0x37, 0xf4, 0x46, 0x36, 
+    0xdc, 0xb5, 0xcc, 0xd5, 0x68, 0x5d, 0x8a, 0x0e, 0xc3, 0xa4, 
+    0xab, 0xd5, 0xf0, 0x70, 0xe7, 0x80, 0xdc, 0xcb, 0xe2, 0xba, 
+    0x79, 0xa9, 0xed, 0x91, 0x4b, 0x20, 0x5e, 0xab, 0x6f, 0xb7, 
+    0xa2, 0x09, 0x18, 0x57, 0xad, 0xee, 0xc9, 0xa1, 0x4e, 0xa0, 
+    0x78, 0xb5, 0x37, 0x87, 0xad, 0xee, 0xf0, 0x5e, 0x28, 0x40, 
+    0x4b, 0xab, 0x86, 0xb4, 0x36, 0x83, 0x48, 0xe5, 0xc0, 0x9f, 
+    0x1d, 0x3f, 0xdc, 0x19, 0xf7, 0xd1, 0x7b, 0x9c, 0xe5, 0xe6, 
+    0x06, 0xbb, 0x95, 0xae, 0x8d, 0xea, 0x22, 0x3e, 0x49, 0x23, 
+    0xf0, 0x9c, 0xa4, 0x10, 0x41, 0x12, 0x59, 0x60, 0xad, 0xbb, 
+    0xa6, 0xf1, 0xec, 0xb2, 0xb5, 0xfe, 0x2e, 0x35, 0x26, 0x7e, 
+    0x6e, 0x06, 0xc7, 0xa2, 0x61, 0xe3, 0x22, 0x3f, 0x89, 0x8e, 
+    0x6a, 0xeb, 0xe3, 0x2c, 0xcf, 0xcd, 0xb0, 0xd6, 0x65, 0xd0, 
+    0x9b, 0xd5, 0x6a, 0x79, 0xd6, 0xf7, 0x3e, 0x87, 0xde, 0x15, 
+    0xb6, 0x19, 0x79, 0xc3, 0xbf, 0x7b, 0xc0, 0xf6, 0x90, 0x5c, 
+    0x50, 0x2a, 0x85, 0x03, 0x82, 0x56, 0x31, 0x62, 0x78, 0x84, 
+    0x14, 0x8e, 0x3e, 0x0c, 0x3f, 0x6d, 0x49, 0x9b, 0x33, 0x82, 
+    0x19, 0x39, 0x2f, 0x40, 0x8e, 0x20, 0x71, 0x64, 0x6a, 0x07, 
+    0x55, 0xed, 0xd4, 0xc9, 0x77, 0xa8, 0xf2, 0xb2, 0x40, 0xb3, 
+    0x91, 0xa1, 0xa5, 0xcf, 0x4d, 0x43, 0x3a, 0x7c, 0xa7, 0xa9, 
+    0xe0, 0x10, 0x60, 0xf3, 0x10, 0xb9, 0xb8, 0x8b, 0x93, 0x23, 
+    0x58, 0x18, 0x0f, 0x20, 0x87, 0x1f, 0x3f, 0x5e, 0xc5, 0x66, 
+    0xdd, 0x06, 0x13, 0x54, 0x8d, 0xfe, 0x83, 0xeb, 0x19, 0x3e, 
+    0xd3, 0x7b, 0x51, 0x11, 0xcd, 0xc8, 0x4d, 0x93, 0x50, 0x63, 
+    0x66, 0xa7, 0xcd, 0x61, 0x1a, 0x03, 0xdb, 0xcf, 0x7d, 0x4a, 
+    0x88, 0xd3, 0xa4, 0x10, 0x1f, 0x28, 0x31, 0xe0, 0x8a, 0xdf, 
+    0xd2, 0xac, 0x48, 0xd3, 0x8e, 0x79, 0x69, 0xd3, 0xf5, 0xc1, 
+    0x86, 0x79, 0x71, 0x73, 0x1f, 0x47, 0x7e, 0x8b, 0x3c, 0x12, 
+    0x24, 0xfe, 0x98, 0xdf, 0xc4, 0x10, 0xae, 0x84, 0xb8, 0xfe, 
+    0xa4, 0x49, 0x67, 0xf9, 0xbd, 0xa7, 0x03, 0x1b, 0xd5, 0x40, 
+    0x3a, 0xd5, 0xb7, 0x5c, 0xd4, 0x94, 0x1e, 0x96, 0x46, 0x01, 
+    0xf8, 0xb0, 0x20, 0xcf, 0x02, 0x59, 0xcf, 0x11, 0x37, 0xd2, 
+    0xe7, 0x56, 0xc9, 0x89, 0xc7, 0x77, 0x0b, 0x24, 0x04, 0xca, 
+    0xa6, 0x44, 0x43, 0xda, 0x28, 0x86, 0x9f, 0x5b, 0x06, 0x1b, 
+    0xc4, 0xcd, 0xa0, 0x9c, 0x89, 0x18, 0x03, 0x47, 0x50, 0x6d, 
+    0xe8, 0xcd, 0x18, 0x53, 0x10, 0xef, 0xa2, 0xa2, 0xf6, 0xcf, 
+    0xdf, 0x92, 0x1a, 0x50, 0x3f, 0xdb, 0x8e, 0x58, 0x40, 0x24, 
+    0xeb, 0xd0, 0xbd, 0x5f, 0x8b, 0x01, 0xb1, 0x65, 0x59, 0x18, 
+    0xd9, 0x62, 0xda, 0x70, 0x64, 0x10, 0x15, 0xce, 0x0b, 0x26, 
+    0xc0, 0x98, 0xd2, 0xa9, 0x19, 0x65, 0xdb, 0x22, 0x77, 0x8f, 
+    0x80, 0x79, 0x51, 0xc4, 0xa4, 0x67, 0xcf, 0x89, 0x8c, 0x9a, 
+    0xd3, 0x80, 0x2f, 0x18, 0x78, 0x26, 0x16, 0xeb, 0xe0, 0x17, 
+    0x69, 0xa5, 0x17, 0x9c, 0xa2, 0x8f, 0x08, 0x74, 0xad, 0x67, 
+    0x8f, 0xaa, 0x91, 0x0a, 0x9e, 0x56, 0x83, 0x14, 0x36, 0x34, 
+    0xcc, 0xef, 0xfe, 0x17, 0x9f, 0xc3, 0xdd, 0x12, 0x5c, 0x23, 
+    0xa7, 0x70, 0x58, 0x48, 0x06, 0x9c, 0x36, 0xe5, 0x2a, 0xfa, 
+    0x65, 0xe3, 0xd4, 0xdb, 0x48, 0xd7, 0xc6, 0xd0, 0xc8, 0x0a, 
+    0x29, 0x01, 0x76, 0x0a, 0x7f, 0x08, 0xd6, 0xda, 0x50, 0x0f, 
+    0x37, 0x30, 0xc4, 0xdd, 0x4e, 0x4a, 0x1f, 0x05, 0x0c, 0x30, 
+    0x6b, 0xa1, 0x55, 0x5a, 0xc8, 0x1d, 0x05, 0x81, 0xc6, 0x66, 
+    0xc0, 0x05, 0xd6, 0x65, 0xde, 0x29, 0x8e, 0xf3, 0xc7, 0x50, 
+    0x9a, 0x0e, 0x70, 0xbf, 0x32, 0xec, 0xa9, 0x3e, 0xf2, 0x09, 
+    0x36, 0xdb, 0xd9, 0x6f, 0xfc, 0x95, 0xb4, 0x8b, 0x4d, 0x4a, 
+    0x1a, 0xce, 0x31, 0x3f, 0xf3, 0x8b, 0x2f, 0x86, 0xb0, 0x34, 
+    0xa8, 0x0d, 0x5f, 0x19, 0x16, 0x18, 0x57, 0xf0, 0xa7, 0x4d, 
+    0x00, 0xc7, 0x81, 0x4a, 0x38, 0x5a, 0x3e, 0x03, 0x53, 0x87, 
+    0xd1, 0x55, 0xc8, 0x9b, 0xe7, 0xe5, 0x2f, 0xd8, 0x22, 0x3d, 
+    0xe2, 0xe2, 0xa1, 0x09, 0xf7, 0xc3, 0x77, 0x18, 0x06, 0xd6, 
+    0x45, 0x65, 0x3c, 0x53, 0xe0, 0x49, 0x80, 0x83, 0x52, 0x95, 
+    0x91, 0x51, 0x03, 0xb4, 0x06, 0x43, 0x84, 0x75, 0x90, 0xd9, 
+    0xfe, 0x3e, 0xc4, 0x0f, 0xee, 0x2d, 0x5e, 0xdc, 0x40, 0xaf, 
+    0x83, 0xb8, 0x77, 0x70, 0x7d, 0x68, 0xef, 0xe8, 0x58, 0x4a, 
+    0x2d, 0x15, 0xba, 0x41, 0x22, 0xd4, 0xc0, 0xf3, 0x00, 0x3f, 
+    0xb3, 0x13, 0x06, 0xe5, 0xe8, 0xfe, 0xfc, 0x40, 0xda, 0x39, 
+    0x38, 0x74, 0x87, 0xf2, 0x55, 0x01, 0x7c, 0x84, 0x0f, 0x16, 
+    0x52, 0x80, 0x0a, 0x91, 0xe8, 0x0d, 0xde, 0xc6, 0xeb, 0x5d, 
+    0x06, 0x5e, 0x89, 0x98, 0x60, 0xc9, 0x31, 0x38, 0x20, 0x29, 
+    0xe1, 0xcc, 0x1c, 0x29, 0xc8, 0xb6, 0xf3, 0x6b, 0x9b, 0x84, 
+    0xf6, 0x62, 0x4e, 0x5c, 0xa2, 0x2a, 0x7a, 0x71, 0xf6, 0x6f, 
+    0xb0, 0xa3, 0x8e, 0x37, 0x79, 0xa7, 0x35, 0x85, 0xbf, 0xc4, 
+    0x59, 0x51, 0xb8, 0x50, 0x1a, 0xd4, 0x30, 0x88, 0x14, 0x0d, 
+    0x21, 0x7f, 0x5c, 0xe0, 0xf3, 0x5d, 0x02, 0x92, 0xee, 0x68, 
+    0x8b, 0x07, 0x71, 0x71, 0xd9, 0x93, 0x51, 0x5e, 0xb4, 0xe5, 
+    0x50, 0x12, 0x1e, 0x14, 0x7b, 0x4d, 0xf8, 0xb0, 0x1a, 0x59, 
+    0x96, 0x9b, 0x8d, 0x3e, 0x6d, 0x6b, 0x8d, 0xf1, 0x1f, 0x4f, 
+    0x61, 0x22, 0x15, 0xa4, 0x61, 0x25, 0x8d, 0x4d, 0x86, 0xa8, 
+    0xfd, 0xf4, 0x7a, 0xdb, 0x50, 0x9c, 0xce, 0xe8, 0x9e, 0x19, 
+    0xe7, 0x4e, 0x28, 0x7b, 0x06, 0xab, 0x56, 0x22, 0xea, 0xa0, 
+    0xb2, 0x33, 0xb1, 0x4f, 0x6e, 0x29, 0xa4, 0x40, 0x3d, 0xa3, 
+    0x0e, 0x5b, 0x0e, 0xdc, 0x4b, 0x63, 0x66, 0xee, 0x0f, 0x1b, 
+    0xee, 0x74, 0x04, 0xd9, 0x23, 0x76, 0xf7, 0x01, 0x2e, 0x91, 
+    0x48, 0x2e, 0xc1, 0x02, 0x7b, 0x97, 0xde, 0x27, 0x81, 0x53, 
+    0x29, 0xd7, 0x6c, 0xe4, 0x9d, 0x42, 0xcf, 0xea, 0x3d, 0xf8, 
+    0xbb, 0xf2, 0x58, 0x7b, 0x06, 0x09, 0x57, 0xb8, 0xb2, 0x37, 
+    0x62, 0xc1, 0x92, 0xd9, 0xea, 0x0c, 0x54, 0xe6, 0xcd, 0x05, 
+    0x78, 0xe3, 0x90, 0x0d, 0x1e, 0xeb, 0x8e, 0x80, 0x9b, 0x14, 
+    0x0a, 0xae, 0xa7, 0x5c, 0xab, 0x91, 0x55, 0x6f, 0x8c, 0x24, 
+    0xe8, 0x13, 0x78, 0x39, 0x7f, 0x85, 0x22, 0xcc, 0xde, 0x6a, 
+    0x3a, 0x6a, 0x7b, 0x9e, 0x07, 0x2d, 0xb9, 0x7c, 0x01, 0x17, 
+    0x59, 0x2a, 0x7f, 0xd6, 0x0c, 0xe0, 0xdc, 0xc4, 0x55, 0xdd, 
+    0x1c, 0x30, 0x75, 0xdb, 0x81, 0x96, 0xe9, 0x90, 0xa6, 0x48, 
+    0x4a, 0x88, 0x76, 0xdb, 0x3a, 0xc7, 0xd0, 0x1b, 0x9f, 0x9d, 
+    0x39, 0xba, 0x70, 0x8a, 0xbe, 0xdc, 0x8f, 0xd0, 0x42, 0x6c, 
+    0x67, 0xf8, 0x7f, 0x49, 0x4e, 0x5d, 0x22, 0xee, 0x02, 0xec, 
+    0x29, 0xb0, 0xb1, 0xf8, 0x06, 0x16, 0x67, 0xd2, 0x3f, 0x8e, 
+    0x88, 0x25, 0x3b, 0x96, 0x89, 0x49, 0xf4, 0x58, 0xb0, 0xd4, 
+    0x05, 0xfa, 0x48, 0x8f, 0x40, 0xcd, 0xa5, 0xae, 0x3d, 0xec, 
+    0x3a, 0x26, 0xa4, 0x71, 0x51, 0x19, 0x1d, 0x5a, 0x2a, 0x08, 
+    0xc8, 0x12, 0x5a, 0x19, 0x66, 0x84, 0xa6, 0xb6, 0xd3, 0xc0, 
+    0x34, 0x22, 0xe9, 0xfe, 0x84, 0x03, 0x09, 0x87, 0xbe, 0xa9, 
+    0xb5, 0x2e, 0xdb, 0xa4, 0xdb, 0x2b, 0x94, 0xc4, 0xd1, 0x72, 
+    0xf3, 0x87, 0xad, 0x49, 0x10, 0x8d, 0x65, 0x27, 0x82, 0x12, 
+    0x96, 0x3d, 0x7c, 0x6f, 0xd2, 0xad, 0x9d, 0xfb, 0x0e, 0x51, 
+    0x3e, 0x5a, 0x4d, 0x0f, 0xb9, 0x3e, 0x67, 0xce, 0x2b, 0x90, 
+    0xbe, 0x11, 0x9f, 0xd2, 0x42, 0x15, 0x57, 0x31, 0x9a, 0x40, 
+    0x98, 0x60, 0x24, 0xce, 0x95, 0x9e, 0x06, 0xeb, 0x09, 0xd1, 
+    0x16, 0x9b, 0x3b, 0xbf, 0x8c, 0x55, 0x90, 0x90, 0x63, 0x4b, 
+    0x4e, 0x6b, 0xc2, 0x50, 0xe2, 0x5f, 0x48, 0x8d, 0xc2, 0x3f, 
+    0xc4, 0xf0, 0x19, 0xbe, 0x6c, 0xb2, 0x1b, 0xe7, 0xcd, 0xa5, 
+    0x1f, 0xa6, 0x10, 0xeb, 0x33, 0x41, 0xf9, 0x30, 0x4c, 0x8d, 
+    0x23, 0xc3, 0xa0, 0xee, 0x77, 0x90, 0x93, 0x75, 0x8d, 0x6b, 
+    0x8f, 0xea, 0x33, 0xf3, 0xc1, 0xea, 0x57, 0x72, 0x39, 0x2d, 
+    0x5c, 0xa6, 0xb8, 0xe6, 0x9e, 0x5d, 0x89, 0x2f, 0xbe, 0xf5, 
+    0x1f, 0x67, 0x8c, 0xf0, 0x06, 0x74, 0x3a, 0x83, 0x24, 0x63, 
+    0xf7, 0x41, 0x36, 0xa3, 0x6e, 0x4c, 0x3a, 0xe6, 0x45, 0x92, 
+    0xe7, 0x54, 0x28, 0xe2, 0x25, 0xaf, 0x3a, 0xd8, 0xe4, 0x29, 
+    0x32, 0x43, 0x67, 0x12, 0x34, 0x34, 0x3b, 0x34, 0xea, 0xed, 
+    0x7d, 0xd6, 0x1d, 0x86, 0x22, 0xd8, 0x89, 0xb5, 0xdc, 0x23, 
+    0x57, 0x75, 0x50, 0xc5, 0x08, 0x24, 0xc7, 0x1a, 0x94, 0xbe, 
+    0x30, 0x35, 0x7d, 0x53, 0x55, 0x5f, 0xe0, 0x16, 0x43, 0x1c, 
+    0x82, 0xa6, 0x83, 0xfb, 0x57, 0x5b, 0x38, 0x8f, 0x43, 0xc7, 
+    0x17, 0x0b, 0x33, 0x3c, 0x96, 0x7e, 0xce, 0x10, 0x8b, 0x60, 
+    0xab, 0xa2, 0x26, 0x6d, 0x78, 0x26, 0x5b, 0xb8, 0xd7, 0x58, 
+    0x8c, 0x09, 0x2f, 0x3a, 0x61, 0x1f, 0x65, 0xe1, 0xcb, 0xfd, 
+    0x9c, 0x5a, 0x31, 0x09, 0x67, 0x93, 0xed, 0x91, 0xa7, 0x4b, 
+    0xaa, 0xe5, 0xa1, 0xb9, 0x99, 0x15, 0x7f, 0xe0, 0xa6, 0x3b, 
+    0x6e, 0x79, 0x5d, 0xe2, 0x05, 0xd6, 0x25, 0xe8, 0xbb, 0x26, 
+    0xc1, 0x0f, 0x88, 0xce, 0xd7, 0xfd, 0xcf, 0x34, 0x89, 0xd1, 
+    0xb2, 0xdb, 0xf6, 0xba, 0xc0, 0x81, 0x67, 0x8f, 0x9a, 0xdf, 
+    0x05, 0xe2, 0x56, 0xdd, 0xdd, 0xee, 0xa5, 0x4c, 0x74, 0xc6, 
+    0x7b, 0x99, 0xec, 0xf5, 0x0b, 0x6c, 0xd1, 0xbc, 0x1c, 0x29, 
+    0x58, 0x0a, 0x13, 0xe3, 0x13, 0xb9, 0x77, 0x4c, 0xc1, 0x96, 
+    0xb7, 0xa8, 0x3f, 0x6d, 0x1a, 0x8c, 0x9d, 0xae, 0xa2, 0x08, 
+    0x09, 0x1c, 0x52, 0xe2, 0xa4, 0xd8, 0x67, 0x6a, 0xdf, 0x73, 
+    0xfc, 0x2a, 0x26, 0xce, 0x29, 0x5e, 0x06, 0x38, 0x43, 0xcc, 
+    0xb8, 0x34, 0xe0, 0x57, 0xbb, 0x15, 0x11, 0xf7, 0x64, 0x5c, 
+    0x36, 0x7a, 0x71, 0x92, 0x9d, 0xb2, 0x83, 0xb1, 0xa1, 0xdb, 
+    0xbd, 0xe5, 0xcd, 0xe1, 0x0b, 0x73, 0x71, 0x98, 0x34, 0xde, 
+    0x2d, 0x4d, 0x57, 0x10, 0x83, 0xe2, 0x6f, 0xe0, 0x77, 0xa1, 
+    0x7c, 0x75, 0x8e, 0x92, 0xea, 0xec, 0x40, 0x8d, 0xc2, 0x86, 
+    0x6b, 0x09, 0xd5, 0x30, 0x97, 0xb5, 0xc3, 0x1a, 0x79, 0x41, 
+    0x37, 0x87, 0x2e, 0xcc, 0x08, 0xb0, 0x04, 0xf2, 0x24, 0x04, 
+    0x23, 0xab, 0x6a, 0x0d, 0xf6, 0x5d, 0x41, 0xfb, 0x9d, 0x2b, 
+    0x0e, 0xf9, 0xee, 0x49, 0x53, 0x59, 0x8b, 0xc4, 0x9c, 0x42, 
+    0xcf, 0xec, 0x60, 0xb3, 0xd1, 0xe6, 0x74, 0x5a, 0xd3, 0x2e, 
+    0x85, 0xbe, 0x86, 0xb1, 0x94, 0x4e, 0x2b, 0xaf, 0x1b, 0xc3, 
+    0xe6, 0xdd, 0x0f, 0x5e, 0x6d, 0x46, 0x86, 0x8a, 0xe4, 0x32, 
+    0x93, 0x46, 0x32, 0x12, 0x48, 0x16, 0x2a, 0x81, 0x50, 0x95, 
+    0x40, 0xaf, 0x2a, 0xe6, 0x3b, 0x2e, 0xab, 0x91, 0xb2, 0x3d, 
+    0x81, 0x54, 0xd1, 0x8a, 0xc4, 0xb9, 0xed, 0x0a, 0x42, 0xdf, 
+    0xc5, 0x89, 0x1e, 0x3f, 0x61, 0xc5, 0x1e, 0x7c, 0xb3, 0x3d, 
+    0x0c, 0x95, 0x55, 0x34, 0xfa, 0x18, 0xce, 0x91, 0x5e, 0xb3, 
+    0x57, 0xfd, 0xcb, 0x5d, 0xaa, 0x38, 0x81, 0xae, 0xf4, 0xbf, 
+    0xf4, 0xc2, 0x38, 0x41, 0x68, 0x2c, 0x27, 0xeb, 0xfc, 0xc5, 
+    0xa2, 0x5c, 0x51, 0x92, 0x65, 0xf1, 0x45, 0xf9, 0xbd, 0x55, 
+    0x31, 0xb3, 0xa4, 0x56, 0x14, 0x92, 0xa7, 0x18, 0xcd, 0x54, 
+    0x80, 0x6e, 0x68, 0x39, 0xec, 0x7f, 0x3e, 0x91, 0x52, 0xfd, 
+    0xca, 0x05, 0x10, 0xb5, 0xe2, 0x23, 0x4e, 0x3f, 0xe7, 0xc6, 
+    0xef, 0xa0, 0x06, 0xe0, 0x6f, 0x36, 0x2b, 0xd1, 0xf7, 0x7d, 
+    0x07, 0xcd, 0xa9, 0xaf, 0xca, 0x46, 0xf2, 0x5b, 0x75, 0x25, 
+    0xa6, 0xba, 0x16, 0xa8, 0x28, 0xb5, 0xb3, 0x4f, 0xcc, 0x0a, 
+    0xf3, 0xaf, 0x60, 0xbd, 0x8d, 0x50, 0xb3, 0xbd, 0x29, 0xe2, 
+    0x43, 0xd8, 0xbf, 0x68, 0xec, 0x86, 0x3b, 0xac, 0xcc, 0xaa, 
+    0x70, 0xb8, 0x5b, 0xc3, 0xa4, 0xfc, 0xfb, 0x5d, 0x5c, 0x38, 
+    0x09, 0xcf, 0xee, 0x27, 0x87, 0x3c, 0x88, 0x48, 0xa0, 0x67, 
+    0x24, 0xf5, 0xcb, 0x60, 0xae, 0x8c, 0x7c, 0xe6, 0x23, 0x1d, 
+    0x23, 0x67, 0xa5, 0xb9, 0x20, 0xc1, 0xed, 0xae, 0xfd, 0xe9, 
+    0x35, 0x43, 0x99, 0x39, 0x27, 0x55, 0x7c, 0xe4, 0x30, 0x76, 
+    0xc7, 0x68, 0xc2, 0xb9, 0x95, 0xc6, 0x9a, 0x3f, 0x14, 0x1e, 
+    0xd5, 0x1a, 0x7e, 0xf9, 0x76, 0x99, 0x4b, 0x42, 0x5b, 0xfa, 
+    0x78, 0x4a, 0x15, 0xa3, 0xef, 0x2e, 0xae, 0x58, 0x2a, 0x81, 
+    0xb8, 0xf2, 0xb4, 0x49, 0x13, 0x32, 0x20, 0xd2, 0x8d, 0x66, 
+    0xc9, 0x21, 0xf9, 0x1a, 0x60, 0x75, 0x47, 0xc7, 0xab, 0x00, 
+    0xe8, 0xe7, 0xd8, 0xa2, 0x2b, 0x01, 0x48, 0x0d, 0x93, 0xfc, 
+    0x40, 0xc8, 0x88, 0x73, 0xd3, 0x29, 0x73, 0xda, 0x2d, 0x6d, 
+    0xc8, 0xd0, 0x66, 0x78, 0x69, 0x37, 0xc2, 0xd3, 0x1d, 0xab, 
+    0x5d, 0xa2, 0xbc, 0xd5, 0x0d, 0x74, 0xd4, 0xf4, 0x21, 0x86, 
+    0xc6, 0x00, 0x56, 0x28, 0x35, 0xc1, 0xfd, 0xad, 0x6b, 0x17, 
+    0xf0, 0xf4, 0x39, 0x82, 0x3b, 0xd6, 0xce, 0xbc, 0xa0, 0x59, 
+    0xcb, 0xec, 0xe6, 0xf2, 0xb9, 0xcb, 0x0e, 0x56, 0x58, 0x2e, 
+    0x11, 0xf6, 0x4e, 0x0b, 0xa6, 0xde, 0xb5, 0xde, 0x76, 0x32, 
+    0xbf, 0xa3, 0x52, 0x62, 0xbf, 0x3e, 0x13, 0x81, 0x4e, 0x2e, 
+    0xb7, 0x0b, 0xe4, 0xa3, 0x4e, 0xf1, 0x8f, 0x4d, 0x62, 0x2a, 
+    0x0a, 0x5e, 0xe9, 0xe3, 0xf7, 0x7b, 0x81, 0x52, 0xd5, 0x38, 
+    0x6d, 0x1f, 0x21, 0x41, 0x82, 0xbf, 0x58, 0x03, 0x94, 0xa0, 
+    0x89, 0x77, 0xac, 0x02, 0xe5, 0x51, 0x80, 0xd4, 0x37, 0xa7, 
+    0x90, 0x43, 0x47, 0x8d, 0x54, 0xf1, 0xe3, 0xce, 0x84, 0x28, 
+    0xb5, 0xb1, 0xef, 0xf2, 0x03, 0x26, 0xb9, 0x61, 0x8c, 0x74, 
+    0xed, 0x9b, 0xd7, 0x1e, 0xf3, 0x45, 0xe0, 0x8b, 0xa1, 0x8c, 
+    0xfc, 0xb4, 0x9b, 0x4e, 0x42, 0x92, 0x0c, 0xf2, 0x35, 0xf6, 
+    0x1b, 0x3f, 0x21, 0xea, 0x88, 0x8c, 0xce, 0x17, 0x1b, 0x99, 
+    0x13, 0x48, 0xe3, 0x66, 0x6a, 0x2f, 0xf6, 0x9b, 0x18, 0xf3, 
+    0xa3, 0x1c, 0x8c, 0x11, 0x93, 0xfe, 0x98, 0x85, 0x5c, 0x6e, 
+    0x84, 0x6e, 0x53, 0xbc, 0x46, 0xd7, 0x99, 0xea, 0xc1, 0xab, 
+    0x91, 0x61, 0x8f, 0xf5, 0xe6, 0xe7, 0xd1, 0x97, 0x43, 0x44, 
+    0xe7, 0x3a, 0x3c, 0x24, 0xe6, 0x30, 0x9a, 0x2d, 0x3f, 0x4f, 
+    0x67, 0x14, 0x73, 0x4c, 0x87, 0x70, 0xb8, 0x98, 0x03, 0xc8, 
+    0xf1, 0x83, 0x07, 0x93, 0xb4, 0x36, 0xdb, 0xd8, 0xce, 0x39, 
+    0x0d, 0xc3, 0x29, 0x50, 0x08, 0xe7, 0xb5, 0x51, 0xed, 0xb6, 
+    0xca, 0x0d, 0x67, 0x45, 0xed, 0x61, 0xdf, 0x39, 0x3a, 0x63, 
+    0xd6, 0x9a, 0x9c, 0x95, 0x13, 0x58, 0x0b, 0x80, 0x41, 0x4b, 
+    0x9e, 0x56, 0xd2, 0xb6, 0x81, 0x39, 0x2b, 0x94, 0x48, 0xc3, 
+    0x81, 0xbe, 0x09, 0x1a, 0xd3, 0x98, 0xb2, 0xfa, 0x47, 0x1b, 
+    0x26, 0xd8, 0xe2, 0x46, 0xe0, 0x46, 0x6e, 0xb9, 0x6a, 0x54, 
+    0x37, 0xa5, 0x67, 0x11, 0xfc, 0xca, 0x2b, 0xe3, 0xb6, 0xd1, 
+    0x4a, 0x99, 0x32, 0x28, 0x0a, 0xdb, 0xfd, 0x6a, 0xb1, 0xab, 
+    0xd5, 0x93, 0x39, 0x24, 0x0c, 0x7a, 0xb4, 0xb5, 0xe1, 0x1e, 
+    0x8c, 0x53, 0x65, 0xca, 0x2c, 0x17, 0x63, 0xba, 0x45, 0x00, 
+    0xdd, 0xbc, 0x60, 0xda, 0x76, 0x20, 0x7b, 0x99, 0xe2, 0xa2, 
+    0x03, 0x8b, 0xb5, 0x44, 0xa3, 0xae, 0x82, 0x8c, 0xdd, 0xfb, 
+    0xbd, 0xb3, 0xc6, 0x04, 0xa6, 0x52, 0x6e, 0x37, 0x3f, 0x92, 
+    0x64, 0x26, 0xc8, 0x0e, 0xd3, 0x8e, 0x71, 0x85, 0x0b, 0x6c, 
+    0xf2, 0x15, 0x78, 0x0a, 0x44, 0x84, 0xcf, 0xd8, 0x9c, 0x88, 
+    0x6a, 0x53, 0x1f, 0x74, 0xd5, 0xdd, 0xb1, 0x4a, 0x11, 0x18, 
+    0xb1, 0x72, 0xc4, 0x7c, 0xe9, 0x83, 0x30, 0x4f, 0xdd, 0xf5, 
+    0x74, 0xb5, 0x5c, 0xa7, 0x07, 0xfc, 0xe5, 0xd8, 0x1f, 0xf2, 
+    0x4a, 0x71, 0xe6, 0x8d, 0xf8, 0xbc, 0x91, 0x8e, 0x1d, 0x12, 
+    0x11, 0xca, 0xf2, 0x17, 0x76, 0x29, 0x03, 0x92, 0x24, 0x17, 
+    0xba, 0xcc, 0x1f, 0x3e, 0xb3, 0xdb, 0xbf, 0xe4, 0x04, 0xd6, 
+    0x12, 0xa2, 0xb2, 0x2e, 0x93, 0x89, 0x5e, 0x8f, 0xb6, 0x8c, 
+    0xf7, 0x80, 0xfc, 0x17, 0xf7, 0x3f, 0x28, 0xaa, 0xc5, 0x8b, 
+    0xa2, 0x84, 0x9a, 0x9f, 0xde, 0x05, 0x0f, 0xa1, 0x2d, 0x05, 
+    0x75, 0xed, 0x53, 0x96, 0xbc, 0xba, 0xcd, 0x99, 0xf7, 0xbb, 
+    0xe0, 0x47, 0x02, 0xb4, 0x4c, 0x10, 0xce, 0xd9, 0x6a, 0x27, 
+    0xbe, 0x35, 0x4d, 0xf8, 0x5c, 0x2b, 0xfb, 0x82, 0x72, 0x0c, 
+    0xec, 0x29, 0x95, 0xab, 0xa2, 0x52, 0xe4, 0x54, 0x30, 0x6c, 
+    0xad, 0xdf, 0x76, 0x29, 0x9c, 0x21, 0x0d, 0x06, 0x61, 0x6d, 
+    0x99, 0x3b, 0x51, 0xee, 0xb0, 0x47, 0x65, 0x61, 0xf8, 0x4b, 
+    0xc7, 0xf0, 0x28, 0x7f, 0xc8, 0x24, 0x80, 0x22, 0x6c, 0xc9, 
+    0x73, 0x39, 0x9b, 0x9f, 0xe5, 0x93, 0x14, 0xc4, 0x7c, 0x91, 
+    0x35, 0xec, 0x04, 0xa6, 0x5c, 0xaa, 0xe5, 0xfd, 0x35, 0x81, 
+    0x05, 0x83, 0x54, 0xc4, 0xad, 0x47, 0x0e, 0x60, 0x6c, 0x52, 
+    0x7d, 0x06, 0x76, 0x9e, 0xe6, 0x91, 0xc3, 0xf4, 0x5b, 0xf1, 
+    0xad, 0x1a, 0x51, 0xac, 0x49, 0x10, 0xb2, 0x18, 0x25, 0x06, 
+    0x90, 0x29, 0x46, 0xe3, 0xd5, 0xe1, 0xca, 0x56, 0x1a, 0x6b, 
+    0x24, 0x52, 0xf8, 0x08, 0xab, 0x52, 0x6c, 0x8d, 0xf7, 0x59, 
+    0xaa, 0x0a, 0xf5, 0x37, 0x31, 0x64, 0x3b, 0xb7, 0x12, 0x76, 
+    0xc4, 0xb4, 0x83, 0x9d, 0x76, 0x39, 0x1f, 0x64, 0x85, 0xed, 
+    0x67, 0x98, 0x97, 0x57, 0x76, 0xe9, 0x0d, 0xc8, 0x4f, 0xc4, 
+    0xc7, 0xf8, 0xad, 0x1d, 0x6f, 0xf3, 0xae, 0x4a, 0xf0, 0x3d, 
+    0x71, 0x73, 0xba, 0x74, 0xa7, 0xd0, 0xc7, 0x62, 0x69, 0xb0, 
+    0x5e, 0x25, 0x87, 0xe8, 0xb7, 0xa5, 0x54, 0x54, 0xe2, 0x7e, 
+    0x13, 0xd1, 0x6d, 0x70, 0x5b, 0x65, 0x49, 0x7c, 0x40, 0xdd, 
+    0xba, 0x95, 0x34, 0xc6, 0xa1, 0x01, 0x47, 0x67, 0xa9, 0x42, 
+    0xde, 0xc7, 0xb1, 0x1e, 0x2e, 0xd6, 0x83, 0x1e, 0x34, 0x30, 
+    0xd1, 0x76, 0x85, 0x33, 0x41, 0x72, 0xd5, 0x28, 0xc4, 0xdd, 
+    0x41, 0x78, 0x6a, 0x64, 0xee, 0x45, 0xa1, 0x90, 0x9e, 0xf2, 
+    0x16, 0xb1, 0x4d, 0xd1, 0xbc, 0x11, 0x2b, 0x0b, 0x19, 0xbf, 
+    0x7e, 0x63, 0xf1, 0x40, 0xbd, 0x16, 0x96, 0xb3, 0xa7, 0x1e, 
+    0x0e, 0x8a, 0x92, 0xbb, 0x7d, 0xcd, 0x03, 0x6e, 0x1c, 0x0a, 
+    0x58, 0x46, 0x88, 0xac, 0x80, 0x63, 0xd5, 0x92, 0xa9, 0xeb, 
+    0xec, 0x17, 0x9a, 0x0a, 0xe3, 0x75, 0xfe, 0x1f, 0xb8, 0x70, 
+    0xf5, 0x3c, 0x01, 0x2a, 0x10, 0x9d, 0xd7, 0xef, 0xa7, 0xb2, 
+    0xb5, 0xab, 0x5a, 0x77, 0x53, 0x61, 0x45, 0x9f, 0x1f, 0xca, 
+    0x95, 0xe0, 0x83, 0xf7, 0x11, 0xef, 0x52, 0x3d, 0xcc, 0x55, 
+    0xba, 0x7d, 0x71, 0x49, 0x0c, 0x26, 0x76, 0x89, 0x4c, 0x44, 
+    0x10, 0xbf, 0x64, 0xad, 0x56, 0x16, 0xb1, 0x83, 0xd0, 0x59, 
+    0xa5, 0x2a, 0x6a, 0x94, 0x78, 0xd4, 0x0f, 0x78, 0x1b, 0x28, 
+    0x15, 0xed, 0x05, 0x14, 0x38, 0xbe, 0x67, 0x76, 0x13, 0x83, 
+    0x6b, 0xca, 0x16, 0xd3, 0x38, 0xf2, 0x57, 0xb7, 0x72, 0x21, 
+    0xa6, 0x1e, 0x94, 0x3f, 0x61, 0x0b, 0x34, 0x9b, 0x30, 0x39, 
+    0x4e, 0x3a, 0x89, 0x03, 0x9c, 0x99, 0x2b, 0x9f, 0x28, 0x3f, 
+    0xde, 0x15, 0xbf, 0xb0, 0xfd, 0x3a, 0xda, 0x58, 0x73, 0x25, 
+    0x05, 0xab, 0x1d, 0xcf, 0xbf, 0xe5, 0x76, 0x30, 0xb5, 0x75, 
+    0x22, 0x3e, 0x6b, 0xfb, 0x59, 0x30, 0xe3, 0x3a, 0x62, 0x72, 
+    0xf8, 0xf5, 0x5a, 0xa7, 0x76, 0xbd, 0xef, 0xa4, 0x7b, 0x7e, 
+    0xdf, 0x4f, 0x94, 0x52, 0xd7, 0x1d, 0x12, 0xb2, 0x00, 0x56, 
+    0xa8, 0x55, 0x3c, 0x49, 0x7c, 0x48, 0x10, 0x9b, 0x97, 0x9a, 
+    0xd7, 0xa3, 0xfa, 0x54, 0xbc, 0xf6, 0x53, 0x56, 0x98, 0x7e, 
+    0xce, 0xab, 0x5b, 0x85, 0xe2, 0x2f, 0x95, 0x8c, 0x36, 0xbd, 
+    0x78, 0xe8, 0x13, 0xc5, 0x9b, 0x16, 0x72, 0xea, 0xba, 0x13, 
+    0x36, 0xfb, 0xda, 0xf2, 0x2c, 0x3f, 0x2d, 0x94, 0x1b, 0xf6, 
+    0x18, 0x83, 0x8d, 0x06, 0x30, 0xaa, 0x59, 0x52, 0x9b, 0xbc, 
+    0xbe, 0xe1, 0x79, 0x22, 0xb0, 0xe4, 0xd8, 0x37, 0xad, 0xe9, 
+    0x86, 0xc6, 0x67, 0x73, 0x1c, 0x2b, 0xda, 0x1f, 0xa5, 0x8e, 
+    0x48, 0x42, 0x52, 0x55, 0x34, 0x59, 0x12, 0xfa, 0x4f, 0xfb, 
+    0xb6, 0x6f, 0xb7, 0x9d, 0x7d, 0x2c, 0x0d, 0xc3, 0x2c, 0x25, 
+    0xd5, 0xb3, 0x90, 0xc6, 0x1a, 0x71, 0x3a, 0x8f, 0xa7, 0x5b, 
+    0xf4, 0xc7, 0xb0, 0x44, 0x98, 0x4a, 0x05, 0xa2, 0x57, 0x92, 
+    0x6e, 0xc1, 0xb4, 0xf8, 0xba, 0xd8, 0x47, 0xa4, 0x09, 0xe7, 
+    0xa1, 0x19, 0x67, 0x45, 0x1a, 0x00, 0x5c, 0x3a, 0x18, 0xf3, 
+    0x29, 0x45, 0x66, 0x3d, 0xfd, 0xc6, 0xa6, 0x67, 0x2f, 0xa5, 
+    0xc6, 0xcb, 0x6e, 0xe7, 0x01, 0xab, 0xd3, 0x39, 0xa9, 0x1e, 
+    0x52, 0x64, 0xd6, 0x30, 0x6c, 0x99, 0x4c, 0x24, 0x5e, 0xed, 
+    0x75, 0xe7, 0xb3, 0x94, 0x60, 0xf4, 0x75, 0xfe, 0x4f, 0xc1, 
+    0x6b, 0xe5, 0x2c, 0x11, 0xd3, 0x08, 0xa2, 0x3b, 0x86, 0x48, 
+    0x7a, 0xb6, 0xbf, 0x91, 0x08, 0x12, 0x71, 0x65, 0xa9, 0xbf, 
+    0x3e, 0xb8, 0xba, 0x9e, 0x07, 0x70, 0x66, 0xdc, 0xab, 0x92, 
+    0x88, 0x25, 0x3e, 0x71, 0xdb, 0x90, 0x29, 0x5f, 0xad, 0xa3, 
+    0x51, 0xc2, 0x82, 0x4c, 0xdb, 0x8f, 0xdf, 0x72, 0x1f, 0xd0, 
+    0x37, 0x75, 0x01, 0xa3, 0x38, 0xed, 0x36, 0x0a, 0x1b, 0xdf, 
+    0xa0, 0x1a, 0x60, 0x65, 0x54, 0xc7, 0x65, 0xd2, 0xc5, 0xe2, 
+    0x78, 0xce, 0xb6, 0x51, 0xe4, 0x84, 0xb0, 0x67, 0x90, 0x85, 
+    0x80, 0x30, 0xdb, 0x88, 0x17, 0x29, 0x0b, 0x36, 0x5a, 0x40, 
+    0x85, 0x80, 0xaa, 0x9c, 0xb8, 0x6d, 0x06, 0xc5, 0xd6, 0xcc, 
+    0xd7, 0x25, 0x83, 0xd1, 0xb2, 0x4d, 0x1e, 0x0d, 0x89, 0x68, 
+    0x26, 0x07, 0x94, 0x92, 0x97, 0x21, 0xda, 0xa4, 0x7e, 0xc7, 
+    0x05, 0x10, 0x1a, 0x98, 0x08, 0xae, 0xb4, 0xf3, 0x33, 0x20, 
+    0xe4, 0xd0, 0x65, 0x94, 0x4c, 0x05, 0xa9, 0xec, 0xec, 0x7f, 
+    0x76, 0x31, 0xe0, 0x33, 0x33, 0x17, 0x0b, 0x6f, 0xb8, 0x8c, 
+    0x8e, 0xb3, 0x54, 0x54, 0x39, 0x91, 0xfe, 0x6e, 0x90, 0xf5, 
+    0x5c, 0x73, 0xbc, 0x2b, 0xb5, 0x1e, 0x7b, 0x2a, 0x7d, 0x83, 
+    0x80, 0x26, 0x8e, 0x9e, 0x41, 0x5c, 0x30, 0x56, 0x12, 0xf8, 
+    0xdd, 0xb9, 0xbe, 0xd7, 0x73, 0xa9, 0xba, 0x7f, 0x24, 0xf5, 
+    0x11, 0xe4, 0xf1, 0x2e, 0xd1, 0x23, 0xc9, 0x4b, 0xa0, 0xfa, 
+    0x50, 0x49, 0x9c, 0x0a, 0xf5, 0xc0, 0x4d, 0xcf, 0x0a, 0xfd, 
+    0xa5, 0xcc, 0x2b, 0x4c, 0x62, 0x95, 0x8c, 0xd0, 0x6b, 0x28, 
+    0x20, 0xf3, 0xdc, 0xec, 0xa4, 0x7b, 0xd6, 0xea, 0xb2, 0x03, 
+    0xaf, 0x31, 0x68, 0x0f, 0xed, 0xe6, 0x05, 0x67, 0xf2, 0x35, 
+    0x87, 0xd6, 0xb2, 0xb0, 0x9b, 0x20, 0x1e, 0xf3, 0x1e, 0x0f, 
+    0x6d, 0x18, 0x91, 0xdd, 0x1f, 0x49, 0xb4, 0x63, 0x8d, 0x92, 
+    0x01, 0x2d, 0x14, 0x70, 0x81, 0x60, 0x60, 0x85, 0xe5, 0xdf, 
+    0x84, 0x10, 0x27, 0xfc, 0x83, 0x9e, 0x89, 0x06, 0x79, 0x0c, 
+    0x38, 0xbe, 0x4d, 0xca, 0xa5, 0xc7, 0x55, 0x82, 0x6c, 0x22, 
+    0x90, 0x11, 0x6c, 0xc6, 0x74, 0xe7, 0x19, 0x26, 0xb5, 0xc0, 
+    0x3b, 0x4e, 0x7c, 0x58, 0xf5, 0xd4, 0xce, 0x5e, 0xcd, 0x1b, 
+    0x8e, 0xb8, 0xe2, 0x17, 0xec, 0xf5, 0xaa, 0x00, 0x58, 0x1d, 
+    0x6f, 0xa0, 0x2a, 0xf4, 0x59, 0x0a, 0xa2, 0x48, 0xa0, 0x28, 
+    0xbb, 0x9a, 0x60, 0x84, 0x58, 0x88, 0x0f, 0xdd, 0x2e, 0x59, 
+    0x6b, 0x0d, 0x10, 0x04, 0x2b, 0x9a, 0x16, 0x5e, 0x2d, 0xed, 
+    0x87, 0x55, 0x4b, 0xd2, 0x41, 0x4d, 0xfe, 0x1e, 0x7e, 0x44, 
+    0xa9, 0xbf, 0xdf, 0x61, 0x9f, 0x8c, 0xad, 0x21, 0xfe, 0x03, 
+    0x84, 0xb5, 0xdc, 0x47, 0x02, 0x33, 0xc5, 0x54, 0xa5, 0xdc, 
+    0x99, 0x45, 0x3e, 0x20, 0x4d, 0xb3, 0x31, 0x37, 0x64, 0xd7, 
+    0xab, 0x7b, 0x3e, 0x1b, 0x76, 0xdc, 0x66, 0xa0, 0xa2, 0x9c, 
+    0x59, 0xbd, 0xe9, 0xcc, 0xe0, 0xae, 0x2a, 0x78, 0xc0, 0x10, 
+    0x85, 0x5a, 0x3c, 0x35, 0x75, 0x1e, 0xe6, 0x97, 0x68, 0x62, 
+    0x10, 0xb4, 0x77, 0xdf, 0xc2, 0x81, 0xe7, 0x53, 0x79, 0xdb, 
+    0x8e, 0xc3, 0x01, 0x19, 0xc3, 0xbc, 0xd9, 0x2c, 0x17, 0xb2, 
+    0xa9, 0x65, 0x5c, 0x75, 0xee, 0x1a, 0x9a, 0x65, 0x5b, 0x44, 
+    0xd5, 0xe2, 0xd2, 0xd5, 0x79, 0x27, 0x53, 0x2b, 0x5a, 0xe5, 
+    0x61, 0x8f, 0x21, 0x52, 0xb7, 0x7a, 0x27, 0x33, 0x94, 0x74, 
+    0x61, 0x0b, 0x82, 0x72, 0xc6, 0x76, 0xd7, 0x67, 0x1b, 0x18, 
+    0x88, 0x2a, 0x67, 0xf3, 0xea, 0x45, 0x8a, 0x38, 0x82, 0x61, 
+    0xbb, 0x02, 0x5b, 0x1f, 0x93, 0xa4, 0xa4, 0x55, 0x3c, 0x83, 
+    0x67, 0x43, 0xb2, 0x0f, 0xb6, 0x9c, 0x88, 0x9b, 0x4b, 0xa8, 
+    0x3c, 0xdc, 0xa1, 0x88, 0x08, 0xf7, 0x16, 0x29, 0xef, 0x36, 
+    0xab, 0x69, 0x09, 0x08, 0xa9, 0x4c, 0xb0, 0x0b, 0x21, 0xc9, 
+    0xc6, 0x12, 0xf3, 0xfa, 0x2a, 0x92, 0x17, 0xc3, 0xf4, 0xe7, 
+    0x71, 0x1e, 0xab, 0x4b, 0x78, 0xdd, 0x65, 0xa8, 0xb7, 0x63, 
+    0xf8, 0xe4, 0x82, 0x5e, 0x21, 0xa9, 0xc8, 0x87, 0x0f, 0x15, 
+    0xf3, 0xf2, 0x8e, 0x44, 0xf3, 0x73, 0x21, 0xfe, 0x52, 0x43, 
+    0x2b, 0xe7, 0xae, 0x14, 0x50, 0xaa, 0x3d, 0x07, 0x91, 0x1d, 
+    0xc1, 0x37, 0xcd, 0xd1, 0x45, 0x36, 0xa1, 0x68, 0xe9, 0x67, 
+    0x5c, 0x38, 0x9e, 0xcb, 0xf2, 0x09, 0x5c, 0xf8, 0x86, 0xfe, 
+    0x46, 0x53, 0xcc, 0x96, 0xbb, 0xd6, 0xca, 0xc1, 0x8d, 0x55, 
+    0xe2, 0xc6, 0x26, 0xba, 0x4f, 0xab, 0x65, 0x0a, 0x65, 0x55, 
+    0xf1, 0x0d, 0x6c, 0xbe, 0x31, 0x05, 0xab, 0xa1, 0xec, 0x1d, 
+    0xc9, 0x76, 0xf6, 0x26, 0xc7, 0x15, 0x8a, 0xf9, 0x8c, 0x23, 
+    0x7a, 0x0b, 0x66, 0xc5, 0xb5, 0x2a, 0xf2, 0x16, 0x84, 0xef, 
+    0x44, 0x94, 0x4b, 0xcb, 0x18, 0x96, 0x05, 0x44, 0xb8, 0xe3, 
+    0xa3, 0xe5, 0x5a, 0x98, 0x3b, 0xd3, 0x3f, 0x4a, 0x80, 0xbc, 
+    0x12, 0x51, 0x92, 0xc6, 0x1d, 0xeb, 0x44, 0x1d, 0x6e, 0x80, 
+    0xfe, 0x3c, 0x7f, 0x89, 0x6d, 0xaf, 0xc3, 0x71, 0xdb, 0xcf, 
+    0xc5, 0xa9, 0x81, 0x87, 0xec, 0xc4, 0x61, 0xf4, 0x3e, 0x16, 
+    0x25, 0x34, 0x0e, 0x15, 0x07, 0xc8, 0xe1, 0x74, 0xc4, 0x7c, 
+    0x8c, 0xfe, 0xa8, 0x1b, 0xd5, 0xad, 0xbb, 0x46, 0x72, 0x6b, 
+    0x46, 0x04, 0x90, 0xed, 0x5f, 0x64, 0xfd, 0x23, 0x35, 0xe0, 
+    0x23, 0xda, 0xa1, 0x76, 0xb0, 0x73, 0xb3, 0x7b, 0xf7, 0xa3, 
+    0x6a, 0xd2, 0xb8, 0x2f, 0xc2, 0x6b, 0xe3, 0x01, 0xd2, 0x28, 
+    0xcd, 0xf4, 0xd2, 0xc2, 0x6c, 0x75, 0xec, 0xc1, 0x76, 0x8f, 
+    0xd4, 0x4c, 0xc0, 0xe0, 0xd6, 0x33, 0x45, 0xbb, 0x21, 0xe7, 
+    0x1f, 0xf9, 0x9e, 0x66, 0x18, 0x52, 0xfa, 0x35, 0x7b, 0x33, 
+    0x75, 0x6d, 0x3b, 0x39, 0xf8, 0xce, 0x75, 0x22, 0x57, 0x65, 
+    0x6a, 0x34, 0x1e, 0xe9, 0x99, 0x66, 0x78, 0xa1, 0xbc, 0xfd, 
+    0xb7, 0xab, 0x98, 0xa7, 0x4b, 0x62, 0x45, 0x10, 0xc4, 0xf4, 
+    0xb0, 0xc2, 0xf1, 0x6d, 0xd3, 0x31, 0x19, 0x8b, 0xc4, 0xa3, 
+    0x36, 0x7e, 0xf4, 0x7f, 0xf4, 0x72, 0xb3, 0xae, 0x75, 0xdb, 
+    0x5b, 0xf6, 0x5d, 0xe7, 0xf3, 0x23, 0x90, 0xf3, 0x0d, 0x5c, 
+    0x0d, 0x54, 0x62, 0x51, 0xbc, 0x20, 0xf2, 0x45, 0x10, 0xb6, 
+    0x3c, 0x93, 0xb6, 0x01, 0x6c, 0x7e, 0x45, 0xc1, 0xad, 0xe2, 
+    0xe1, 0x1c, 0x15, 0x37, 0x30, 0x1f, 0x90, 0x62, 0x6e, 0x6d, 
+    0x6a, 0x83, 0xfd, 0xca, 0xd7, 0x76, 0x2b, 0xb6, 0xc5, 0xf9, 
+    0x1d, 0x54, 0x28, 0xd8, 0xcf, 0xa8, 0xac, 0xf8, 0xc1, 0x54, 
+    0xf8, 0x5c, 0x23, 0x48, 0x6e, 0xb8, 0x7a, 0x02, 0x72, 0x30, 
+    0x0e, 0x11, 0x83, 0xb5, 0x2d, 0x83, 0xad, 0xc3, 0x3e, 0xf6, 
+    0x8f, 0xf3, 0xb6, 0xa0, 0xd7, 0xbb, 0x43, 0xc1, 0x75, 0xfe, 
+    0x9e, 0x1d, 0x69, 0x2e, 0x83, 0x03, 0x43, 0x43, 0xf6, 0xfa, 
+    0x43, 0x54, 0xd3, 0xab, 0xfc, 0x53, 0xde, 0x87, 0x92, 0x7b, 
+    0xc4, 0x3a, 0x5a, 0xe4, 0xee, 0xef, 0x18, 0x96, 0x9c, 0xe0, 
+    0xea, 0xd1, 0xce, 0x3a, 0x67, 0x56, 0x0c, 0xe8, 0x23, 0x0d, 
+    0x34, 0x1e, 0x0d, 0x80, 0x03, 0x8f, 0x90, 0xc5, 0xba, 0x01, 
+    0x43, 0xfb, 0x94, 0x99, 0x3b, 0x19, 0x13, 0x30, 0x5f, 0xd8, 
+    0x0b, 0xb1, 0x10, 0x8d, 0x61, 0x20, 0x99, 0x3b, 0x2d, 0xbe, 
+    0x64, 0xf7, 0x7a, 0x97, 0x84, 0x12, 0xc0, 0x44, 0x85, 0x0b, 
+    0x4b, 0x2c, 0x1c, 0x83, 0x7f, 0x6a, 0xd7, 0x71, 0x6b, 0x5b, 
+    0x71, 0xf0, 0xc2, 0xbb, 0x9a, 0xfb, 0xad, 0x5c, 0x27, 0x07, 
+    0x02, 0x5f, 0x54, 0xd7, 0xb5, 0x5f, 0x95, 0xd1, 0x15, 0xb0, 
+    0x27, 0x1d, 0xda, 0x3f, 0x87, 0x5d, 0x8b, 0x20, 0x2c, 0x12, 
+    0x6b, 0x11, 0x67, 0xcb, 0xd7, 0xa7, 0xc5, 0xa1, 0x55, 0x91, 
+    0xed, 0x2a, 0x11, 0xc4, 0x84, 0x1c, 0xab, 0xeb, 0x68, 0x84, 
+    0x18, 0x94, 0x5b, 0xb1, 0x39, 0x65, 0xca, 0x6c, 0x14, 0xe1, 
+    0x65, 0xec, 0x18, 0xea, 0x7c, 0x5c, 0x7f, 0x4f, 0xda, 0xe9, 
+    0x4d, 0x30, 0xaf, 0x96, 0x47, 0x01, 0x7a, 0xfa, 0xd2, 0x97, 
+    0x37, 0x73, 0xa6, 0xd5, 0x4b, 0x36, 0xb4, 0xda, 0xa7, 0x69, 
+    0x43, 0x38, 0xa6, 0x59, 0x20, 0x4b, 0xd4, 0x1c, 0x2e, 0xbb, 
+    0x18, 0xed, 0xd3, 0x3e, 0xef, 0xd2, 0xac, 0xb1, 0xac, 0x51, 
+    0x23, 0x8b, 0xbc, 0xc3, 0x01, 0x99, 0xd4, 0x2c, 0x6c, 0x33, 
+    0xf5, 0xe4, 0x3d, 0x3e, 0x6d, 0xae, 0x05, 0x62, 0xeb, 0xb3, 
+    0xb6, 0x66, 0x64, 0x42, 0x32, 0xba, 0x81, 0xd0, 0xa1, 0x03, 
+    0xef, 0xc9, 0x4b, 0xee, 0x22, 0x0a, 0x64, 0x08, 0x8a, 0x99, 
+    0xf1, 0xc0, 0x4e, 0x9e, 0x79, 0x4a, 0xa0, 0x9b, 0xb1, 0xba, 
+    0x84, 0xbd, 0xe9, 0x5b, 0xec, 0x24, 0x4b, 0x34, 0x5d, 0x2a, 
+    0xcc, 0xd2, 0x5f, 0xc7, 0x58, 0xe6, 0x47, 0xbb, 0xb9, 0xf7, 
+    0xb4, 0xc0, 0x1d, 0x2b, 0x21, 0xc0, 0x7d, 0x76, 0xd5, 0xc2, 
+    0x40, 0xa5, 0xe4, 0x9b, 0x13, 0x0d, 0xb9, 0xb7, 0xb3, 0x2f, 
+    0xcd, 0xc4, 0xef, 0xef, 0x48, 0x28, 0xe7, 0x87, 0x11, 0x46, 
+    0xed, 0xac, 0xbf, 0xec, 0x69, 0x30, 0xa7, 0x61, 0x92, 0xcf, 
+    0x5b, 0xa2, 0x69, 0xe0, 0x14, 0xa2, 0xf9, 0x29, 0xf8, 0xc5, 
+    0x7e, 0xe2, 0xf0, 0xfc, 0xf0, 0x32, 0xf3, 0x0c, 0xab, 0x7d, 
+    0xbb, 0xbd, 0x89, 0x54, 0x98, 0x47, 0x6a, 0xb2, 0x87, 0x45, 
+    0xd6, 0x06, 0xd4, 0xbf, 0x77, 0x95, 0xf4, 0x61, 0x38, 0x96, 
+    0xad, 0x56, 0xfa, 0x71, 0x9b, 0xde, 0x15, 0x0f, 0xc3, 0x3d, 
+    0xf5, 0x21, 0x63, 0x9b, 0xfe, 0x70, 0x2d, 0xfe, 0xc2, 0xcf, 
+    0x26, 0xfd, 0xd6, 0xb6, 0x3e, 0xcc, 0x05, 0x36, 0x11, 0xe2, 
+    0xb7, 0xdd, 0xc4, 0x6f, 0xc7, 0xac, 0xfe, 0x6f, 0x5b, 0xb4, 
+    0xe6, 0x47, 0x53, 0x30, 0x69, 0x56, 0x5b, 0x5c, 0xee, 0xc4, 
+    0xd0, 0x5b, 0x1e, 0x1b, 0xb8, 0xbe, 0xcd, 0xde, 0x28, 0xe6, 
+    0x90, 0x2d, 0xcf, 0xa0, 0xfb, 0x11, 0x3b, 0xda, 0xba, 0x2e, 
+    0x05, 0xa8, 0x9e, 0x3b, 0x07, 0xce, 0x87, 0xf0, 0xc4, 0xb4, 
+    0x2c, 0x33, 0x32, 0xda, 0x5a, 0x86, 0xba, 0x97, 0x1a, 0xf4, 
+    0xbf, 0xbf, 0x7c, 0x0c, 0xa8, 0xf1, 0x17, 0x99, 0x59, 0xa5, 
+    0x32, 0xc2, 0x48, 0x2d, 0xa6, 0x02, 0x18, 0xc1, 0x97, 0x95, 
+    0x7b, 0x6f, 0x62, 0xf8, 0xbb, 0x80, 0x83, 0xb9, 0xe3, 0x27, 
+    0xdd, 0x38, 0xb0, 0x26, 0xea, 0xde, 0xa5, 0x45, 0xf3, 0x05, 
+    0xde, 0x51, 0x36, 0x97, 0xc2, 0x80, 0x13, 0x9d, 0xd8, 0x58, 
+    0xd8, 0xc3, 0xac, 0xaf, 0xfd, 0xf5, 0x19, 0x21, 0x26, 0x9d, 
+    0xe8, 0xeb, 0xc5, 0x94, 0xd2, 0xb2, 0x56, 0x7e, 0xe6, 0xab, 
+    0x3c, 0x9d, 0x17, 0xaa, 0x8a, 0x15, 0x6d, 0xd0, 0x01, 0x20, 
+    0x01, 0xe0, 0x76, 0x44, 0x9c, 0xd0, 0x36, 0x56, 0xbf, 0xc2, 
+    0x3d, 0xee, 0x5a, 0x11, 0x22, 0xbf, 0x05, 0xf5, 0x98, 0xfe, 
+    0x88, 0x86, 0x9f, 0x5b, 0x55, 0xce, 0x8b, 0x65, 0x49, 0x24, 
+    0x13, 0x30, 0x44, 0xb4, 0x32, 0x69, 0x23, 0x13, 0xab, 0xb9, 
+    0x76, 0x7c, 0x7d, 0xb1, 0x14, 0xdf, 0x35, 0x72, 0x82, 0xbd, 
+    0xda, 0x6a, 0xa2, 0x91, 0x5c, 0x9d, 0x2a, 0xe0, 0x89, 0x4d, 
+    0x89, 0xa7, 0xf4, 0x16, 0x7e, 0xc0, 0x74, 0x7e, 0x7d, 0xcc, 
+    0x2f, 0xf5, 0x63, 0x1a, 0xcb, 0x03, 0x18, 0x17, 0x2a, 0x21, 
+    0x10, 0x47, 0x83, 0x77, 0x71, 0x2a, 0xdb, 0x82, 0x09, 0xb8, 
+    0x0e, 0x5c, 0x2d, 0x9d, 0x56, 0x12, 0x20, 0x3d, 0x7c, 0xea, 
+    0xb1, 0x2f, 0x2a, 0x65, 0xc3, 0xb3, 0xc1, 0xc8, 0x03, 0x74, 
+    0xe4, 0xe9, 0x24, 0x3a, 0x31, 0xcf, 0x5b, 0x66, 0x3d, 0x23, 
+    0x1c, 0x39, 0x3b, 0x8b, 0x66, 0x46, 0x0b, 0xcc, 0x57, 0xe7, 
+    0xb5, 0xe7, 0x35, 0x7c, 0xd0, 0xe2, 0x94, 0x22, 0x33, 0x2a, 
+    0xb4, 0x01, 0x70, 0x1f, 0x03, 0x8c, 0xde, 0x3b, 0x70, 0x13, 
+    0x11, 0x1b, 0xe4, 0x22, 0xf3, 0x2b, 0x1f, 0x9d, 0x76, 0x08, 
+    0xec, 0x7c, 0x88, 0x19, 0xc0, 0x73, 0xef, 0x23, 0x94, 0xda, 
+    0xa5, 0xa2, 0x2d, 0xf4, 0xd9, 0xc5, 0x60, 0x87, 0x23, 0xb7, 
+    0x35, 0x29, 0x37, 0x04, 0xae, 0x7f, 0x7c, 0x81, 0x78, 0x49, 
+    0x30, 0xfc, 0x27, 0x98, 0x83, 0xa9, 0xf2, 0xcc, 0xe3, 0x45, 
+    0xa1, 0xe2, 0xe9, 0xf2, 0x36, 0x3b, 0x43, 0x8d, 0xe9, 0x39, 
+    0xa6, 0x2f, 0x00, 0xed, 0xf4, 0x88, 0x61, 0xd7, 0xfc, 0xad, 
+    0x5a, 0x94, 0x8f, 0x84, 0xc3, 0x38, 0xa0, 0x4e, 0xc3, 0x0d, 
+    0x30, 0xe1, 0x39, 0x36, 0xc8, 0x86, 0x41, 0xb8, 0x72, 0xd8, 
+    0x48, 0x4a, 0x08, 0x56, 0xa1, 0xbc, 0x52, 0x94, 0xee, 0x36, 
+    0x7f, 0xc4, 0xf3, 0x73, 0x59, 0x12, 0x0a, 0x03, 0x4c, 0xbe, 
+    0x10, 0x96, 0xdd, 0x37, 0xdd, 0x48, 0x41, 0xab, 0xcd, 0xf7, 
+    0x99, 0xc8, 0xac, 0x23, 0x23, 0x56, 0xba, 0x8b, 0x24, 0x0d, 
+    0x18, 0x15, 0x6d, 0x5b, 0x88, 0x01, 0x24, 0xe9, 0x99, 0x27, 
+    0x56, 0x3d, 0xdc, 0xe4, 0xf2, 0x41, 0x2b, 0x65, 0x1f, 0xe0, 
+    0x84, 0x2f, 0x31, 0xbd, 0x27, 0xd8, 0x0e, 0x99, 0xea, 0xc1, 
+    0xa9, 0x0b, 0x21, 0x2a, 0xe7, 0x57, 0xc0, 0xa3, 0x79, 0x1f, 
+    0xf7, 0xbb, 0x4e, 0x50, 0xf2, 0x37, 0x53, 0xed, 0x73, 0xb2, 
+    0xde, 0x08, 0xef, 0xa5, 0x9e, 0xcd, 0x38, 0x51, 0x55, 0xa1, 
+    0xf6, 0xac, 0x78, 0xf5, 0xd6, 0xd9, 0x7d, 0x92, 0x34, 0x57, 
+    0x78, 0xd1, 0x47, 0x17, 0xde, 0x10, 0x7e, 0xce, 0xf8, 0x1c, 
+    0x09, 0xcc, 0xe0, 0xea, 0x6d, 0x55, 0xec, 0xcb, 0x49, 0xaa, 
+    0x22, 0x5a, 0x18, 0xe7, 0x10, 0xb2, 0xb5, 0xf0, 0x9d, 0xdf, 
+    0xc5, 0x49, 0xe8, 0x4d, 0x9d, 0xcd, 0x51, 0x53, 0xf5, 0x95, 
+    0x37, 0x4b, 0xe4, 0x57, 0xca, 0x3f, 0x55, 0x68, 0x75, 0xc1, 
+    0xcb, 0xd6, 0x5b, 0xb7, 0x4b, 0x62, 0xb5, 0xce, 0x74, 0x98, 
+    0x84, 0x01, 0x28, 0xc8, 0x1c, 0x21, 0xc2, 0xc5, 0xa8, 0xdf, 
+    0x69, 0xd8, 0xd5, 0xe2, 0x4d, 0x72, 0xc1, 0xd5, 0x2f, 0x13, 
+    0x94, 0x6d, 0x1c, 0xc6, 0x46, 0xdf, 0x70, 0x57, 0xdd, 0x9b, 
+    0xb8, 0x90, 0x37, 0xad, 0xe9, 0xf0, 0x09, 0x1f, 0xfb, 0xb1, 
+    0x99, 0x6e, 0xa5, 0xe7, 0x9f, 0x94, 0xf1, 0xd4, 0xbd, 0x85, 
+    0x13, 0xfd, 0xbf, 0x18, 0x86, 0xc2, 0xb7, 0x95, 0xa9, 0x4c, 
+    0x26, 0x2f, 0x54, 0xcf, 0xc1, 0x64, 0xf6, 0x85, 0x13, 0xc9, 
+    0xd7, 0xd2, 0xef, 0xd7, 0xf0, 0x38, 0xf9, 0x19, 0x10, 0x88, 
+    0xd6, 0x13, 0x93, 0xc5, 0xb4, 0xc3, 0x58, 0x2e, 0x38, 0xbf, 
+    0xa9, 0xf7, 0x7d, 0xbc, 0x6d, 0xb6, 0x08, 0x3e, 0xd4, 0xab, 
+    0x7b, 0xd1, 0xe0, 0x68, 0x5c, 0xc7, 0xa8, 0x6a, 0xab, 0x9c, 
+    0x86, 0xe9, 0x0a, 0x7d, 0x71, 0x92, 0xdb, 0xd3, 0xa3, 0x8f, 
+    0x0e, 0x5c, 0xad, 0x86, 0xac, 0x5f, 0x9d, 0xd6, 0x75, 0x3b, 
+    0xf3, 0x64, 0x61, 0x27, 0xa5, 0x10, 0x74, 0x3f, 0x30, 0xce, 
+    0xa3, 0x40, 0x22, 0xb4, 0x29, 0xa3, 0x9b, 0x94, 0x10, 0xe5, 
+    0xc0, 0x8f, 0x7f, 0xe3, 0x51, 0x13, 0x93, 0x2f, 0xb2, 0xba, 
+    0x4e, 0xb8, 0x6e, 0xc2, 0x43, 0x25, 0xb6, 0xb1, 0x50, 0xe2, 
+    0xe6, 0x8a, 0xbe, 0x1c, 0xc6, 0x0a, 0x49, 0xcd, 0x87, 0xa4, 
+    0x36, 0x8b, 0xcf, 0x8c, 0xa1, 0xf3, 0x69, 0x3a, 0x73, 0x04, 
+    0x9d, 0xde, 0xf1, 0x8e, 0x7d, 0x62, 0x43, 0xf5, 0x38, 0x4b, 
+    0x81, 0x1f, 0xb7, 0xac, 0x06, 0x29, 0x08, 0xd3, 0x06, 0xd0, 
+    0x67, 0x7f, 0xf4, 0x25, 0x53, 0x54, 0x20, 0x2e, 0xdf, 0x6b, 
+    0x33, 0x03, 0x64, 0x04, 0x36, 0xd0, 0x25, 0x02, 0x2a, 0x6e, 
+    0xa6, 0xfc, 0xd5, 0x36, 0x6b, 0x55, 0x6c, 0x9f, 0x43, 0x34, 
+    0x29, 0xc7, 0xe9, 0xe2, 0xe4, 0x9a, 0x28, 0xbb, 0x40, 0x8d, 
+    0x4d, 0x99, 0x13, 0x84, 0x71, 0x31, 0x6d, 0xaa, 0x99, 0x34, 
+    0xb0, 0x9a, 0x7d, 0x16, 0xe6, 0xdf, 0x83, 0xaf, 0xbd, 0xf8, 
+    0x77, 0x88, 0xb1, 0x4b, 0x18, 0xa8, 0x62, 0xe2, 0xb5, 0xdb, 
+    0x17, 0x31, 0x47, 0xeb, 0xca, 0x9e, 0x9e, 0xe7, 0x19, 0x99, 
+    0x09, 0x00, 0x9a, 0x40, 0x55, 0x38, 0xc4, 0xea, 0xd7, 0xbf, 
+    0xae, 0x00, 0x95, 0x8b, 0x78, 0x01, 0x60, 0x01, 0xf9, 0xa3, 
+    0xac, 0x49, 0xc8, 0xdd, 0xc5, 0x78, 0xc2, 0x69, 0xf2, 0x2c, 
+    0x2b, 0x7a, 0x2f, 0x01, 0x09, 0x13, 0xe9, 0x85, 0x3e, 0xb9, 
+    0x63, 0xd9, 0x48, 0xea, 0xd1, 0x9e, 0x21, 0xda, 0x4b, 0x26, 
+    0xc7, 0xc9, 0xfd, 0xb5, 0x42, 0x37, 0x24, 0x1a, 0x97, 0x51, 
+    0x8c, 0x20, 0x51, 0xfe, 0x0d, 0x14, 0x0c, 0xc8, 0x50, 0x74, 
+    0xb5, 0x1d, 0xba, 0x8c, 0xd5, 0x8a, 0x80, 0xf4, 0x1e, 0x37, 
+    0xf2, 0x6a, 0x3c, 0x31, 0xf3, 0x90, 0xb4, 0x0b, 0xab, 0x7f, 
+    0x44, 0x2c, 0xe1, 0x09, 0x25, 0x42, 0xf9, 0x81, 0xd4, 0x49, 
+    0x4c, 0x5b, 0xd4, 0xc1, 0x14, 0x5e, 0x5e, 0x3c, 0xdc, 0xe8, 
+    0x7d, 0xee, 0x39, 0xe6, 0x10, 0x78, 0x7e, 0xf6, 0xb5, 0xb2, 
+    0xf1, 0x32, 0xac, 0xa2, 0x48, 0xc4, 0xc6, 0xbe, 0x0d, 0xeb, 
+    0x8a, 0x53, 0xf9, 0x0a, 0x6f, 0x50, 0x71, 0x36, 0x10, 0x88, 
+    0xc5, 0x8c, 0xda, 0xb9, 0x99, 0xe3, 0x56, 0xb6, 0xad, 0x59, 
+    0x1e, 0x42, 0x72, 0x97, 0x53, 0x75, 0x2f, 0x4e, 0x08, 0x43, 
+    0x53, 0xb1, 0x02, 0xbc, 0xd7, 0xc9, 0x13, 0x7a, 0x05, 0x84, 
+    0x8c, 0x0a, 0xf3, 0x62, 0x11, 0x5d, 0x0d, 0x56, 0x2a, 0x5a, 
+    0x7e, 0x05, 0x62, 0x57, 0xe5, 0x05, 0xb7, 0x5a, 0xf7, 0x69, 
+    0x39, 0xd7, 0xf2, 0x0d, 0x93, 0x82, 0x0c, 0x9a, 0xa4, 0xc8, 
+    0x77, 0x6c, 0xe3, 0x60, 0x21, 0x6c, 0x1b, 0x5e, 0xe2, 0xf1, 
+    0xb6, 0xbb, 0x69, 0x21, 0xca, 0xf6, 0xf0, 0x5f, 0xb4, 0xdb, 
+    0xeb, 0x20, 0x47, 0xa9, 0x82, 0x11, 0x0d, 0x72, 0xde, 0xd1, 
+    0xa4, 0xda, 0x6c, 0xfa, 0xaa, 0xe9, 0x2b, 0x30, 0xab, 0x6a, 
+    0x83, 0x48, 0xea, 0xfd, 0xf9, 0xaa, 0x76, 0x93, 0xe6, 0xe2, 
+    0x25, 0xb0, 0x6d, 0x4f, 0x11, 0xd4, 0xee, 0x9e, 0x6d, 0x72, 
+    0x20, 0x00, 0x8e, 0x90, 0x62, 0x0a, 0x57, 0x57, 0x08, 0x1f, 
+    0xfd, 0x24, 0x64, 0x69, 0x6e, 0x8e, 0x76, 0x8c, 0xbb, 0x8d, 
+    0x42, 0x8d, 0xd9, 0x94, 0xbf, 0xf2, 0x20, 0x6e, 0xcd, 0xca, 
+    0xdd, 0x0a, 0x4b, 0x5c, 0xb7, 0xa7, 0x36, 0x8f, 0xea, 0x51, 
+    0x2c, 0x3f, 0x59, 0x66, 0xa1, 0x8a, 0x8b, 0xba, 0x4c, 0x6c, 
+    0xc3, 0x63, 0x5b, 0x6d, 0x29, 0x2f, 0xc2, 0xd1, 0x99, 0x34, 
+    0x23, 0xf0, 0x81, 0x5c, 0xf7, 0x6e, 0xea, 0x84, 0x65, 0x9c, 
+    0x7f, 0x88, 0x1e, 0x8c, 0xf9, 0xaf, 0x57, 0x52, 0x58, 0x4a, 
+    0xd9, 0x3b, 0x6a, 0xf1, 0x02, 0xa6, 0xe0, 0xfe, 0x9b, 0xdf, 
+    0x7e, 0xa1, 0x74, 0x96, 0x37, 0x61, 0x95, 0x46, 0x81, 0xb2, 
+    0x37, 0x50, 0x47, 0x57, 0xf3, 0xc7, 0x07, 0xb1, 0x54, 0xe0, 
+    0x39, 0x0c, 0x48, 0xd7, 0x59, 0x17, 0x37, 0x0c, 0x08, 0x1b, 
+    0x84, 0xc7, 0xc1, 0xad, 0xdc, 0x3a, 0xd8, 0x0f, 0x77, 0x71, 
+    0x25, 0x92, 0x23, 0x1f, 0x51, 0xe6, 0x8d, 0x94, 0x8e, 0xc8, 
+    0xbe, 0x31, 0xbf, 0xce, 0x30, 0xc9, 0x3e, 0xe0, 0x65, 0x71, 
+    0xa2, 0x50, 0xb7, 0xab, 0x1a, 0x81, 0x3c, 0x8e, 0x3c, 0xe4, 
+    0xc3, 0xaa, 0x16, 0xe4, 0x0b, 0x41, 0xdd, 0x7f, 0x76, 0x06, 
+    0xad, 0x82, 0x19, 0x6e, 0x37, 0x9d, 0xc4, 0x3c, 0x26, 0x3a, 
+    0xa1, 0x13, 0x93, 0xbc, 0x9b, 0x6a, 0xd1, 0xb1, 0x03, 0x75, 
+    0xbd, 0x3a, 0x9d, 0x6c, 0x0f, 0xe3, 0x78, 0xf6, 0x12, 0x73, 
+    0x29, 0xef, 0x10, 0xdf, 0xfb, 0xfb, 0x58, 0xd6, 0xd8, 0x14, 
+    0xd7, 0xa2, 0xdb, 0x5f, 0x71, 0x49, 0x98, 0x5d, 0x3f, 0xb5, 
+    0xed, 0xc3, 0x1e, 0x77, 0xec, 0x3f, 0x5e, 0x99, 0x15, 0x41, 
+    0xc5, 0xb0, 0x82, 0xe5, 0x1c, 0x81, 0x85, 0x9e, 0xf1, 0x5a, 
+    0x58, 0x80, 0xb2, 0x27, 0x2b, 0x19, 0xb7, 0xf4, 0x29, 0x79, 
+    0x01, 0x41, 0x3e, 0x9c, 0x7a, 0x76, 0x49, 0x53, 0xaa, 0x8c, 
+    0xd1, 0x2f, 0xc5, 0x72, 0xdb, 0x2a, 0x4b, 0x2b, 0x9e, 0x76, 
+    0x20, 0x08, 0xfd, 0x3e, 0xb9, 0x97, 0xda, 0x6c, 0xdb, 0xf1, 
+    0x48, 0x6a, 0x3e, 0xbd, 0x8a, 0x15, 0x22, 0xbc, 0x13, 0x67, 
+    0xe1, 0xd4, 0x1a, 0xa3, 0xc1, 0xdb, 0xa4, 0x54, 0x16, 0x05, 
+    0x79, 0x13, 0x2d, 0x8e, 0x7c, 0x85, 0xc9, 0x4e, 0x4b, 0x00, 
+    0xc7, 0xc5, 0x7f, 0x47, 0xf0, 0xee, 0x3a, 0x36, 0x8f, 0x96, 
+    0xfe, 0x2a, 0x77, 0x48, 0x45, 0x3b, 0x52, 0x8d, 0x6d, 0x21, 
+    0x3f, 0x40, 0x52, 0x22, 0x83, 0x26, 0x06, 0xd3, 0x81, 0x53, 
+    0x18, 0x4f, 0xeb, 0xc8, 0x20, 0x1f, 0xc8, 0xab, 0x7b, 0x40, 
+    0x88, 0x9b, 0x6e, 0x5b, 0x96, 0x26, 0x64, 0xbf, 0x35, 0x33, 
+    0x06, 0x16, 0x7a, 0xf5, 0x43, 0xd1, 0xa7, 0xb9, 0x4c, 0x11, 
+    0x91, 0x6a, 0x52, 0x21, 0xa0, 0x1a, 0x23, 0x08, 0x64, 0x51, 
+    0x43, 0x3a, 0xb9, 0xf2, 0xa6, 0xa2, 0xc8, 0x46, 0xf1, 0xd0, 
+    0x8f, 0x0d, 0xe0, 0xd0, 0x1a, 0xf8, 0xcd, 0xcf, 0xaa, 0xaa, 
+    0x81, 0xa5, 0x69, 0xab, 0x53, 0xd6, 0xfb, 0x63, 0xe0, 0x9e, 
+    0xc6, 0x27, 0xc6, 0xb1, 0xf6, 0x49, 0x96, 0xea, 0x80, 0x19, 
+    0xf9, 0xe4, 0xb6, 0xec, 0x6b, 0x77, 0x14, 0xee, 0x93, 0xa3, 
+    0xa1, 0xd9, 0xac, 0x83, 0x63, 0xa1, 0x31, 0x1d, 0x5e, 0x8b, 
+    0x86, 0x1f, 0xa2, 0xc0, 0x97, 0x3a, 0x4a, 0xeb, 0xc3, 0xbb, 
+    0x1a, 0x30, 0xd3, 0xcd, 0x75, 0x8c, 0x93, 0x66, 0x3b, 0xdb, 
+    0x93, 0xfa, 0xf1, 0x62, 0xeb, 0xd7, 0x81, 0xb5, 0xe4, 0xba, 
+    0xdc, 0x7a, 0x11, 0x4c, 0x79, 0x6c, 0xe5, 0xdd, 0x69, 0x71, 
+    0x77, 0xf0, 0x4f, 0x90, 0x4b, 0x00, 0xa5, 0xd3, 0xe7, 0xba, 
+    0xad, 0x32, 0x5b, 0x34, 0x07, 0x0a, 0x70, 0x1f, 0xe1, 0xd8, 
+    0xb0, 0x9b, 0xc2, 0x35, 0x9c, 0x8c, 0x2b, 0xe2, 0xdd, 0x31, 
+    0x48, 0xcd, 0xf1, 0x8b, 0x18, 0xfa, 0x21, 0x9a, 0x2a, 0x3e, 
+    0xcd, 0xeb, 0x5e, 0x06, 0xfc, 0xf6, 0x9e, 0x41, 0x5f, 0xc8, 
+    0x7b, 0x13, 0xaf, 0x91, 0xac, 0xe6, 0xe9, 0xcd, 0xc1, 0x7c, 
+    0x84, 0x81, 0xaf, 0x75, 0x0a, 0xe9, 0x30, 0x6d, 0xfa, 0x02, 
+    0x2a, 0x1f, 0x13, 0x2e, 0x2d, 0x34, 0x7b, 0x87, 0x6c, 0x5a, 
+    0x43, 0xaa, 0x7d, 0x25, 0xd0, 0xc3, 0xcf, 0x32, 0x8b, 0x3f, 
+    0x28, 0xb5, 0x0b, 0xa8, 0x7d, 0x06, 0x8e, 0x04, 0x39, 0xd0, 
+    0x96, 0x6a, 0xa2, 0x26, 0x64, 0xe1, 0xdc, 0xcb, 0xcc, 0xd5, 
+    0xfe, 0x33, 0xef, 0x9c, 0xba, 0xdb, 0xb8, 0xe1, 0xf2, 0x2a, 
+    0xb8, 0x5c, 0x50, 0xb8, 0x2a, 0xc8, 0x82, 0xc0, 0x1c, 0xc9, 
+    0x9b, 0xca, 0x53, 0x0a, 0x52, 0x3d, 0xbe, 0x23, 0x71, 0xdc, 
+    0x5c, 0xf8, 0x41, 0x6e, 0xb1, 0x00, 0x17, 0x96, 0xec, 0xdc, 
+    0xfb, 0xac, 0x61, 0xb4, 0x62, 0xcd, 0x97, 0x5e, 0xc9, 0x82, 
+    0xf8, 0x00, 0xc7, 0x50, 0x2c, 0xf5, 0xe5, 0x6b, 0xd8, 0xe1, 
+    0x2c, 0x65, 0x50, 0x46, 0xbd, 0x1e, 0x4f, 0xe3, 0x5c, 0x3b, 
+    0x90, 0x22, 0xf4, 0x24, 0x4f, 0x4f, 0x8f, 0x08, 0x3a, 0x8e, 
+    0x02, 0x9c, 0x99, 0xf9, 0xe4, 0x0b, 0x34, 0x37, 0x66, 0x0b, 
+    0x29, 0x36, 0xa6, 0xde, 0x8a, 0x9d, 0x21, 0x2b, 0xab, 0x6a, 
+    0x0f, 0xe3, 0x55, 0x2c, 0xb9, 0xa7, 0x35, 0x68, 0x5e, 0xf4, 
+    0xaf, 0x38, 0x66, 0x1d, 0x2a, 0xcc, 0xe9, 0x26, 0x7a, 0x72, 
+    0xb4, 0x16, 0xde, 0x98, 0x4a, 0xe0, 0x01, 0xa4, 0x36, 0x11, 
+    0x00, 0x26, 0xc6, 0x08, 0xeb, 0x7e, 0xfd, 0x2f, 0x1a, 0xe7, 
+    0x8d, 0x17, 0xe2, 0xb4, 0xf3, 0x41, 0x53, 0xc5, 0x56, 0x64, 
+    0x92, 0x1f, 0xcf, 0x7f, 0xf2, 0xf6, 0x2c, 0x06, 0x10, 0x5e, 
+    0x21, 0xd1, 0xc6, 0x22, 0x66, 0x66, 0x5b, 0x70, 0x97, 0xbc, 
+    0xbf, 0x9f, 0x16, 0x92, 0x95, 0x66, 0xfb, 0xe3, 0x8b, 0xf6, 
+    0x36, 0xb7, 0x53, 0x8a, 0xd3, 0xbe, 0xa3, 0x3b, 0xa4, 0xc3, 
+    0x10, 0xc0, 0x7c, 0x28, 0xea, 0x7c, 0xa7, 0x35, 0x79, 0x0a, 
+    0x67, 0xe4, 0x47, 0xa0, 0x13, 0xea, 0xe8, 0x2e, 0x36, 0xbe, 
+    0xae, 0x83, 0x5f, 0xdc, 0xad, 0x35, 0x7b, 0xcd, 0x6d, 0x44, 
+    0x35, 0x40, 0xd7, 0xbf, 0x3a, 0xd0, 0xb9, 0x3d, 0xdc, 0xed, 
+    0x14, 0x4b, 0xcd, 0x34, 0x2c, 0x06, 0xd9, 0x61, 0x18, 0x80, 
+    0xe1, 0x06, 0x41, 0xb1, 0x9e, 0x46, 0x44, 0xa5, 0xb0, 0xaf, 
+    0xc9, 0x26, 0xa0, 0x7a, 0xca, 0xf5, 0x8a, 0xee, 0x37, 0xa6, 
+    0xc2, 0xd2, 0x04, 0x12, 0xc7, 0x1e, 0x0c, 0x69, 0x35, 0x61, 
+    0xe3, 0x61, 0x73, 0xdf, 0xad, 0xe8, 0x3d, 0x3b, 0xaa, 0xc2, 
+    0x19, 0x77, 0x84, 0xb8, 0x3d, 0x6d, 0x2b, 0xb5, 0xcf, 0xd9, 
+    0x8f, 0x75, 0x33, 0xc3, 0x04, 0xb5, 0xa4, 0xf8, 0xcc, 0xea, 
+    0x1b, 0x41, 0x11, 0x18, 0x09, 0x9d, 0xf4, 0x64, 0x8b, 0x59, 
+    0xee, 0x9e, 0xf8, 0x72, 0xdf, 0x97, 0x1d, 0xca, 0x4e, 0x55, 
+    0xca, 0x2b, 0xe0, 0x95, 0x9c, 0x0e, 0x91, 0xf5, 0xbf, 0x76, 
+    0x58, 0x57, 0xbc, 0xa0, 0xc0, 0xd1, 0x01, 0xef, 0x8f, 0x3b, 
+    0x21, 0x0d, 0x18, 0x4d, 0x1f, 0x98, 0x59, 0x2b, 0x94, 0xc4, 
+    0x0b, 0xb6, 0x26, 0x93, 0xcc, 0xe5, 0x43, 0x5c, 0xf5, 0xb5, 
+    0x48, 0x4d, 0x1e, 0xdd, 0x6c, 0x3f, 0x65, 0x5c, 0xb1, 0x00, 
+    0x62, 0x79, 0xf2, 0xf9, 0xaa, 0xb5, 0xbc, 0x36, 0x17, 0x0d, 
+    0x80, 0xc3, 0xcc, 0xc3, 0x13, 0x17, 0x5b, 0x90, 0xbc, 0xe8, 
+    0x6b, 0xbd, 0x9e, 0x8d, 0x15, 0x23, 0xd8, 0x2a, 0x54, 0x37, 
+    0x2d, 0x04, 0x3c, 0xeb, 0xe4, 0x22, 0x7c, 0x2f, 0x14, 0x8e, 
+    0xca, 0xbc, 0xbd, 0x8a, 0xb2, 0xd3, 0x85, 0xb4, 0xd3, 0x88, 
+    0x96, 0x9e, 0x45, 0xdb, 0xd0, 0xd9, 0xf4, 0x8e, 0x8a, 0x80, 
+    0x7b, 0xb0, 0xfb, 0x2a, 0xed, 0x9d, 0xb5, 0x96, 0xac, 0x96, 
+    0xcf, 0x0d, 0xfb, 0x82, 0x38, 0x41, 0xbb, 0x48, 0x38, 0x59, 
+    0x03, 0xb8, 0x41, 0xac, 0xbb, 0x91, 0x82, 0x9c, 0x2c, 0x17, 
+    0x8b, 0xe2, 0x71, 0x51, 0xdd, 0xe1, 0x14, 0x36, 0x7a, 0xc7, 
+    0xb0, 0x19, 0x5c, 0x75, 0xca, 0x6a, 0xf4, 0x65, 0x9e, 0xab, 
+    0xfb, 0x9e, 0xaf, 0x7b, 0x87, 0x96, 0xf5, 0xbf, 0x42, 0xf9, 
+    0x26, 0x69, 0xfd, 0xfb, 0x52, 0x11, 0x81, 0xb0, 0x52, 0x4a, 
+    0x24, 0xa8, 0x2d, 0xa0, 0x31, 0x2b, 0xcf, 0x7a, 0x0b, 0xd2, 
+    0x5b, 0x21, 0xfb, 0x20, 0x28, 0xac, 0x68, 0xd1, 0x11, 0x4e, 
+    0xe0, 0x35, 0x3c, 0xad, 0x51, 0x56, 0xdc, 0x5c, 0x8d, 0x8e, 
+    0x13, 0x87, 0xbc, 0x96, 0x3c, 0x03, 0x5f, 0xb2, 0x93, 0xbb, 
+    0x8c, 0x37, 0x9d, 0xb2, 0x0b, 0x87, 0x26, 0x07, 0x32, 0x43, 
+    0x73, 0xcb, 0xfc, 0x5c, 0xa6, 0x5d, 0x58, 0xa5, 0x3d, 0xe8, 
+    0x03, 0xfc, 0x6a, 0xdf, 0xe0, 0x67, 0x85, 0x05, 0xd1, 0x44, 
+    0x45, 0xbf, 0x2d, 0xd5, 0xa6, 0xd1, 0x7d, 0xa7, 0x5a, 0x82, 
+    0x44, 0x07, 0x02, 0x70, 0x07, 0x08, 0x86, 0x37, 0x23, 0x98, 
+    0x50, 0x54, 0x95, 0x99, 0x68, 0xa8, 0x3d, 0xb6, 0x23, 0x5c, 
+    0x7c, 0x38, 0x74, 0x77, 0x6b, 0xc4, 0xcf, 0xda, 0x6b, 0x7a, 
+    0x19, 0x20, 0xe1, 0x29, 0xdb, 0xf6, 0xe2, 0x3e, 0x43, 0x56, 
+    0x87, 0xeb, 0x12, 0x83, 0xb6, 0x0a, 0xcc, 0x20, 0x68, 0xdf, 
+    0x5b, 0xf8, 0xe0, 0x35, 0x65, 0x2e, 0xb8, 0xce, 0xa2, 0xa0, 
+    0x7d, 0xa0, 0x58, 0x36, 0xa6, 0x97, 0x4d, 0x3f, 0x55, 0x3f, 
+    0x30, 0x6f, 0x66, 0x06, 0x2f, 0x22, 0xd3, 0x6c, 0xab, 0x52, 
+    0x81, 0x51, 0x88, 0xdb, 0x9c, 0x3c, 0x2d, 0xce, 0xe0, 0x67, 
+    0x98, 0x00, 0x6e, 0xbb, 0xb7, 0xc6, 0x82, 0xb2, 0x7b, 0x41, 
+    0x41, 0xe6, 0x62, 0x99, 0x84, 0xe5, 0x58, 0xc5, 0xc9, 0xf0, 
+    0x5b, 0xc6, 0x2b, 0x2b, 0x75, 0x1d, 0xfe, 0x5b, 0x86, 0x9a, 
+    0x09, 0xdd, 0x60, 0x39, 0x50, 0xa0, 0xdb, 0xc5, 0x32, 0x7e, 
+    0x54, 0xc5, 0xad, 0xde, 0x65, 0x37, 0x1e, 0xf2, 0x71, 0x05, 
+    0x9a, 0x24, 0x24, 0xbb, 0x3f, 0x94, 0x6c, 0x7e, 0x46, 0x54, 
+    0x05, 0x8f, 0x49, 0x48, 0x29, 0x5e, 0x1a, 0x2b, 0x28, 0xca, 
+    0xbf, 0xb2, 0x56, 0x97, 0x01, 0x18, 0x7a, 0xf1, 0x22, 0xb6, 
+    0x72, 0x88, 0x76, 0xb2, 0x72, 0x70, 0x47, 0x6e, 0x74, 0x19, 
+    0x36, 0x76, 0x50, 0x99, 0x7e, 0x8e, 0x61, 0x54, 0x2c, 0x55, 
+    0xef, 0xf7, 0x27, 0x62, 0x2f, 0x7a, 0xa6, 0x0b, 0x33, 0xb3, 
+    0x21, 0xec, 0xde, 0xaf, 0x1e, 0x5d, 0x30, 0x42, 0x82, 0xd9, 
+    0x59, 0x5e, 0xc9, 0xcf, 0x5a, 0x5b, 0xf5, 0xf6, 0x77, 0xdf, 
+    0xc1, 0x55, 0x6d, 0x5a, 0x6f, 0x54, 0x9a, 0x1f, 0x2d, 0xc2, 
+    0x19, 0xe1, 0x54, 0x91, 0xe3, 0x1d, 0x01, 0xae, 0xcb, 0xd0, 
+    0x73, 0x99, 0x0f, 0x6a, 0x7d, 0x3a, 0x88, 0x81, 0x90, 0xf6, 
+    0x15, 0xb9, 0x80, 0xf8, 0x2f, 0x4b, 0x39, 0x85, 0x17, 0xc5, 
+    0xf9, 0x76, 0x1d, 0x35, 0x2e, 0xbf, 0x68, 0x12, 0x6a, 0xb2, 
+    0xee, 0xc2, 0x4f, 0xa0, 0x1b, 0xa8, 0xad, 0xb9, 0x58, 0x3f, 
+    0x3b, 0xbb, 0x38, 0x21, 0x29, 0xa4, 0xf5, 0x53, 0xc7, 0x10, 
+    0xb7, 0x64, 0x52, 0xa4, 0x8b, 0xda, 0x22, 0x0d, 0xc5, 0xbc, 
+    0xb0, 0xb5, 0xea, 0x3c, 0x15, 0x6e, 0x18, 0x3a, 0x03, 0x92, 
+    0x49, 0xd5, 0x07, 0xdc, 0x30, 0x93, 0xb8, 0x8c, 0x3b, 0x38, 
+    0xc4, 0x96, 0xef, 0x7f, 0xb6, 0x6f, 0x0a, 0xe4, 0x8e, 0x4c, 
+    0xbb, 0x15, 0x87, 0x2a, 0xa5, 0x97, 0xfe, 0xb1, 0xdf, 0x65, 
+    0xb4, 0xed, 0x9b, 0x1d, 0x7e, 0x6c, 0xa4, 0xa5, 0x20, 0x8f, 
+    0x09, 0xd6, 0x14, 0x33, 0xad, 0xcd, 0x25, 0xe8, 0x3b, 0x74, 
+    0x8e, 0x3b, 0x67, 0x07, 0x0f, 0x2b, 0x19, 0x35, 0x10, 0x98, 
+    0xab, 0x33, 0xaa, 0x8f, 0xa6, 0xf3, 0x9a, 0xc2, 0x89, 0x94, 
+    0xfd, 0x72, 0xb2, 0x94, 0x64, 0x77, 0x74, 0x84, 0xf7, 0x0a, 
+    0x47, 0xc9, 0x14, 0x5c, 0xe0, 0x19, 0xc6, 0xa8, 0xf0, 0xb9, 
+    0x13, 0x98, 0x11, 0xae, 0x76, 0x85, 0xc8, 0x38, 0xda, 0xe8, 
+    0x55, 0xb8, 0x1f, 0xe1, 0xf0, 0xfe, 0x77, 0x1c, 0x53, 0xb3, 
+    0xe9, 0x57, 0x1c, 0x2b, 0x08, 0x89, 0xe1, 0x88, 0x13, 0x24, 
+    0x57, 0x83, 0x9c, 0x85, 0xd1, 0x14, 0xf8, 0xf9, 0xdd, 0x29, 
+    0x30, 0x73, 0xae, 0xcd, 0xd0, 0x2a, 0xec, 0xa7, 0x44, 0x75, 
+    0x8a, 0x38, 0x3e, 0xd3, 0x5b, 0x24, 0x2b, 0xe0, 0x5a, 0x85, 
+    0x39, 0x8c, 0x0c, 0x51, 0x9b
+};
+
diff --git a/neureka/pointwise/src/output.c b/neureka/pointwise/src/output.c
new file mode 100644
index 0000000..0172aeb
--- /dev/null
+++ b/neureka/pointwise/src/output.c
@@ -0,0 +1,165 @@
+#include "output.h"
+
+#define OUTPUT_SIZE (1365)
+PI_L1 uint8_t output[OUTPUT_SIZE];
+
+#define GOLDEN_OUTPUT_SIZE (1365)
+PI_L2 uint8_t golden_output[GOLDEN_OUTPUT_SIZE] = {
+    0xcd, 0x00, 0x00, 0x00, 0x63, 0x15, 0x00, 0x00, 0x00, 0x00, 
+    0x36, 0x6f, 0x00, 0x00, 0x27, 0x6d, 0x00, 0x2a, 0xff, 0x31, 
+    0xff, 0x8f, 0x00, 0x00, 0x01, 0x64, 0x55, 0x16, 0xb8, 0x0a, 
+    0x00, 0x00, 0x57, 0x00, 0x00, 0x00, 0x8c, 0x00, 0x00, 0x77, 
+    0x00, 0x1b, 0x00, 0x9b, 0x00, 0x00, 0x0d, 0x00, 0x00, 0x30, 
+    0x60, 0x9b, 0x37, 0x1b, 0xff, 0x00, 0x00, 0xff, 0x00, 0xff, 
+    0xbf, 0x00, 0x00, 0x00, 0xff, 0x3d, 0x22, 0x00, 0x00, 0x00, 
+    0x00, 0x4e, 0x00, 0x00, 0x45, 0xe0, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x93, 0x00, 0x00, 0x00, 0x00, 0x00, 0x3b, 0xff, 
+    0x00, 0x28, 0x00, 0xff, 0x00, 0x00, 0xff, 0x77, 0xf8, 0x4b, 
+    0x00, 0x00, 0x00, 0x8d, 0x37, 0x1c, 0x65, 0x00, 0x05, 0x00, 
+    0x8d, 0x00, 0x01, 0x50, 0xce, 0x00, 0x00, 0x8c, 0x00, 0x93, 
+    0x00, 0x17, 0x00, 0x00, 0x00, 0x00, 0x00, 0x3b, 0x2f, 0x6e, 
+    0x21, 0x00, 0xff, 0x00, 0x00, 0xff, 0x00, 0x4b, 0xcd, 0x00, 
+    0x00, 0x00, 0xff, 0x44, 0x0c, 0x5d, 0x00, 0x2f, 0x00, 0x7c, 
+    0x00, 0x00, 0x1d, 0xf9, 0x00, 0x00, 0x5b, 0x00, 0xd8, 0x00, 
+    0xf6, 0xdc, 0x00, 0x00, 0x00, 0x00, 0x33, 0x34, 0x00, 0x21, 
+    0x19, 0xe8, 0x95, 0x4e, 0xff, 0x00, 0xff, 0x9e, 0x00, 0x00, 
+    0x00, 0x9b, 0x41, 0x23, 0x47, 0x2c, 0x00, 0x00, 0x7d, 0x00, 
+    0x00, 0x00, 0x49, 0x00, 0x00, 0xee, 0x00, 0xee, 0x00, 0x86, 
+    0x71, 0x00, 0x00, 0x00, 0x00, 0x2e, 0x2d, 0xe3, 0x7a, 0x53, 
+    0xdc, 0x74, 0x14, 0xff, 0x66, 0xff, 0xcc, 0x13, 0x00, 0x00, 
+    0xce, 0x2e, 0x15, 0x93, 0x09, 0x24, 0x00, 0x70, 0x00, 0x00, 
+    0x00, 0xe9, 0x0e, 0x00, 0x3a, 0x00, 0x25, 0x00, 0x02, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x34, 0x93, 0xea, 0xc3, 0x59, 0x8a, 
+    0x00, 0xbb, 0xda, 0x3e, 0xff, 0xc3, 0x00, 0x00, 0x00, 0xff, 
+    0x4a, 0x12, 0x00, 0x34, 0x00, 0x00, 0x70, 0x00, 0x00, 0x00, 
+    0xeb, 0x00, 0x00, 0x00, 0x00, 0x1c, 0x08, 0x86, 0xb1, 0x14, 
+    0x14, 0x00, 0x00, 0x30, 0x5a, 0xad, 0x48, 0x00, 0xff, 0x00, 
+    0x00, 0xff, 0x00, 0xff, 0xda, 0x00, 0x00, 0x00, 0x62, 0x4b, 
+    0x22, 0x49, 0x00, 0x00, 0x00, 0x8a, 0x00, 0x00, 0x24, 0x64, 
+    0x00, 0x00, 0xff, 0x00, 0x9d, 0x00, 0x4b, 0xc1, 0x00, 0x00, 
+    0x00, 0x00, 0x2e, 0x00, 0x11, 0x2b, 0x8f, 0xff, 0x63, 0x00, 
+    0xd4, 0x12, 0xff, 0xcc, 0x00, 0x00, 0x00, 0x93, 0x5f, 0x1b, 
+    0x0c, 0x00, 0x43, 0x00, 0x6f, 0x00, 0x00, 0x00, 0xff, 0x00, 
+    0x00, 0x55, 0x00, 0x54, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x35, 0xff, 0x00, 0x00, 0x15, 0xf9, 0x00, 0x00, 0xe8, 
+    0x00, 0xff, 0xff, 0x00, 0x00, 0x02, 0x4f, 0x4a, 0x18, 0xc4, 
+    0x22, 0x09, 0x00, 0x6b, 0x00, 0x00, 0x08, 0xff, 0x00, 0x00, 
+    0x8c, 0x00, 0x00, 0x00, 0x00, 0x00, 0x74, 0x00, 0x00, 0x00, 
+    0x2f, 0xdf, 0x00, 0x00, 0x97, 0xff, 0x00, 0x00, 0xba, 0x00, 
+    0xff, 0xa5, 0x00, 0x00, 0x00, 0x9c, 0x13, 0x12, 0xbb, 0x00, 
+    0x00, 0x00, 0x4d, 0x00, 0x00, 0x00, 0xff, 0x00, 0x00, 0x11, 
+    0x00, 0x00, 0x00, 0x8a, 0x07, 0x00, 0x00, 0x00, 0x00, 0x36, 
+    0xd2, 0x8a, 0x6b, 0x00, 0xa2, 0x08, 0x33, 0xf1, 0x0d, 0x2a, 
+    0xc6, 0x00, 0x00, 0x00, 0x36, 0x54, 0x1a, 0x9e, 0x00, 0x10, 
+    0x00, 0x59, 0x00, 0x00, 0x00, 0xa0, 0x00, 0x00, 0x00, 0x00, 
+    0x4b, 0x00, 0x40, 0x00, 0x00, 0x00, 0x00, 0x00, 0x2e, 0xaf, 
+    0x56, 0x36, 0x18, 0xb5, 0x00, 0x00, 0x87, 0x00, 0x09, 0xaf, 
+    0x00, 0x00, 0x00, 0xec, 0x3b, 0x22, 0x64, 0x00, 0x47, 0x00, 
+    0xb7, 0x00, 0x00, 0x00, 0x9e, 0x15, 0x00, 0x3b, 0x00, 0x00, 
+    0x00, 0x63, 0xa5, 0x00, 0x18, 0x00, 0x00, 0x34, 0x97, 0x00, 
+    0x9e, 0x23, 0xa2, 0x64, 0x57, 0xff, 0x00, 0xf2, 0xef, 0x00, 
+    0x00, 0x00, 0xf2, 0x4f, 0x1b, 0x5e, 0x06, 0x00, 0x00, 0x8c, 
+    0x00, 0x00, 0x0a, 0xd6, 0x00, 0x00, 0x38, 0x00, 0x61, 0x00, 
+    0x90, 0xbf, 0x00, 0x00, 0x00, 0x00, 0x2d, 0x00, 0xc8, 0x80, 
+    0x00, 0xa9, 0x01, 0x01, 0xfb, 0x4b, 0xff, 0xcb, 0x20, 0x00, 
+    0x00, 0xff, 0x4e, 0x13, 0x94, 0x00, 0x00, 0x00, 0x54, 0x00, 
+    0x00, 0x00, 0xff, 0x00, 0x00, 0xe9, 0x00, 0x14, 0x31, 0x0c, 
+    0xd1, 0x00, 0x00, 0x00, 0x00, 0x35, 0x97, 0x46, 0x00, 0x0d, 
+    0xff, 0x00, 0x00, 0xff, 0x00, 0xb7, 0xbf, 0x00, 0x00, 0x00, 
+    0xa0, 0x3d, 0x11, 0x9f, 0x00, 0x51, 0x00, 0x7f, 0x00, 0x00, 
+    0x00, 0x96, 0x00, 0x00, 0x82, 0x00, 0x6a, 0x00, 0x51, 0xae, 
+    0x00, 0x00, 0x00, 0x00, 0x31, 0xff, 0x00, 0xae, 0x00, 0xf2, 
+    0x79, 0x00, 0xd7, 0x00, 0xff, 0xef, 0x05, 0x00, 0x00, 0xf0, 
+    0x44, 0x20, 0xff, 0xed, 0x4d, 0x00, 0x63, 0x00, 0x00, 0x81, 
+    0xff, 0x00, 0x00, 0xda, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x35, 0x95, 0xff, 0x9d, 0x00, 0xff, 0x00, 
+    0x00, 0xff, 0x62, 0xff, 0xbb, 0x00, 0x00, 0x00, 0x9e, 0x3c, 
+    0x1e, 0x97, 0x1f, 0x52, 0x00, 0x53, 0x00, 0x00, 0x43, 0xff, 
+    0x00, 0x00, 0xa6, 0x00, 0x72, 0x00, 0xfb, 0x4b, 0x01, 0x00, 
+    0x00, 0x00, 0x31, 0x86, 0x00, 0x36, 0x00, 0xf9, 0x00, 0xa3, 
+    0xff, 0xed, 0xff, 0xb5, 0x00, 0x00, 0x00, 0x6a, 0x3c, 0x19, 
+    0x8c, 0x63, 0x00, 0x00, 0x57, 0x00, 0x00, 0x00, 0xff, 0x00, 
+    0x00, 0xff, 0x00, 0x00, 0x00, 0x00, 0xc9, 0x00, 0x00, 0x00, 
+    0x00, 0x34, 0x9e, 0x11, 0x2b, 0x00, 0x36, 0x00, 0x00, 0xd6, 
+    0x30, 0xff, 0x96, 0x00, 0x00, 0x00, 0x4b, 0x27, 0x1d, 0x31, 
+    0x00, 0x25, 0x00, 0x53, 0x00, 0x00, 0x00, 0xad, 0x00, 0x6c, 
+    0x70, 0x00, 0x0a, 0x00, 0x62, 0x28, 0x00, 0x00, 0x00, 0x00, 
+    0x32, 0xba, 0x49, 0x3e, 0x13, 0xeb, 0x27, 0x14, 0xff, 0x11, 
+    0x09, 0xdd, 0x00, 0x00, 0x00, 0xff, 0x3e, 0x1b, 0x92, 0x00, 
+    0x00, 0x00, 0x9b, 0x00, 0x00, 0x00, 0xf2, 0x00, 0x00, 0xf5, 
+    0x00, 0x16, 0x00, 0x00, 0xb4, 0x00, 0x00, 0x00, 0x00, 0x33, 
+    0x45, 0x01, 0x37, 0x96, 0xff, 0x00, 0xa8, 0xff, 0x00, 0xff, 
+    0xd5, 0x00, 0x00, 0x00, 0xb0, 0x2e, 0x19, 0xbe, 0x00, 0x00, 
+    0x00, 0x80, 0x00, 0x00, 0x3e, 0xf9, 0x00, 0x00, 0x69, 0x00, 
+    0x3d, 0x00, 0x7e, 0x00, 0x00, 0x00, 0x00, 0x00, 0x36, 0xd3, 
+    0x0e, 0x2a, 0x3c, 0xc3, 0x00, 0x32, 0xff, 0x71, 0xff, 0x77, 
+    0x00, 0x00, 0x00, 0xff, 0x56, 0x23, 0x9c, 0x00, 0x52, 0x00, 
+    0x7d, 0x00, 0x16, 0x5d, 0xa2, 0x00, 0x00, 0x77, 0x00, 0x16, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x33, 0x46, 0x00, 
+    0x2c, 0x17, 0xdc, 0x0a, 0x00, 0xff, 0x00, 0xff, 0xba, 0x00, 
+    0x00, 0x00, 0x9e, 0x33, 0x27, 0x5f, 0x3f, 0x00, 0x00, 0x75, 
+    0x00, 0x00, 0xbd, 0x54, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x36, 0xe6, 0x00, 0x00, 
+    0x00, 0xc2, 0xbb, 0x00, 0xff, 0x2a, 0xff, 0xff, 0x00, 0x00, 
+    0x00, 0x47, 0x51, 0x15, 0x71, 0x00, 0x00, 0x00, 0x8a, 0x00, 
+    0x00, 0x00, 0xff, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0xff, 
+    0x6b, 0x00, 0x00, 0x00, 0x00, 0x30, 0x3e, 0x5d, 0x8b, 0x32, 
+    0xde, 0x00, 0x00, 0xff, 0x49, 0xff, 0x92, 0x00, 0x00, 0x00, 
+    0x08, 0x4f, 0x12, 0xb3, 0x00, 0x71, 0x00, 0x83, 0x00, 0x0c, 
+    0x00, 0xe1, 0x00, 0x00, 0x00, 0x00, 0x3f, 0x00, 0x00, 0x57, 
+    0x00, 0x18, 0x00, 0x00, 0x2f, 0xd2, 0x00, 0x00, 0x55, 0xd9, 
+    0x5c, 0x26, 0xff, 0x00, 0xff, 0xc0, 0x00, 0x00, 0x00, 0xe2, 
+    0x3c, 0x23, 0x18, 0x0e, 0x28, 0x00, 0x66, 0x00, 0x00, 0x00, 
+    0xff, 0x00, 0x00, 0x26, 0x00, 0xb4, 0x00, 0x5f, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x35, 0xc6, 0x84, 0x01, 0x53, 0xff, 0x00, 
+    0x00, 0xff, 0x00, 0xfe, 0x7f, 0x10, 0x00, 0x00, 0xed, 0x2d, 
+    0x1d, 0xb2, 0x55, 0x00, 0x00, 0x5e, 0x00, 0x26, 0xd2, 0xff, 
+    0x00, 0x00, 0x03, 0x00, 0x5b, 0x00, 0x2e, 0x20, 0x00, 0x00, 
+    0x00, 0x00, 0x34, 0xff, 0x9a, 0x3e, 0x00, 0xff, 0x00, 0x00, 
+    0xff, 0x00, 0xff, 0xff, 0x00, 0x00, 0x00, 0x26, 0x55, 0x08, 
+    0x6a, 0x00, 0x1d, 0x00, 0x83, 0x00, 0x00, 0x7a, 0xff, 0x00, 
+    0x00, 0x28, 0x00, 0xe7, 0x00, 0x08, 0x0b, 0x0b, 0x00, 0x00, 
+    0x00, 0x30, 0xab, 0x00, 0x00, 0x00, 0x64, 0x0c, 0x1e, 0xe4, 
+    0x27, 0xff, 0x93, 0x05, 0x00, 0x00, 0xff, 0x00, 0x2b, 0x8a, 
+    0x65, 0x0a, 0x00, 0x5f, 0x00, 0x00, 0x30, 0xff, 0x00, 0x00, 
+    0x45, 0x00, 0x7d, 0x00, 0x19, 0x6c, 0x00, 0x00, 0x00, 0x00, 
+    0x3b, 0xf9, 0x00, 0x00, 0x06, 0xa7, 0x00, 0x6f, 0xfd, 0x40, 
+    0xff, 0xb7, 0x00, 0x00, 0x00, 0xff, 0x22, 0x1e, 0xb5, 0x00, 
+    0x15, 0x00, 0x92, 0x00, 0x00, 0x54, 0xff, 0x00, 0x00, 0x66, 
+    0x00, 0x00, 0x00, 0x62, 0x00, 0x00, 0x00, 0x00, 0x00, 0x33, 
+    0x5b, 0x82, 0x26, 0x60, 0xff, 0x00, 0x00, 0xff, 0x6c, 0xff, 
+    0xfe, 0x00, 0x00, 0x00, 0x89, 0x63, 0x14, 0x69, 0xe5, 0x00, 
+    0x00, 0x80, 0x00, 0x00, 0x2b, 0xcf, 0x00, 0x00, 0x00, 0x00, 
+    0x3d, 0x00, 0x84, 0x3d, 0x00, 0x00, 0x00, 0x00, 0x37, 0xaf, 
+    0x00, 0x00, 0x3e, 0xff, 0x00, 0x42, 0xff, 0x00, 0xff, 0xff, 
+    0x00, 0x00, 0x04, 0x09, 0x3f, 0x1e, 0xaf, 0x17, 0x00, 0x00, 
+    0x5d, 0x00, 0x00, 0x22, 0xff, 0x00, 0x1c, 0x56, 0x00, 0x00, 
+    0x00, 0xd5, 0x19, 0x00, 0x00, 0x00, 0x00, 0x35, 0xc9, 0xff, 
+    0x00, 0x00, 0x88, 0x00, 0x0b, 0xff, 0x23, 0xff, 0xff, 0x00, 
+    0x00, 0x00, 0x0b, 0x4e, 0x17, 0x3c, 0x00, 0x00, 0x00, 0x61, 
+    0x00, 0x00, 0x00, 0x52, 0x00, 0x00, 0xb8, 0x00, 0xff, 0x00, 
+    0x45, 0x07, 0x00, 0x00, 0x00, 0x00, 0x2d, 0xe5, 0x00, 0x47, 
+    0x01, 0xff, 0x00, 0x00, 0xf7, 0x27, 0xff, 0x44, 0x00, 0x00, 
+    0x00, 0x4e, 0x5e, 0x17, 0xb2, 0x00, 0x00, 0x00, 0x5a, 0x00, 
+    0x00, 0x00, 0xff, 0x00, 0x36
+};
+
+int check_output() {
+        printf("Checking the output vector:\n");
+
+        int n_err = 0;
+        for (int i = 0; i < OUTPUT_SIZE; i++) {
+            if (output[i] != golden_output[i]) {
+                printf("ERROR: wrong value of output @ %d: %d vs. golden: %d\n", i, output[i], golden_output[i]);
+                n_err++;
+            }
+        }
+
+        if (n_err == 0)
+            printf("> Success! No errors found.\n");
+        else
+            printf("> Failure! Found %d/%d errors.\n", n_err, OUTPUT_SIZE);
+        return n_err;
+    }
+
+    
\ No newline at end of file
diff --git a/neureka/pointwise/src/scale.c b/neureka/pointwise/src/scale.c
new file mode 100644
index 0000000..5eb2c42
--- /dev/null
+++ b/neureka/pointwise/src/scale.c
@@ -0,0 +1,10 @@
+#include "scale.h"
+
+#define SCALE_SIZE (39)
+PI_L1 uint8_t scale[SCALE_SIZE] = {
+    0x1e, 0x02, 0x1a, 0x1d, 0x15, 0x1b, 0x13, 0x07, 0x15, 0x0e, 
+    0x01, 0x11, 0x1b, 0x11, 0x0e, 0x11, 0x19, 0x13, 0x0d, 0x13, 
+    0x1e, 0x0f, 0x0b, 0x0b, 0x05, 0x1b, 0x04, 0x02, 0x0c, 0x14, 
+    0x0c, 0x04, 0x05, 0x0f, 0x07, 0x0f, 0x0f, 0x16, 0x1c
+};
+
diff --git a/neureka/pointwise/src/weight.c b/neureka/pointwise/src/weight.c
new file mode 100644
index 0000000..140952b
--- /dev/null
+++ b/neureka/pointwise/src/weight.c
@@ -0,0 +1,1130 @@
+#include "weight.h"
+
+#define WEIGHT_SIZE (11232)
+PI_L1 uint8_t weight[WEIGHT_SIZE] = {
+    0x9f, 0x3b, 0x8a, 0x72, 0xf7, 0x43, 0x00, 0x69, 0xa1, 0xc9, 
+    0x09, 0x17, 0xa7, 0xab, 0xc5, 0xf7, 0xa7, 0xfa, 0x97, 0xa8, 
+    0xa5, 0x92, 0x11, 0xdf, 0xa7, 0x92, 0x91, 0xff, 0x58, 0x6d, 
+    0x6e, 0x00, 0x18, 0xb6, 0x1f, 0x16, 0x49, 0x91, 0x15, 0x60, 
+    0x06, 0x38, 0x2d, 0x55, 0x2d, 0xa3, 0x6e, 0x7e, 0xd2, 0x59, 
+    0x80, 0x79, 0x8f, 0x11, 0x4a, 0x7e, 0x8f, 0x11, 0x4a, 0x7e, 
+    0x70, 0xee, 0xb5, 0x81, 0x68, 0x39, 0xc7, 0xb5, 0x59, 0x5c, 
+    0xcd, 0x89, 0xfa, 0x75, 0x35, 0xdd, 0x0d, 0x89, 0x5c, 0xcd, 
+    0x9a, 0xee, 0x05, 0xba, 0x3d, 0xbc, 0x05, 0xca, 0x1d, 0xbc, 
+    0x05, 0xca, 0xe2, 0x43, 0xfa, 0x35, 0xee, 0xf7, 0xfa, 0x2d, 
+    0x88, 0xfb, 0x79, 0x5b, 0x2a, 0xc9, 0x15, 0x24, 0xe4, 0xbe, 
+    0x84, 0x12, 0x1d, 0x2d, 0x28, 0x42, 0xc6, 0xac, 0x00, 0xb2, 
+    0xc6, 0xac, 0x00, 0x12, 0x39, 0x53, 0xff, 0xed, 0xe0, 0x32, 
+    0xc6, 0x63, 0x0c, 0xea, 0x91, 0xf4, 0x00, 0x85, 0x05, 0x7f, 
+    0x6c, 0x88, 0x55, 0x3e, 0xfd, 0x98, 0x18, 0x1d, 0x6d, 0x9e, 
+    0x3b, 0x25, 0x6d, 0x9e, 0x3b, 0x25, 0x92, 0x61, 0xc4, 0xda, 
+    0x1e, 0x15, 0x6a, 0x82, 0x3d, 0x68, 0xeb, 0x2f, 0x6e, 0x76, 
+    0x1d, 0x53, 0xbb, 0x91, 0x91, 0xa7, 0x5c, 0x4a, 0xfa, 0x3a, 
+    0xc9, 0x01, 0x2a, 0x1b, 0xc9, 0x01, 0x6a, 0x1b, 0x36, 0xfe, 
+    0x95, 0xe4, 0x13, 0x69, 0xdd, 0xf5, 0xdc, 0xed, 0xb0, 0x39, 
+    0xdc, 0xda, 0xf6, 0x76, 0xaf, 0x02, 0xaa, 0xc4, 0x42, 0x3a, 
+    0xe9, 0x4f, 0x4a, 0x1b, 0xff, 0x4f, 0xca, 0x1a, 0xfb, 0x4f, 
+    0x35, 0xe5, 0x04, 0xb0, 0x2f, 0x58, 0xab, 0xfa, 0x4c, 0x98, 
+    0x7f, 0x23, 0x21, 0xa9, 0xfd, 0xce, 0x34, 0xf0, 0x3a, 0x67, 
+    0x5b, 0x89, 0x6f, 0x3b, 0x01, 0x88, 0x26, 0x3b, 0x01, 0x88, 
+    0x2e, 0xbb, 0xfe, 0x77, 0xd1, 0x44, 0x00, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0xf7, 0xd1, 
+    0x59, 0xb4, 0x26, 0x04, 0x6e, 0x6b, 0xb0, 0x5c, 0x36, 0xaf, 
+    0xde, 0x83, 0x05, 0xfe, 0x12, 0xc3, 0xe5, 0x71, 0xf6, 0x8f, 
+    0x0c, 0xcd, 0xf6, 0x8f, 0x24, 0xcd, 0x09, 0x70, 0xdb, 0x32, 
+    0xb0, 0x48, 0x23, 0x45, 0xb7, 0x93, 0xae, 0xcc, 0x66, 0x56, 
+    0x86, 0x3f, 0xa5, 0xe8, 0x76, 0xbf, 0xcf, 0x08, 0xa1, 0xed, 
+    0xff, 0x08, 0x1d, 0xad, 0xff, 0x08, 0x15, 0xad, 0x00, 0xf7, 
+    0xea, 0x52, 0x78, 0x44, 0x14, 0xba, 0x06, 0x42, 0x66, 0xcb, 
+    0xc3, 0xe8, 0x9d, 0x75, 0x48, 0x9b, 0x03, 0x94, 0x7b, 0x80, 
+    0xff, 0x8a, 0xfb, 0x11, 0xce, 0x96, 0x5b, 0x11, 0xce, 0x92, 
+    0xa4, 0xee, 0x31, 0x6d, 0xe5, 0xe4, 0x8d, 0x49, 0x4d, 0xb4, 
+    0x11, 0x98, 0x0c, 0x8e, 0x94, 0xbd, 0x3e, 0xfc, 0xbc, 0x24, 
+    0xe0, 0x7c, 0xfa, 0x85, 0x95, 0xcc, 0x9d, 0x87, 0x94, 0x4c, 
+    0x9d, 0x87, 0x6b, 0xb3, 0x62, 0x78, 0x01, 0x54, 0xa5, 0xaa, 
+    0xd9, 0xdd, 0x03, 0xe6, 0xfc, 0x51, 0xce, 0x8e, 0xd4, 0x0c, 
+    0x6a, 0xe7, 0x4c, 0xf4, 0xfd, 0x13, 0x58, 0x74, 0x87, 0xb3, 
+    0x58, 0x74, 0x87, 0xbb, 0xa7, 0x8b, 0x78, 0x44, 0x8a, 0x0d, 
+    0x4a, 0xe6, 0x03, 0x6e, 0x2c, 0xfe, 0xb5, 0x8b, 0x96, 0xbd, 
+    0xc1, 0xdc, 0xf7, 0xb1, 0x51, 0xc7, 0x5f, 0x03, 0x51, 0xf7, 
+    0x7b, 0x95, 0x51, 0xe7, 0x7b, 0x95, 0xae, 0x18, 0x84, 0x6a, 
+    0x54, 0x66, 0x1d, 0xda, 0x22, 0x63, 0xdd, 0x34, 0x7a, 0x6a, 
+    0x2e, 0x56, 0x99, 0x82, 0x9a, 0xc7, 0xf2, 0x7e, 0x90, 0x5e, 
+    0x7a, 0x4e, 0xda, 0x76, 0xfa, 0x4e, 0xda, 0x76, 0x05, 0xb1, 
+    0x25, 0x89, 0x0d, 0xef, 0x72, 0xd8, 0x3c, 0xe2, 0x2a, 0x60, 
+    0x9d, 0xd5, 0xf0, 0x43, 0x2b, 0x50, 0x86, 0x30, 0xea, 0x59, 
+    0x9d, 0x45, 0xe8, 0x18, 0xd9, 0xe7, 0xe8, 0x18, 0xd9, 0xe7, 
+    0x17, 0xe7, 0x26, 0x18, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x69, 0x4a, 0x80, 0xce, 
+    0xd0, 0x4e, 0x9b, 0x8a, 0xba, 0xa5, 0xd9, 0x53, 0x75, 0x4c, 
+    0x98, 0x23, 0xcc, 0xe4, 0xd7, 0x31, 0x3d, 0x00, 0xfd, 0x0b, 
+    0x7d, 0x00, 0xdd, 0x0b, 0x82, 0xff, 0x22, 0xf4, 0x75, 0x49, 
+    0x5f, 0xcc, 0xdb, 0xae, 0xde, 0xab, 0xdd, 0x30, 0x8b, 0x2c, 
+    0xa6, 0xc4, 0xb5, 0xf1, 0x97, 0x01, 0x7a, 0x23, 0xa3, 0x95, 
+    0x0e, 0x2d, 0xa3, 0x95, 0x2e, 0x2d, 0x5c, 0x6a, 0xd1, 0xd2, 
+    0x65, 0xf0, 0x53, 0x99, 0xf0, 0x91, 0xef, 0x32, 0x96, 0xf7, 
+    0x12, 0xc3, 0xe0, 0x17, 0x13, 0xac, 0x3c, 0x53, 0x2f, 0x1f, 
+    0x54, 0x77, 0x33, 0xde, 0x14, 0x77, 0x33, 0x9e, 0xeb, 0x88, 
+    0xcc, 0x61, 0x02, 0x02, 0xca, 0xed, 0xa2, 0xa4, 0xcf, 0x20, 
+    0xf7, 0x92, 0xdb, 0x88, 0x11, 0x86, 0xee, 0xcf, 0x9a, 0xe0, 
+    0xc9, 0x6f, 0x9b, 0x82, 0x88, 0xee, 0x9b, 0x82, 0xc8, 0xee, 
+    0x64, 0x7d, 0x37, 0x11, 0x46, 0x01, 0xa3, 0xbc, 0x0e, 0xc4, 
+    0x72, 0xd1, 0xc3, 0x85, 0xad, 0xcb, 0x11, 0x25, 0x56, 0x89, 
+    0x4e, 0x27, 0xcd, 0xcd, 0xa6, 0xa7, 0x6c, 0x7d, 0x86, 0x27, 
+    0xec, 0xfd, 0x79, 0xd8, 0x13, 0x02, 0xa0, 0x26, 0xfe, 0xe8, 
+    0xfc, 0xf5, 0x64, 0x2d, 0x59, 0xa7, 0x92, 0xd4, 0x5a, 0xd3, 
+    0xbe, 0x77, 0x08, 0x4e, 0x04, 0x86, 0x50, 0xce, 0x36, 0x84, 
+    0x50, 0xce, 0x36, 0x84, 0xaf, 0x31, 0xc9, 0x7b, 0xbb, 0x25, 
+    0x67, 0x70, 0x0c, 0x20, 0x92, 0x7f, 0x30, 0xc2, 0xd0, 0x88, 
+    0x27, 0xaa, 0x32, 0xaf, 0x65, 0xd0, 0x19, 0x4f, 0x83, 0x98, 
+    0x1e, 0xb3, 0x03, 0x98, 0x1e, 0xb3, 0xfc, 0x67, 0xe1, 0x4c, 
+    0x73, 0x1e, 0x1f, 0xd8, 0xb7, 0x16, 0x15, 0x8b, 0x96, 0xc5, 
+    0x15, 0xa3, 0xa5, 0x2b, 0xae, 0x7a, 0x30, 0xb5, 0x0f, 0x89, 
+    0x2f, 0x92, 0x37, 0xa8, 0x27, 0x92, 0x17, 0xa8, 0xd8, 0x6d, 
+    0xe8, 0x57, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x4b, 0x8a, 0x05, 0xdc, 0x97, 0x61, 
+    0xcf, 0x05, 0xb4, 0x98, 0x55, 0x98, 0xc4, 0xc5, 0xd8, 0xb1, 
+    0x28, 0x49, 0x74, 0x9d, 0x89, 0x8b, 0x74, 0xbd, 0x89, 0xcb, 
+    0x74, 0xbd, 0x76, 0x34, 0x8b, 0x42, 0xe3, 0x67, 0x8e, 0x28, 
+    0x5e, 0x2e, 0x32, 0x1f, 0xdf, 0x9e, 0xa1, 0x1f, 0xcf, 0x0f, 
+    0x1b, 0x16, 0xaa, 0x3e, 0x33, 0x31, 0x83, 0x37, 0xb1, 0x75, 
+    0x83, 0x36, 0xb1, 0x75, 0x7c, 0xc9, 0x4e, 0x8a, 0xe4, 0x3f, 
+    0xc5, 0x8b, 0xd2, 0x10, 0xf2, 0xc2, 0x99, 0x4f, 0x15, 0xe1, 
+    0x72, 0xfc, 0x83, 0xf4, 0x79, 0x03, 0x05, 0x83, 0x79, 0x0f, 
+    0xc7, 0x80, 0x79, 0x0f, 0xc7, 0x80, 0x86, 0xf0, 0x38, 0x7f, 
+    0x91, 0x81, 0xee, 0x27, 0xf3, 0x13, 0xbd, 0x47, 0x55, 0xbf, 
+    0xe1, 0x5e, 0xa4, 0x47, 0xed, 0xf3, 0x1b, 0x7b, 0x41, 0x5b, 
+    0x26, 0x5b, 0xe5, 0x7b, 0x26, 0x5b, 0x65, 0x7b, 0xd9, 0xa4, 
+    0x9a, 0x84, 0x3b, 0x70, 0x1e, 0x38, 0x31, 0x21, 0x9c, 0x48, 
+    0xe7, 0x9c, 0x1b, 0x8d, 0x67, 0xc9, 0x9c, 0x89, 0x26, 0x1a, 
+    0xc3, 0x2f, 0x27, 0xd8, 0xf3, 0xbb, 0x27, 0x98, 0xb3, 0xab, 
+    0xd8, 0x67, 0x4c, 0x54, 0x48, 0xbd, 0xf9, 0x4f, 0x2c, 0x37, 
+    0xc9, 0xdd, 0x4d, 0xef, 0x59, 0x09, 0x73, 0x9d, 0xb8, 0x36, 
+    0x9b, 0x06, 0x11, 0xc0, 0xb3, 0x16, 0x5d, 0x09, 0xb3, 0x16, 
+    0x59, 0x09, 0x4c, 0xe9, 0xa6, 0xf6, 0x25, 0xf5, 0xf8, 0x08, 
+    0x9b, 0x75, 0x87, 0x32, 0x0a, 0xd0, 0x64, 0x04, 0xbb, 0x9d, 
+    0x5c, 0x30, 0x75, 0xa3, 0x9e, 0x55, 0xb7, 0x94, 0xcd, 0x01, 
+    0xb7, 0x97, 0xcd, 0x01, 0x48, 0x68, 0x32, 0xfe, 0x4f, 0x22, 
+    0x33, 0xd1, 0x46, 0x2a, 0x5d, 0xc7, 0x69, 0x59, 0xf0, 0xc2, 
+    0x1e, 0xc6, 0xb6, 0xca, 0xa0, 0x1f, 0xd6, 0x1f, 0x2b, 0xc6, 
+    0xdc, 0xfb, 0x29, 0xc6, 0xd4, 0xfb, 0xd6, 0x39, 0x2b, 0x04, 
+    0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x78, 0xdb, 0x23, 0x5a, 0xc8, 0xac, 0xb1, 0x69, 
+    0x2d, 0x8a, 0xc1, 0x6a, 0x7b, 0x9c, 0x59, 0x1c, 0xc6, 0x09, 
+    0x3e, 0x88, 0xfe, 0x00, 0x5b, 0x68, 0xfe, 0x00, 0x5b, 0x68, 
+    0x01, 0xff, 0xa4, 0x97, 0x62, 0xd2, 0xf9, 0xdf, 0x04, 0x54, 
+    0x66, 0xcf, 0x69, 0x13, 0x3e, 0x79, 0xc4, 0xa3, 0x28, 0x1c, 
+    0x81, 0xb4, 0x86, 0xa5, 0xc4, 0xa4, 0xa8, 0x11, 0xc4, 0xa4, 
+    0xa0, 0x11, 0x3b, 0x5b, 0x5f, 0xee, 0xbe, 0x03, 0x85, 0x09, 
+    0x4c, 0x70, 0xa5, 0x8a, 0x8a, 0xe9, 0x64, 0x34, 0x4d, 0x6b, 
+    0x08, 0xe0, 0x6f, 0x60, 0xdc, 0x8c, 0x4f, 0x72, 0xc8, 0xa4, 
+    0x4f, 0x70, 0xc8, 0xac, 0xb0, 0x8f, 0x37, 0x53, 0xd0, 0xd6, 
+    0xbe, 0x82, 0x0b, 0x60, 0xbc, 0x0b, 0xea, 0xc9, 0x00, 0xb8, 
+    0x4d, 0xf7, 0xc5, 0x2b, 0x43, 0xc2, 0xcd, 0x2c, 0x4b, 0x4a, 
+    0x85, 0x16, 0x4b, 0xc8, 0xc5, 0x16, 0xb4, 0x37, 0x3a, 0xe9, 
+    0x1f, 0x07, 0x35, 0x29, 0x54, 0x16, 0x57, 0xfc, 0xac, 0x94, 
+    0x3b, 0xf2, 0xb7, 0x54, 0xa6, 0x89, 0xc4, 0xa5, 0x94, 0x63, 
+    0x85, 0x87, 0xe3, 0xe7, 0x85, 0x87, 0xa2, 0xe3, 0x7a, 0x78, 
+    0x5d, 0x1c, 0x11, 0xa7, 0xf8, 0x78, 0x8c, 0x11, 0xa1, 0xe8, 
+    0x75, 0xbf, 0x6e, 0x3f, 0x05, 0xe7, 0x2d, 0xae, 0x9e, 0xed, 
+    0x9a, 0x32, 0xd1, 0xef, 0xdb, 0x22, 0xd1, 0xef, 0x9b, 0x22, 
+    0x2e, 0x10, 0x64, 0xdd, 0x00, 0xb6, 0x03, 0x5d, 0x88, 0x5b, 
+    0x8f, 0xc7, 0x71, 0x40, 0x83, 0x28, 0x6f, 0x1d, 0xf3, 0xf3, 
+    0x7c, 0x40, 0xd4, 0xe9, 0x4c, 0x51, 0xc8, 0xe3, 0x4c, 0x51, 
+    0xc4, 0xe3, 0xb3, 0xae, 0x3b, 0x1c, 0xa8, 0x94, 0x08, 0x8f, 
+    0x88, 0x6d, 0x96, 0x3d, 0xcd, 0x50, 0xa6, 0xf4, 0x58, 0x41, 
+    0xd7, 0xa1, 0x58, 0x3a, 0x2d, 0x4e, 0x74, 0x78, 0x2f, 0xcc, 
+    0x74, 0x78, 0x2f, 0xcc, 0x8b, 0x87, 0xd0, 0x33, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0xca, 0xd3, 0x0c, 0x83, 0x34, 0xb6, 0x8c, 0x65, 0xb7, 0x0f, 
+    0x20, 0x7d, 0xf7, 0xaa, 0x57, 0xc3, 0x1b, 0x4a, 0xb4, 0xcf, 
+    0x6b, 0x2a, 0xb4, 0x76, 0x7b, 0x2a, 0xb4, 0x76, 0x84, 0xd5, 
+    0x4b, 0x89, 0x47, 0xe6, 0x1e, 0xd2, 0x99, 0x2c, 0x18, 0xf7, 
+    0x80, 0x02, 0x61, 0x8a, 0x12, 0xfc, 0x89, 0xa3, 0x99, 0x93, 
+    0x64, 0xec, 0x8d, 0xb2, 0xe1, 0xf7, 0x89, 0xb3, 0xe1, 0xf7, 
+    0x76, 0x4c, 0x1e, 0x08, 0x1f, 0xff, 0x88, 0x4b, 0x72, 0x31, 
+    0x0e, 0x14, 0xd1, 0x68, 0x91, 0x9d, 0x2d, 0x90, 0xa8, 0x8d, 
+    0xc1, 0x85, 0xf8, 0x6e, 0x14, 0x80, 0xb2, 0x6d, 0x04, 0x80, 
+    0xb8, 0x6d, 0xfb, 0x7f, 0x47, 0x92, 0x6c, 0x25, 0xfe, 0xbb, 
+    0x22, 0x74, 0xac, 0x32, 0x08, 0xd0, 0xb9, 0x98, 0x8a, 0x2c, 
+    0xc9, 0xba, 0xc4, 0x03, 0x70, 0x24, 0xcc, 0x07, 0xbd, 0x38, 
+    0xcc, 0x07, 0xbc, 0x38, 0x33, 0xf8, 0x43, 0xc7, 0x03, 0x8f, 
+    0x12, 0x85, 0x7e, 0xc2, 0xa2, 0x58, 0x2e, 0xdf, 0xee, 0xd9, 
+    0x5a, 0x4e, 0x35, 0x0f, 0xdc, 0x19, 0x68, 0xc3, 0x56, 0x9a, 
+    0x01, 0x5b, 0x5e, 0x1a, 0x41, 0x5b, 0xa1, 0xe5, 0xbe, 0xa4, 
+    0x61, 0x55, 0xb5, 0x36, 0x7c, 0xaf, 0x79, 0x28, 0xee, 0x65, 
+    0x4c, 0x17, 0x29, 0xa6, 0x9a, 0xe1, 0xdb, 0xd6, 0xfa, 0x83, 
+    0x5f, 0xe8, 0xac, 0x29, 0x5d, 0xc6, 0xa8, 0x29, 0xa2, 0x39, 
+    0x57, 0xd6, 0xf4, 0xa1, 0x69, 0x75, 0x8b, 0x66, 0x66, 0xbd, 
+    0xe9, 0xd5, 0x3f, 0xda, 0x69, 0x1e, 0x0a, 0x95, 0x47, 0xe9, 
+    0x00, 0x5b, 0xa3, 0xe0, 0x79, 0x0e, 0xa3, 0xe8, 0x58, 0x1a, 
+    0x5c, 0x17, 0xa7, 0xe5, 0xc2, 0x17, 0xe0, 0xda, 0x9e, 0x76, 
+    0x68, 0x43, 0x21, 0xd1, 0xff, 0x6c, 0x24, 0x85, 0x79, 0x50, 
+    0x75, 0xf4, 0x02, 0xcb, 0xe5, 0xfc, 0x44, 0xe7, 0x65, 0xf4, 
+    0x40, 0xee, 0x9a, 0x0b, 0xbf, 0x11, 0x01, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x07, 0x5f, 
+    0x4b, 0x87, 0xac, 0x0d, 0xca, 0xed, 0x8e, 0x1c, 0x34, 0xe7, 
+    0xaa, 0x88, 0xa2, 0x36, 0x5c, 0x47, 0xa8, 0xd2, 0x00, 0x04, 
+    0xeb, 0x96, 0x04, 0x04, 0xea, 0x96, 0xfb, 0xfb, 0x15, 0x69, 
+    0x0c, 0x78, 0xa1, 0x71, 0x29, 0x4b, 0x5e, 0x7f, 0x2a, 0xdd, 
+    0x8e, 0x99, 0xc7, 0x3a, 0x91, 0x94, 0xee, 0x9b, 0x1d, 0x25, 
+    0xbe, 0x11, 0x15, 0x00, 0xbe, 0x19, 0x1d, 0x00, 0x41, 0xe6, 
+    0xe2, 0xff, 0x88, 0xdb, 0x6b, 0x02, 0x47, 0xe6, 0x35, 0xe8, 
+    0xa7, 0x9b, 0x5b, 0x99, 0xc6, 0xb8, 0xd7, 0xa2, 0x08, 0x03, 
+    0x76, 0x40, 0xe0, 0x97, 0xfb, 0xba, 0xc0, 0x97, 0xfb, 0xaa, 
+    0x3f, 0x68, 0x04, 0x55, 0x4a, 0x42, 0x8e, 0xa3, 0x33, 0x66, 
+    0xa3, 0x39, 0x4e, 0xc0, 0xf9, 0x7e, 0x05, 0x86, 0x00, 0x34, 
+    0x8b, 0x7f, 0x5e, 0x3b, 0x80, 0x6f, 0x5c, 0xbb, 0x81, 0x6f, 
+    0x5c, 0xbb, 0x7e, 0x90, 0xa3, 0x44, 0xea, 0xb7, 0xac, 0x08, 
+    0x62, 0x5d, 0x4a, 0x25, 0x38, 0xd8, 0xef, 0xb8, 0x15, 0xa7, 
+    0xe9, 0xda, 0x80, 0xcf, 0x71, 0x23, 0x16, 0xa6, 0x3b, 0x73, 
+    0x14, 0xae, 0x7b, 0x73, 0xeb, 0x51, 0x84, 0x8c, 0x7a, 0x61, 
+    0x8d, 0x92, 0x48, 0x9f, 0xd7, 0x0e, 0x62, 0x8e, 0x5a, 0x58, 
+    0x47, 0x60, 0x35, 0x9c, 0x5a, 0x6e, 0x10, 0x2f, 0x0a, 0x19, 
+    0x1c, 0x3f, 0x0a, 0x19, 0x1c, 0x3f, 0xf5, 0xe6, 0xe3, 0xc0, 
+    0x07, 0x93, 0x9e, 0x91, 0x7b, 0x6d, 0x0a, 0x4a, 0x2a, 0x21, 
+    0xdb, 0xfb, 0x40, 0x1c, 0x96, 0xff, 0xa1, 0x8e, 0x14, 0xde, 
+    0x12, 0xff, 0x30, 0x7e, 0x02, 0xff, 0x12, 0xfe, 0xfd, 0x00, 
+    0xed, 0x01, 0x42, 0x11, 0x91, 0xaa, 0xfe, 0x89, 0x1f, 0x1a, 
+    0x9b, 0xb7, 0xb2, 0xf6, 0xef, 0x73, 0x92, 0x49, 0x8d, 0x84, 
+    0xcc, 0x21, 0x98, 0x86, 0x98, 0x06, 0x98, 0x86, 0x98, 0x02, 
+    0x67, 0x79, 0x67, 0xfd, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x38, 0xcd, 0x21, 0x7b, 
+    0x6a, 0xaf, 0x7c, 0xf0, 0xc4, 0x58, 0x07, 0x7f, 0x2d, 0xfb, 
+    0xdb, 0xfe, 0xa7, 0x81, 0x9d, 0x89, 0x8d, 0xd9, 0xd9, 0xca, 
+    0xad, 0xd9, 0xd9, 0xca, 0x52, 0x26, 0x26, 0x35, 0x26, 0xe5, 
+    0xf5, 0xfe, 0x3b, 0xa4, 0x71, 0x1b, 0x44, 0xcb, 0xfd, 0xb6, 
+    0x9a, 0xf3, 0x27, 0x22, 0xc8, 0xfa, 0x78, 0x6c, 0xe9, 0xfa, 
+    0x6e, 0x74, 0xe8, 0xfa, 0x6e, 0x74, 0x17, 0x05, 0x91, 0x8b, 
+    0x93, 0x49, 0xa2, 0x64, 0xfc, 0x78, 0x4b, 0xb5, 0x20, 0x51, 
+    0x90, 0xb5, 0x4e, 0x65, 0x9e, 0xa3, 0x28, 0xa0, 0x04, 0xba, 
+    0x40, 0xa6, 0x48, 0x39, 0x40, 0xa6, 0x48, 0x39, 0xbf, 0x59, 
+    0xb7, 0xc6, 0x2f, 0xf4, 0xdf, 0x7b, 0x46, 0x8f, 0x30, 0x10, 
+    0x65, 0x59, 0x79, 0x8e, 0xc7, 0x96, 0xe4, 0xed, 0x3f, 0x2b, 
+    0x6d, 0x9f, 0xae, 0xad, 0xe9, 0x99, 0xae, 0xaf, 0x69, 0x9d, 
+    0x51, 0x50, 0x96, 0x62, 0xb9, 0x23, 0x8e, 0xa3, 0x0b, 0x15, 
+    0x87, 0x3a, 0x9b, 0x56, 0x2d, 0x71, 0x0e, 0x35, 0xb6, 0xea, 
+    0x8d, 0xce, 0xa1, 0x29, 0x8e, 0x2a, 0xfa, 0x0b, 0x8e, 0x0a, 
+    0xea, 0x09, 0x71, 0xf5, 0x15, 0xf6, 0x61, 0x8e, 0xce, 0xa5, 
+    0x47, 0xce, 0x20, 0x70, 0x3b, 0xb3, 0x8b, 0xa1, 0x0b, 0xbe, 
+    0x46, 0x5f, 0x2a, 0x45, 0xdc, 0x1b, 0x8b, 0x9a, 0x10, 0xbf, 
+    0x8b, 0x9a, 0x90, 0xbf, 0x74, 0x65, 0x6f, 0x40, 0xfd, 0x99, 
+    0x2b, 0x2c, 0x34, 0xd2, 0x98, 0x6c, 0xa2, 0x96, 0x7f, 0xeb, 
+    0x01, 0x7f, 0xce, 0xf3, 0x57, 0xde, 0x3a, 0xbc, 0x73, 0xbe, 
+    0x1d, 0xd9, 0x73, 0xbe, 0x1f, 0xd9, 0x8c, 0x41, 0xe0, 0x26, 
+    0xce, 0xa8, 0xec, 0x37, 0xd6, 0x13, 0x47, 0xb2, 0x58, 0xbd, 
+    0x88, 0xa5, 0x22, 0xbb, 0x85, 0x61, 0xaf, 0xbd, 0x6d, 0x5a, 
+    0x32, 0x89, 0xed, 0x6b, 0x3a, 0x99, 0xed, 0x6b, 0xc5, 0x66, 
+    0x12, 0x94, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x7c, 0x06, 0x63, 0x8a, 0x5e, 0x18, 
+    0xbb, 0x88, 0x3a, 0x7a, 0xf9, 0xd3, 0x4d, 0x85, 0xfe, 0x2a, 
+    0x19, 0x8f, 0x50, 0xc3, 0xcd, 0x23, 0x9f, 0xcc, 0xcd, 0x2b, 
+    0xdf, 0xce, 0x32, 0xd4, 0x20, 0x31, 0x81, 0x32, 0xe7, 0x3e, 
+    0x93, 0x3a, 0x17, 0x89, 0x21, 0x48, 0xf5, 0xd7, 0x60, 0x44, 
+    0x37, 0x04, 0x78, 0x53, 0xb3, 0xed, 0xb0, 0xcc, 0x3e, 0xad, 
+    0xb0, 0xc4, 0x36, 0xad, 0x4f, 0x3b, 0xc9, 0x52, 0x3a, 0x01, 
+    0x6b, 0xa2, 0xf2, 0x11, 0x50, 0x3d, 0x30, 0xb4, 0x5e, 0xfe, 
+    0x87, 0x19, 0x58, 0x7c, 0x7d, 0x90, 0xa4, 0x7a, 0xb9, 0x3c, 
+    0xb8, 0xf6, 0xb9, 0x3c, 0xb8, 0xfe, 0x46, 0xc3, 0x47, 0x01, 
+    0xbb, 0xb5, 0xfc, 0x72, 0xc6, 0x36, 0x7c, 0xbd, 0x08, 0x16, 
+    0xf8, 0x39, 0x0e, 0xcf, 0xed, 0xa8, 0x1a, 0x8f, 0xd0, 0x19, 
+    0x53, 0xe7, 0x55, 0x39, 0x13, 0xe7, 0x55, 0x39, 0xec, 0x18, 
+    0xaa, 0xc6, 0xc2, 0x10, 0x66, 0xbf, 0x83, 0xbf, 0x7a, 0x46, 
+    0x84, 0xe2, 0x90, 0xf9, 0x16, 0xf6, 0x0d, 0x2e, 0x1c, 0x8f, 
+    0xb0, 0x81, 0x08, 0x96, 0xf4, 0x1b, 0x18, 0x96, 0xf4, 0x1b, 
+    0xe7, 0x69, 0x0b, 0xe4, 0x7f, 0x88, 0xfd, 0xa9, 0xaa, 0x9c, 
+    0xb8, 0x31, 0x91, 0xf0, 0xe1, 0xa6, 0xe2, 0x4e, 0x8d, 0x99, 
+    0x15, 0xbe, 0x00, 0x0c, 0x79, 0xfa, 0x69, 0x97, 0x79, 0xfe, 
+    0x49, 0x95, 0x86, 0x01, 0xb6, 0x6a, 0xc8, 0xde, 0x64, 0x2c, 
+    0x07, 0x35, 0xdf, 0x26, 0x62, 0x1f, 0xa3, 0xc3, 0x95, 0xfc, 
+    0x2d, 0x04, 0x46, 0x90, 0x31, 0xbd, 0x25, 0x7d, 0x73, 0x0c, 
+    0x25, 0x7c, 0x73, 0x0c, 0xda, 0x83, 0x8c, 0xf3, 0xfe, 0x6b, 
+    0xd7, 0x46, 0x86, 0xf6, 0x0b, 0x5e, 0x55, 0xde, 0x4c, 0x5d, 
+    0x67, 0xec, 0x0f, 0x6c, 0x7c, 0x97, 0x6a, 0x2a, 0xe6, 0x8f, 
+    0x2a, 0xa3, 0x66, 0xdf, 0x2a, 0xab, 0x99, 0x20, 0xd5, 0x54, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x12, 0x00, 0x77, 0x8b, 0xee, 0x27, 0x42, 0x81, 
+    0x7c, 0x43, 0x31, 0xf3, 0x45, 0x89, 0xed, 0x09, 0x18, 0x79, 
+    0xda, 0x83, 0xce, 0xf8, 0x4f, 0x03, 0xce, 0xf8, 0xdb, 0x03, 
+    0x31, 0x07, 0x24, 0xfc, 0x84, 0xb3, 0x04, 0x7b, 0x22, 0xf6, 
+    0xa6, 0x45, 0x8a, 0x6a, 0xcb, 0x6f, 0x5e, 0x6b, 0x31, 0x55, 
+    0xe7, 0x2d, 0x17, 0x2a, 0xf4, 0x03, 0x33, 0x3b, 0xf6, 0x43, 
+    0x33, 0x3b, 0x09, 0xbc, 0xcc, 0xc4, 0x16, 0x4c, 0x18, 0xd8, 
+    0xc8, 0xa2, 0xa8, 0x45, 0xed, 0xd9, 0xc8, 0x6a, 0xe1, 0x39, 
+    0x8c, 0x26, 0x2c, 0x21, 0x8e, 0x26, 0x40, 0xf1, 0x86, 0x04, 
+    0x78, 0x71, 0x8e, 0x26, 0x87, 0x8e, 0x71, 0xd9, 0x61, 0x1e, 
+    0xb0, 0x82, 0x4b, 0x54, 0xb5, 0xde, 0x8d, 0xba, 0x26, 0xc7, 
+    0xbd, 0x54, 0xc7, 0x6f, 0x2c, 0x1f, 0x50, 0x38, 0x18, 0x5d, 
+    0x40, 0x3e, 0x18, 0x5d, 0x40, 0x3e, 0xe7, 0xa2, 0xbf, 0xc1, 
+    0x2d, 0x17, 0x5a, 0xb0, 0xa9, 0xdd, 0x60, 0x10, 0xe9, 0xd3, 
+    0x15, 0x8a, 0x7d, 0x94, 0x20, 0xf1, 0x8e, 0x0c, 0xdb, 0x48, 
+    0xef, 0x4b, 0xab, 0x9e, 0xef, 0x4b, 0xab, 0xde, 0x10, 0xb4, 
+    0x54, 0x21, 0x2c, 0x3e, 0x3a, 0x45, 0x53, 0x0e, 0x7f, 0x2f, 
+    0x44, 0x31, 0x55, 0x48, 0xec, 0x31, 0x2b, 0x38, 0xb4, 0xa0, 
+    0xe1, 0x0a, 0xbc, 0x0b, 0x61, 0x2b, 0xbc, 0x03, 0x61, 0x6b, 
+    0x43, 0xfc, 0x9e, 0xd4, 0x6c, 0x85, 0xa1, 0x59, 0x6c, 0x25, 
+    0x95, 0x9c, 0x63, 0xd9, 0xbc, 0x35, 0x18, 0x03, 0x89, 0x6b, 
+    0xc7, 0xd2, 0xf5, 0xee, 0xe3, 0x71, 0xc5, 0xaa, 0xe3, 0x51, 
+    0xc5, 0xea, 0x1c, 0xae, 0x3a, 0x15, 0x05, 0x1b, 0x2c, 0xd0, 
+    0x30, 0xc2, 0x94, 0x31, 0x54, 0xc8, 0x8b, 0x8f, 0xde, 0x62, 
+    0x1d, 0x58, 0x99, 0x77, 0x2c, 0xc9, 0x1c, 0xf3, 0x34, 0x4d, 
+    0x1c, 0xf3, 0x3c, 0x4d, 0xe3, 0x0c, 0xc3, 0xb2, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x14, 0x2e, 0x50, 0x55, 0x21, 0x4f, 0xb4, 0x93, 0x75, 0x95, 
+    0x3f, 0x2f, 0x6e, 0x70, 0x02, 0x44, 0x9e, 0xd4, 0xee, 0x10, 
+    0xeb, 0x75, 0x42, 0x4c, 0xeb, 0x75, 0x42, 0x44, 0x14, 0x8a, 
+    0xbd, 0xbb, 0xef, 0x2d, 0xb8, 0x77, 0x9f, 0x2c, 0x1d, 0x29, 
+    0xe0, 0x05, 0x9c, 0x5a, 0xcf, 0xf6, 0xf6, 0x79, 0x4b, 0x06, 
+    0xac, 0xef, 0xaa, 0x94, 0xac, 0xed, 0xca, 0x14, 0xac, 0xef, 
+    0x35, 0xeb, 0x53, 0x10, 0xb9, 0x22, 0x21, 0xcc, 0x95, 0x41, 
+    0x36, 0xec, 0x7c, 0x83, 0x66, 0xe0, 0x36, 0x48, 0x65, 0xe4, 
+    0x7a, 0xf2, 0xe2, 0x56, 0xb6, 0xf0, 0xe0, 0xee, 0xb6, 0xf0, 
+    0xe0, 0xee, 0x49, 0x0f, 0x1f, 0x11, 0x7d, 0x76, 0xd8, 0xa0, 
+    0xc1, 0x10, 0xb9, 0x4c, 0xcd, 0xb2, 0xa4, 0x27, 0x53, 0x97, 
+    0xe3, 0x7e, 0xe0, 0xfb, 0x95, 0x97, 0x61, 0x6a, 0x96, 0x1b, 
+    0x61, 0x7b, 0x96, 0x9b, 0x9e, 0x84, 0x69, 0x64, 0x3e, 0x37, 
+    0xa1, 0xb8, 0x3a, 0x09, 0xee, 0x34, 0x2a, 0xd8, 0x7a, 0x51, 
+    0x8b, 0x6c, 0xe8, 0x7d, 0x9c, 0x51, 0x8f, 0xc3, 0x90, 0x1d, 
+    0xbe, 0xc4, 0x90, 0x19, 0xbf, 0xc4, 0x6f, 0xe6, 0x40, 0x3b, 
+    0x13, 0xa1, 0x45, 0x1b, 0x0b, 0x8b, 0x22, 0xec, 0x34, 0x25, 
+    0x8e, 0x44, 0x8c, 0x9b, 0x72, 0x12, 0xb8, 0x8b, 0x5f, 0x54, 
+    0xb0, 0x70, 0x1a, 0x74, 0xb0, 0xba, 0x1a, 0x14, 0x4f, 0x45, 
+    0xe5, 0xeb, 0x52, 0xdc, 0x5b, 0x77, 0xdb, 0xa7, 0xac, 0x22, 
+    0x15, 0xf9, 0xd9, 0xa4, 0xc9, 0x6a, 0xf6, 0x9f, 0xb3, 0x62, 
+    0xd4, 0x57, 0xb1, 0x6a, 0xa9, 0xc5, 0xb1, 0x6a, 0xb9, 0xc5, 
+    0x4e, 0x95, 0x46, 0x3a, 0x82, 0xa5, 0x81, 0x62, 0x47, 0xa5, 
+    0xa6, 0xca, 0x90, 0x67, 0xee, 0x22, 0x57, 0x4d, 0x11, 0x48, 
+    0xe9, 0x55, 0x2d, 0xf7, 0x7a, 0x17, 0x29, 0x8e, 0x7a, 0x55, 
+    0x29, 0xce, 0x85, 0xaa, 0xd6, 0x31, 0x01, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x1c, 0xea, 
+    0x42, 0x9b, 0x87, 0x51, 0xd4, 0xb1, 0x0d, 0xa3, 0x13, 0x96, 
+    0xed, 0xf5, 0x90, 0xce, 0x49, 0x6b, 0xd9, 0xb8, 0x19, 0x62, 
+    0x70, 0xb2, 0x19, 0x63, 0x50, 0xb2, 0xe6, 0x9c, 0xaf, 0x4d, 
+    0x25, 0x1a, 0x7e, 0x0d, 0x88, 0x4b, 0x2e, 0xb3, 0xea, 0x33, 
+    0x81, 0xe3, 0xb2, 0x90, 0xcb, 0x97, 0xd5, 0x33, 0x13, 0x84, 
+    0xb4, 0x03, 0x11, 0x31, 0xb4, 0x03, 0x11, 0x10, 0x4b, 0xfc, 
+    0xee, 0xef, 0x5a, 0x56, 0xd7, 0x2a, 0xc2, 0x83, 0x8a, 0x07, 
+    0xa1, 0x6f, 0xf9, 0x84, 0x26, 0x21, 0x67, 0xa5, 0xe2, 0x8a, 
+    0xc6, 0xb3, 0x52, 0x0e, 0xce, 0x21, 0x52, 0x0e, 0xce, 0x21, 
+    0xad, 0xf1, 0x31, 0xde, 0x28, 0x6a, 0x17, 0x5b, 0x92, 0xd0, 
+    0xf5, 0x33, 0xfa, 0x22, 0x54, 0xf7, 0x82, 0xc3, 0x71, 0x73, 
+    0x80, 0x48, 0x1f, 0x7a, 0x9c, 0x70, 0x74, 0xba, 0x98, 0x70, 
+    0x55, 0xba, 0x67, 0x8f, 0xaa, 0x45, 0x28, 0x9f, 0xb7, 0xcc, 
+    0x72, 0x83, 0x2a, 0xe7, 0x9c, 0xa6, 0xa3, 0x8f, 0x38, 0xe7, 
+    0x6d, 0x7d, 0x3d, 0xb0, 0x83, 0xfb, 0x23, 0xf1, 0x82, 0xff, 
+    0x2b, 0xb1, 0x83, 0xff, 0xd4, 0x4e, 0x7c, 0x00, 0x7c, 0x79, 
+    0x36, 0xcc, 0x2d, 0xf5, 0xae, 0x00, 0x8d, 0x7b, 0xa7, 0xcf, 
+    0xb2, 0xde, 0xa2, 0x7c, 0xf2, 0xd5, 0x95, 0x0d, 0x94, 0xb0, 
+    0x75, 0x26, 0x96, 0xf0, 0x35, 0x24, 0x69, 0x0f, 0xca, 0xdb, 
+    0x20, 0x9d, 0x8f, 0x24, 0xc9, 0x97, 0x2b, 0x1e, 0xf5, 0xaf, 
+    0x7a, 0x22, 0xa6, 0xa7, 0x1e, 0x76, 0xf4, 0x95, 0x7e, 0x7a, 
+    0x3d, 0x97, 0x3e, 0xda, 0xbd, 0x97, 0x3e, 0xfa, 0x42, 0x68, 
+    0xc1, 0x05, 0x78, 0xe2, 0x57, 0x31, 0xcd, 0xe2, 0x53, 0x25, 
+    0x2a, 0x66, 0x68, 0x7d, 0x3d, 0x09, 0xcf, 0x8a, 0x29, 0xb7, 
+    0xe3, 0x4c, 0x0b, 0x77, 0x09, 0x01, 0x09, 0x77, 0x49, 0x01, 
+    0xf6, 0x88, 0xb6, 0xfe, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x98, 0xac, 0x21, 0x34, 
+    0xf3, 0xbb, 0x7f, 0x8f, 0x29, 0x3a, 0x0f, 0x0a, 0xb4, 0xb8, 
+    0x5d, 0x00, 0x89, 0x59, 0xd7, 0x43, 0x0b, 0x3a, 0x64, 0x19, 
+    0x09, 0x3a, 0x65, 0x08, 0xf6, 0xc5, 0x9a, 0xf7, 0xb2, 0x0d, 
+    0xb7, 0xbd, 0xbc, 0x0d, 0x50, 0xca, 0xd4, 0x1d, 0x93, 0x44, 
+    0xc0, 0xa0, 0x14, 0x0f, 0xfe, 0x7a, 0x3c, 0x1b, 0x37, 0xba, 
+    0x61, 0x77, 0x37, 0xba, 0x70, 0x7f, 0xc8, 0x45, 0x8f, 0x80, 
+    0x2e, 0x33, 0xd8, 0x88, 0x61, 0xa4, 0xce, 0x30, 0xe1, 0xea, 
+    0x8c, 0x6f, 0xf2, 0x1f, 0x2d, 0xaa, 0xb0, 0x7b, 0x9c, 0x43, 
+    0xd1, 0xff, 0xed, 0x17, 0xd0, 0xff, 0xad, 0x13, 0x2f, 0x00, 
+    0x52, 0xec, 0x9c, 0xda, 0x41, 0x4d, 0xaa, 0x56, 0xab, 0x95, 
+    0x5e, 0x54, 0xb2, 0xae, 0x30, 0xcd, 0x2c, 0x5b, 0x12, 0x29, 
+    0xe0, 0xe4, 0x02, 0x41, 0xbb, 0x65, 0x02, 0x49, 0xb9, 0x64, 
+    0xfd, 0xb6, 0x46, 0x9b, 0x35, 0xf0, 0xb1, 0x56, 0xa9, 0x8e, 
+    0x45, 0xa5, 0xe7, 0x9e, 0xf1, 0xc1, 0x65, 0xa3, 0x38, 0x96, 
+    0x73, 0x29, 0xa8, 0x51, 0x54, 0x77, 0x31, 0x4a, 0x65, 0x77, 
+    0xb1, 0x43, 0x9a, 0x88, 0x4e, 0xbc, 0x32, 0x24, 0x99, 0x19, 
+    0xd4, 0x08, 0x13, 0x2f, 0xbf, 0x5b, 0x34, 0xfb, 0x2f, 0x93, 
+    0xd1, 0x9b, 0xdc, 0x70, 0x6e, 0x19, 0x56, 0x53, 0x44, 0xf3, 
+    0x5e, 0x53, 0x44, 0x7b, 0xa1, 0xac, 0xbb, 0x84, 0x08, 0xd9, 
+    0x8e, 0x10, 0xb1, 0x88, 0x4e, 0x33, 0x86, 0xdd, 0xc1, 0xbc, 
+    0x0a, 0x83, 0x84, 0x68, 0x75, 0x6f, 0x67, 0x77, 0x44, 0x85, 
+    0x07, 0x76, 0x44, 0x8f, 0x07, 0x77, 0xbb, 0x70, 0xf8, 0x88, 
+    0x83, 0x18, 0x11, 0xdb, 0xf2, 0xea, 0x5e, 0x78, 0x9d, 0xb4, 
+    0xa7, 0xd0, 0x77, 0x32, 0xfd, 0xb2, 0x90, 0xad, 0xb7, 0xdf, 
+    0x30, 0x02, 0xff, 0xd0, 0x20, 0x00, 0xff, 0xd0, 0xcf, 0xff, 
+    0x00, 0x2f, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x80, 0xde, 0x9b, 0x24, 0x79, 0xde, 
+    0x52, 0x83, 0xf4, 0x1f, 0x16, 0x0a, 0xa5, 0xf0, 0x7a, 0xfc, 
+    0xd9, 0x0d, 0x5f, 0x0d, 0xf9, 0x9a, 0x33, 0x4b, 0xf9, 0x9a, 
+    0x33, 0x09, 0x06, 0x65, 0xcc, 0xf6, 0x35, 0x08, 0x23, 0xaa, 
+    0x06, 0xb1, 0x5b, 0x50, 0xdf, 0xa4, 0xa3, 0xcd, 0xc0, 0x08, 
+    0xe6, 0x85, 0xcd, 0xb1, 0x2d, 0x1a, 0xdf, 0x32, 0xe6, 0x11, 
+    0xcf, 0x30, 0xa7, 0x11, 0x30, 0xcf, 0x58, 0xee, 0x4c, 0xad, 
+    0x9a, 0x46, 0xe1, 0x56, 0x69, 0x88, 0x9a, 0xbe, 0xe3, 0x93, 
+    0xa5, 0x8d, 0x59, 0x33, 0x0a, 0x86, 0x13, 0x46, 0x08, 0xa6, 
+    0x93, 0x86, 0x08, 0xa6, 0x93, 0x06, 0xf7, 0x59, 0x6c, 0xf9, 
+    0xc0, 0x96, 0x57, 0x1e, 0xef, 0xc6, 0x33, 0x25, 0x5d, 0x30, 
+    0x18, 0x77, 0x04, 0x7f, 0x63, 0x76, 0xc3, 0x81, 0x5d, 0x55, 
+    0xdf, 0xe5, 0xda, 0x74, 0xd7, 0xe5, 0xda, 0x74, 0x28, 0x1a, 
+    0x25, 0x8b, 0x58, 0xa9, 0xfa, 0x1c, 0xd9, 0xb2, 0xaa, 0xa8, 
+    0x52, 0x74, 0xba, 0x35, 0x97, 0xc3, 0x11, 0x61, 0x0d, 0x89, 
+    0x5b, 0x70, 0x11, 0xed, 0x57, 0x73, 0x15, 0xed, 0x53, 0x71, 
+    0xea, 0x12, 0xac, 0x8e, 0xaf, 0x5b, 0xae, 0xc5, 0x6d, 0x5b, 
+    0x68, 0xc1, 0x77, 0x6d, 0xba, 0x91, 0xfb, 0xd0, 0x12, 0x23, 
+    0xbe, 0x02, 0x12, 0xfc, 0x2a, 0xb2, 0x88, 0xfc, 0xaa, 0x92, 
+    0x98, 0xfc, 0x55, 0x6d, 0x67, 0x03, 0xaf, 0x1b, 0xbf, 0x3e, 
+    0xe4, 0x3c, 0x17, 0x8b, 0x4d, 0x1f, 0x23, 0xa6, 0x59, 0x00, 
+    0x3b, 0x4e, 0x54, 0x67, 0xe3, 0xc9, 0x5d, 0x46, 0x01, 0x4b, 
+    0x5d, 0x44, 0x21, 0x4b, 0xa2, 0xbb, 0xde, 0xb4, 0xcf, 0x48, 
+    0x2a, 0x76, 0x2d, 0xb7, 0x31, 0x7f, 0xc7, 0x35, 0xb1, 0x88, 
+    0x94, 0x5f, 0xa7, 0xe9, 0xde, 0xb7, 0xbf, 0x5a, 0x16, 0x37, 
+    0xb7, 0xdd, 0x96, 0x37, 0xb7, 0xd9, 0x69, 0xc8, 0x48, 0x26, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x35, 0x0b, 0x81, 0x6e, 0x7b, 0xa4, 0x51, 0x57, 
+    0x5a, 0xb1, 0xff, 0xd2, 0x53, 0x0a, 0xdc, 0xfc, 0x94, 0x89, 
+    0x24, 0x1a, 0x00, 0x9e, 0x78, 0x3b, 0x10, 0x9b, 0x38, 0x3b, 
+    0xef, 0x64, 0xc7, 0xc4, 0xa4, 0x4f, 0xee, 0x1a, 0x58, 0xb4, 
+    0x30, 0x49, 0x03, 0xc1, 0x50, 0xdc, 0xb1, 0xec, 0x5b, 0x94, 
+    0x83, 0x06, 0x8e, 0x2c, 0x43, 0x84, 0xd9, 0xac, 0x03, 0x84, 
+    0xda, 0xac, 0xfc, 0x7b, 0x25, 0x53, 0x0c, 0x64, 0x40, 0xb9, 
+    0x49, 0x48, 0x41, 0x1d, 0xca, 0x4a, 0x67, 0x57, 0x1c, 0x2e, 
+    0x2b, 0x2d, 0xb4, 0x15, 0x0b, 0xb5, 0x7c, 0x80, 0x96, 0x9d, 
+    0x7c, 0x80, 0x9a, 0x9d, 0x83, 0x7f, 0x65, 0x62, 0x57, 0x29, 
+    0x49, 0x40, 0xbf, 0x1a, 0xbc, 0xa9, 0xbb, 0xa4, 0xb0, 0xa9, 
+    0xe1, 0x39, 0x3b, 0xd6, 0xc7, 0x04, 0x4f, 0xc1, 0x41, 0x2e, 
+    0x2f, 0x85, 0x41, 0x2c, 0x2f, 0xc5, 0xbe, 0xd3, 0xd0, 0x3a, 
+    0xbd, 0x99, 0x11, 0x63, 0xd0, 0xbd, 0x9f, 0xf4, 0xda, 0x71, 
+    0xf1, 0xbd, 0xae, 0xc2, 0x1f, 0xf7, 0x4e, 0x12, 0x29, 0x54, 
+    0x20, 0x52, 0x79, 0xde, 0x28, 0x52, 0x79, 0xd6, 0xd7, 0xad, 
+    0x86, 0x29, 0x3d, 0xe0, 0xe2, 0x58, 0xe4, 0xb3, 0x40, 0x94, 
+    0x24, 0xe9, 0x10, 0xea, 0x70, 0x36, 0x37, 0x84, 0x37, 0xf6, 
+    0xd6, 0x32, 0x35, 0xfc, 0x0a, 0xba, 0x35, 0xfc, 0x1a, 0xba, 
+    0xca, 0x03, 0xe5, 0x45, 0x5c, 0xd7, 0x39, 0x71, 0x09, 0x3b, 
+    0x31, 0x93, 0x18, 0x99, 0x9b, 0x43, 0x33, 0x39, 0x62, 0x7e, 
+    0xd9, 0x02, 0x22, 0xed, 0xd9, 0x08, 0x2a, 0xfe, 0xd9, 0x08, 
+    0x22, 0xff, 0x26, 0xf7, 0xdd, 0x00, 0xf8, 0x80, 0xe6, 0x7a, 
+    0x12, 0xe5, 0x3b, 0x51, 0x04, 0xf9, 0xcc, 0x3f, 0xcb, 0x42, 
+    0xeb, 0x28, 0xae, 0xee, 0x02, 0xd5, 0x0d, 0x18, 0x27, 0x15, 
+    0x0f, 0x18, 0x26, 0x15, 0xf0, 0xe7, 0xd9, 0xea, 0x01, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x4b, 0x1d, 0x31, 0xeb, 0x9e, 0xea, 0x2f, 0x38, 0x01, 0x30, 
+    0xbe, 0x26, 0x94, 0x22, 0x25, 0x73, 0x29, 0x04, 0xe7, 0x8a, 
+    0x85, 0x80, 0x77, 0xfe, 0x84, 0x80, 0x77, 0xfe, 0x7b, 0x7f, 
+    0x88, 0x01, 0x9d, 0x4e, 0xd9, 0xec, 0xa2, 0xd8, 0xd5, 0xa9, 
+    0x8d, 0x42, 0x0f, 0x5e, 0x6f, 0xc6, 0xee, 0xfe, 0xca, 0x2c, 
+    0xd6, 0x5f, 0x66, 0xcd, 0xcf, 0xf2, 0x66, 0x8d, 0xcf, 0xde, 
+    0x99, 0x72, 0x30, 0x21, 0xc2, 0xfb, 0xea, 0x85, 0xb3, 0x92, 
+    0xf2, 0xf9, 0xa3, 0xc8, 0x45, 0xd3, 0x7d, 0x19, 0xf6, 0xcd, 
+    0xd4, 0xc3, 0x45, 0xe7, 0x54, 0xe3, 0x47, 0xdf, 0x54, 0xc3, 
+    0x47, 0xdf, 0xab, 0x3c, 0xb8, 0x20, 0x21, 0x36, 0xfa, 0xa6, 
+    0x35, 0x15, 0x0e, 0xf0, 0x28, 0x44, 0xba, 0x93, 0x70, 0xc8, 
+    0x9e, 0xaf, 0x3c, 0x25, 0x31, 0x0c, 0xdc, 0x55, 0xf0, 0xee, 
+    0x1c, 0x45, 0xf0, 0xee, 0xe3, 0xba, 0x0f, 0x11, 0xe0, 0x28, 
+    0xf4, 0x7e, 0xe5, 0x04, 0xa8, 0xe5, 0x5e, 0x48, 0x1b, 0xc0, 
+    0x8e, 0x17, 0x05, 0xa3, 0x2d, 0xd0, 0x3c, 0xf0, 0x6d, 0x89, 
+    0xb4, 0xe0, 0x6d, 0x88, 0xb4, 0xe0, 0x92, 0x77, 0x4b, 0x1f, 
+    0xc2, 0x0f, 0x82, 0x73, 0xb7, 0x27, 0x88, 0x2e, 0x27, 0x3e, 
+    0x89, 0xfb, 0x61, 0x37, 0xdd, 0x66, 0xb7, 0x19, 0x1e, 0x36, 
+    0x03, 0x59, 0x1c, 0xee, 0x23, 0x19, 0x1c, 0xa6, 0xdc, 0xe6, 
+    0xe3, 0x59, 0xa0, 0x0c, 0x00, 0xd8, 0x55, 0xd9, 0x56, 0x47, 
+    0x6e, 0x8b, 0xc9, 0xb5, 0xd7, 0x39, 0xfb, 0x1a, 0x4d, 0xbc, 
+    0xe5, 0x65, 0x4c, 0x53, 0xc5, 0x60, 0x4c, 0x19, 0xe5, 0x60, 
+    0xb3, 0xe6, 0x1a, 0x9f, 0xf7, 0x25, 0x77, 0x24, 0x51, 0x25, 
+    0x46, 0x11, 0x38, 0x62, 0xc7, 0x6b, 0x52, 0x9a, 0x56, 0x9b, 
+    0x22, 0x56, 0xfa, 0xd4, 0xaa, 0xb2, 0x53, 0x0d, 0x2a, 0xb2, 
+    0x53, 0x0d, 0xd5, 0x4d, 0xac, 0xf2, 0x00, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x0e, 0x64, 
+    0x3e, 0x18, 0x3b, 0xcb, 0xeb, 0xf2, 0x02, 0x44, 0x7b, 0x05, 
+    0x15, 0x20, 0x2e, 0x45, 0x8f, 0x10, 0x33, 0xb6, 0x83, 0xe0, 
+    0xb2, 0x9d, 0x83, 0xe0, 0xb2, 0x9d, 0x7c, 0x1f, 0x4d, 0x62, 
+    0x8e, 0x8a, 0x04, 0x60, 0x17, 0x58, 0xc2, 0x72, 0x0f, 0x5a, 
+    0x00, 0x4e, 0xcc, 0xc9, 0x4f, 0xbf, 0xda, 0x59, 0x02, 0x39, 
+    0x4b, 0xd5, 0x76, 0x29, 0x4b, 0xd1, 0x56, 0x29, 0xb4, 0x2e, 
+    0xa9, 0xd6, 0x9f, 0x51, 0x50, 0xd6, 0xf7, 0x24, 0x06, 0xa5, 
+    0x0f, 0x2a, 0x25, 0xe3, 0x7b, 0xf1, 0x23, 0xb6, 0x2e, 0xa1, 
+    0xf2, 0xeb, 0xbe, 0xed, 0x60, 0xcf, 0xbe, 0xed, 0x60, 0x8f, 
+    0x41, 0x12, 0x9f, 0x70, 0x4a, 0xb0, 0x50, 0x89, 0xe3, 0xe3, 
+    0xd0, 0x22, 0x7e, 0x8f, 0x69, 0x3e, 0x35, 0x72, 0x52, 0xf7, 
+    0x35, 0x03, 0xed, 0xbf, 0xb5, 0x02, 0xcc, 0xbf, 0x35, 0x02, 
+    0xcc, 0xbf, 0xca, 0xfd, 0x33, 0x40, 0xb2, 0x0d, 0x23, 0xcc, 
+    0xf6, 0xcf, 0x89, 0xb3, 0x07, 0x63, 0x92, 0xad, 0x47, 0x3d, 
+    0x7d, 0xd3, 0xb3, 0x7e, 0xf0, 0x2c, 0xa7, 0x93, 0xe7, 0x98, 
+    0xa7, 0xbf, 0xf4, 0x89, 0x58, 0x40, 0x0b, 0x76, 0xda, 0x21, 
+    0x50, 0xb3, 0x08, 0x36, 0xcb, 0x8d, 0xd7, 0x74, 0xec, 0xaf, 
+    0x56, 0x85, 0x2c, 0xf6, 0x28, 0xab, 0xbf, 0xf8, 0xf4, 0x09, 
+    0x8b, 0xe2, 0xf4, 0x09, 0xaf, 0xe2, 0x0b, 0xf6, 0x50, 0x1d, 
+    0x56, 0x3e, 0xd4, 0x03, 0x57, 0x23, 0xf4, 0xa3, 0x10, 0xb0, 
+    0x8d, 0x06, 0x19, 0x69, 0x35, 0x6f, 0xfa, 0xc8, 0x37, 0x69, 
+    0x3f, 0x5d, 0x35, 0x6c, 0x3f, 0x88, 0x35, 0x6d, 0xc0, 0x77, 
+    0xca, 0x92, 0xc8, 0x3f, 0xce, 0x0c, 0xba, 0x1f, 0x44, 0xb7, 
+    0x68, 0xfb, 0xb8, 0x4a, 0xfa, 0x9a, 0x94, 0x72, 0x98, 0x40, 
+    0x89, 0xe2, 0x4f, 0x55, 0x98, 0xc2, 0xcf, 0x53, 0x98, 0xc2, 
+    0x30, 0xac, 0x67, 0x3d, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x0c, 0x43, 0x91, 0x3d, 
+    0xe4, 0x75, 0xe0, 0xb3, 0x42, 0x9e, 0x5c, 0x75, 0x28, 0x41, 
+    0x25, 0xfc, 0x3e, 0xa4, 0xc7, 0x67, 0x23, 0xc8, 0x44, 0x67, 
+    0x23, 0xc8, 0x44, 0x67, 0xdc, 0x37, 0xbb, 0x98, 0x18, 0xb7, 
+    0x3f, 0x2f, 0x0c, 0x34, 0x2e, 0xc8, 0x9f, 0x42, 0xa1, 0xc0, 
+    0x5b, 0x8a, 0xb0, 0x48, 0xc0, 0xbe, 0x45, 0x80, 0x4c, 0x3e, 
+    0xa0, 0x44, 0x4c, 0x3e, 0xa0, 0x40, 0xb3, 0xc1, 0x5f, 0xbf, 
+    0xb8, 0x25, 0xf4, 0xa7, 0x68, 0xb3, 0x3f, 0x24, 0x90, 0xf4, 
+    0xf8, 0x21, 0x7f, 0x86, 0x5d, 0xa6, 0xb4, 0xc9, 0x2b, 0x64, 
+    0xb9, 0xd0, 0xed, 0xac, 0xb9, 0xd0, 0xe9, 0x2c, 0x46, 0x2f, 
+    0x16, 0xd3, 0xd3, 0x0d, 0xa5, 0xf5, 0xb9, 0x05, 0xea, 0x11, 
+    0xfb, 0x92, 0xf6, 0x73, 0xc8, 0xea, 0xcd, 0xe8, 0x0e, 0xe0, 
+    0x65, 0x8f, 0xfc, 0x61, 0x14, 0xcc, 0xfc, 0x61, 0x54, 0xcc, 
+    0x03, 0x9e, 0xab, 0x33, 0xda, 0x88, 0xe9, 0xf2, 0xfd, 0xc8, 
+    0x19, 0x34, 0xde, 0x55, 0x52, 0x7b, 0xe8, 0xbe, 0xd3, 0xf8, 
+    0xa4, 0x55, 0x24, 0xc4, 0xe9, 0x85, 0xe7, 0xf9, 0xe9, 0x05, 
+    0x67, 0xfd, 0x16, 0xfa, 0x98, 0x02, 0xdd, 0xa3, 0x23, 0xec, 
+    0xd5, 0xf7, 0x75, 0x13, 0x5b, 0x8d, 0x54, 0x02, 0x6c, 0xc2, 
+    0x1e, 0x59, 0xc0, 0xfb, 0x5b, 0xc6, 0xf9, 0x5a, 0x77, 0xad, 
+    0xf8, 0xda, 0x7f, 0xe9, 0x07, 0x25, 0x80, 0x16, 0xaa, 0x2f, 
+    0xe7, 0xf3, 0x3b, 0x7d, 0x9d, 0x37, 0x97, 0x81, 0xa0, 0xa8, 
+    0x10, 0x43, 0xe4, 0x3a, 0x4f, 0xed, 0x38, 0xf2, 0x87, 0xe5, 
+    0x27, 0x36, 0x87, 0xe5, 0x25, 0x32, 0x78, 0x1a, 0xda, 0xcd, 
+    0x5d, 0x74, 0xe6, 0xf6, 0x2a, 0x7b, 0xd2, 0x57, 0x76, 0x48, 
+    0xe5, 0x26, 0x6b, 0x86, 0x70, 0x86, 0x56, 0xe3, 0x6a, 0xd7, 
+    0x12, 0x60, 0xce, 0x26, 0x16, 0x60, 0xea, 0x26, 0xe9, 0x9f, 
+    0x15, 0xd9, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x39, 0xda, 0x11, 0x7f, 0x2c, 0x42, 
+    0xe4, 0x4b, 0xb8, 0xe0, 0x90, 0xd0, 0xbb, 0x18, 0x04, 0xc2, 
+    0x7b, 0x8f, 0x86, 0x4d, 0xa5, 0x8d, 0xa6, 0x50, 0xa9, 0x8d, 
+    0x86, 0x40, 0x56, 0x72, 0x79, 0xbf, 0xe5, 0x18, 0x44, 0x8b, 
+    0x1d, 0x18, 0xd4, 0x81, 0xa8, 0xed, 0xc3, 0x76, 0x96, 0x3b, 
+    0xd8, 0x31, 0x74, 0x15, 0x2c, 0x58, 0x46, 0xf5, 0x3e, 0x55, 
+    0x46, 0x55, 0x3c, 0x55, 0xb9, 0xaa, 0xc3, 0xaa, 0xcf, 0xa1, 
+    0xf7, 0x76, 0x8d, 0x9c, 0xef, 0x9a, 0xde, 0x2b, 0x7a, 0x49, 
+    0xc1, 0xc1, 0x38, 0xc8, 0x63, 0x45, 0x42, 0xeb, 0x67, 0x8d, 
+    0xb7, 0x79, 0x67, 0x85, 0x32, 0x79, 0x98, 0x7a, 0xcd, 0x86, 
+    0x05, 0xc0, 0xc6, 0xa4, 0xd0, 0xb0, 0x70, 0xd0, 0xed, 0x54, 
+    0xea, 0x94, 0xf4, 0xaa, 0x1d, 0x75, 0x91, 0x86, 0x3e, 0x6e, 
+    0x1c, 0xeb, 0x0e, 0x6e, 0x94, 0xeb, 0x2e, 0x6e, 0x6b, 0x14, 
+    0xd1, 0x91, 0x6a, 0xde, 0x6a, 0xcb, 0x82, 0x57, 0xf1, 0x95, 
+    0x49, 0xff, 0xe4, 0xf4, 0xd4, 0x20, 0xcc, 0xbd, 0x2c, 0xa4, 
+    0x80, 0x77, 0x05, 0x0c, 0xdb, 0xbe, 0x05, 0x0c, 0xd9, 0x3e, 
+    0xfa, 0xf3, 0x26, 0xc1, 0x9f, 0x92, 0x00, 0x8b, 0x5c, 0xeb, 
+    0x6a, 0xd2, 0x90, 0xab, 0x56, 0x2a, 0xf7, 0xc0, 0x64, 0x11, 
+    0xc6, 0xbc, 0x97, 0x60, 0x90, 0xbc, 0x54, 0x40, 0x90, 0xbc, 
+    0x54, 0x40, 0x6f, 0x43, 0xab, 0xbf, 0x8a, 0xc4, 0xce, 0x3f, 
+    0xae, 0x5b, 0xc5, 0x96, 0x94, 0xa0, 0x67, 0xa2, 0xc7, 0x75, 
+    0x8d, 0x84, 0x30, 0xdb, 0xa1, 0x8b, 0x4e, 0xfb, 0xe1, 0xd8, 
+    0x4e, 0xfb, 0xa1, 0x88, 0xb1, 0x04, 0x5e, 0x77, 0x8a, 0x55, 
+    0xbf, 0xfb, 0x04, 0xfe, 0x21, 0x99, 0x89, 0xbe, 0x4a, 0x99, 
+    0xb5, 0x42, 0x06, 0x84, 0xbc, 0xe0, 0xf0, 0xfc, 0x85, 0x00, 
+    0x05, 0xc3, 0x85, 0x00, 0x04, 0xc3, 0x7a, 0xff, 0xfb, 0x3c, 
+    0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x22, 0xda, 0x8d, 0x48, 0xe1, 0x99, 0x28, 0x49, 
+    0x90, 0x69, 0x60, 0xbe, 0xc5, 0x88, 0x99, 0x9a, 0x1f, 0x3a, 
+    0x78, 0xf0, 0xcb, 0x31, 0x68, 0xc3, 0xcb, 0x31, 0x68, 0xc1, 
+    0x34, 0xce, 0x97, 0x3e, 0x34, 0xff, 0xe7, 0x3a, 0xe6, 0xe1, 
+    0xff, 0xe7, 0xaf, 0x75, 0x3d, 0xb9, 0x2a, 0x06, 0x5b, 0xae, 
+    0x87, 0x59, 0x16, 0xd3, 0xc3, 0x4c, 0x37, 0x53, 0xc3, 0x4c, 
+    0x37, 0xd3, 0x3c, 0xb3, 0xc8, 0x2c, 0x08, 0xdd, 0x74, 0x06, 
+    0x33, 0xf9, 0x61, 0xcc, 0x7c, 0x79, 0xfd, 0x3a, 0x6d, 0x4a, 
+    0x40, 0xb5, 0xea, 0xf0, 0x78, 0x98, 0xec, 0xda, 0x3d, 0xaa, 
+    0xec, 0xda, 0x3d, 0xaa, 0x13, 0x25, 0xc2, 0x55, 0xf2, 0x87, 
+    0x14, 0x8b, 0xb6, 0x29, 0x44, 0xe1, 0xb1, 0x10, 0x9b, 0x13, 
+    0x2f, 0xae, 0xb6, 0x63, 0x8f, 0x8c, 0xce, 0xab, 0x4e, 0x0f, 
+    0x12, 0xab, 0x4f, 0x0c, 0x92, 0xab, 0xb0, 0xf3, 0x6d, 0x54, 
+    0x55, 0x95, 0x9b, 0x2a, 0xf5, 0x17, 0xd0, 0x79, 0x5c, 0xc5, 
+    0x3a, 0xbe, 0xae, 0xf6, 0xb2, 0x68, 0x0a, 0x20, 0xc8, 0xd6, 
+    0x08, 0xc2, 0xcb, 0x68, 0x0a, 0xc0, 0xcb, 0x68, 0xf5, 0x3f, 
+    0x34, 0x97, 0x4c, 0x52, 0x17, 0x9b, 0x91, 0xcb, 0x84, 0x19, 
+    0x7e, 0x61, 0x1a, 0x45, 0x0b, 0x1f, 0x6c, 0x8a, 0xb4, 0x03, 
+    0xb9, 0x8f, 0x95, 0xa3, 0xba, 0x8d, 0x95, 0x23, 0xba, 0x8d, 
+    0x6a, 0xdc, 0x45, 0x72, 0x0d, 0xf0, 0x5b, 0x74, 0xe2, 0x5a, 
+    0x5b, 0xd0, 0x7b, 0x9b, 0x64, 0xe8, 0xb5, 0xbf, 0x72, 0x25, 
+    0xba, 0xd7, 0x2a, 0x2c, 0xbf, 0xa5, 0xc1, 0xfc, 0xbf, 0xa7, 
+    0x40, 0x7c, 0x40, 0x58, 0xbf, 0x83, 0x00, 0xbc, 0xc0, 0xc5, 
+    0xfc, 0xcc, 0x87, 0xf6, 0x82, 0xdb, 0x7f, 0x04, 0x38, 0xbe, 
+    0x87, 0xee, 0xae, 0x59, 0xc0, 0x5a, 0x3c, 0x5d, 0xc8, 0x56, 
+    0x3c, 0x5d, 0xc8, 0x5e, 0xc3, 0xa2, 0x37, 0xa1, 0x00, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0xc1, 0x7c, 0xa3, 0x86, 0x61, 0x37, 0x8d, 0xb0, 0xf0, 0xb9, 
+    0x6b, 0x2c, 0x31, 0xfa, 0xdf, 0xe6, 0x75, 0xbd, 0x0e, 0x31, 
+    0xb9, 0xb9, 0x9b, 0x20, 0xf1, 0xb9, 0x9f, 0x20, 0x0e, 0x46, 
+    0x60, 0xdf, 0xd2, 0x5e, 0xd3, 0xa5, 0xd5, 0xf2, 0xc8, 0xef, 
+    0xb3, 0x18, 0x09, 0x91, 0xaf, 0x4b, 0x40, 0x3b, 0x7a, 0xd1, 
+    0x73, 0xfb, 0x6e, 0x19, 0xf5, 0xe1, 0x7e, 0x19, 0xf5, 0xe1, 
+    0x81, 0xe6, 0x0a, 0x1e, 0x2a, 0x1e, 0x4e, 0x81, 0x9c, 0xf4, 
+    0x27, 0x36, 0x19, 0xba, 0xea, 0x28, 0x92, 0x1e, 0x34, 0x6b, 
+    0xfb, 0x5b, 0x58, 0x2e, 0x62, 0xaa, 0x3a, 0x34, 0x62, 0xaa, 
+    0x38, 0x34, 0x9d, 0x55, 0xc7, 0xcb, 0x89, 0x87, 0xa7, 0x3b, 
+    0xd6, 0x28, 0x54, 0xee, 0x3d, 0x26, 0x3d, 0x67, 0x47, 0x88, 
+    0xe6, 0x9a, 0xf4, 0x5a, 0x1a, 0x10, 0x27, 0xc0, 0x16, 0x9a, 
+    0x27, 0x48, 0x16, 0x9a, 0xd8, 0xb7, 0xe9, 0x65, 0x82, 0x75, 
+    0x6c, 0xad, 0xd8, 0xfe, 0x32, 0x42, 0x39, 0x26, 0x17, 0x9b, 
+    0x35, 0xf4, 0x6d, 0x26, 0x45, 0x9f, 0x6b, 0x5c, 0x50, 0x1d, 
+    0x22, 0x55, 0x50, 0x1d, 0x62, 0x45, 0xaf, 0xe2, 0x9d, 0xba, 
+    0x1d, 0x4e, 0xf8, 0xcb, 0x37, 0x49, 0x44, 0xcd, 0x29, 0xe3, 
+    0xf8, 0xfd, 0xa3, 0x23, 0x64, 0x27, 0x05, 0x49, 0xee, 0x3e, 
+    0xa5, 0x26, 0xec, 0x5f, 0xa5, 0x26, 0xec, 0x5f, 0x5a, 0xd9, 
+    0x13, 0xa0, 0xcb, 0xf5, 0x5a, 0xd2, 0x63, 0xec, 0xfb, 0xba, 
+    0x79, 0x40, 0x19, 0x76, 0xcc, 0xed, 0x49, 0x94, 0x2f, 0x43, 
+    0x01, 0x61, 0xa5, 0x45, 0x27, 0x76, 0xa5, 0x45, 0x07, 0x72, 
+    0x5a, 0xba, 0xf8, 0x8d, 0xdc, 0x78, 0x48, 0xb6, 0x79, 0xf7, 
+    0x3f, 0x87, 0x1c, 0xb1, 0x2c, 0xc6, 0xef, 0x6c, 0x62, 0x37, 
+    0x05, 0x08, 0xe7, 0xcc, 0xe5, 0x08, 0xae, 0x81, 0xc5, 0x08, 
+    0xae, 0x84, 0x3a, 0xf7, 0x51, 0x7b, 0x00, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x6a, 0x83, 
+    0xce, 0x4a, 0xa5, 0x2a, 0xf8, 0xbb, 0x6d, 0x45, 0x91, 0x4d, 
+    0x62, 0x8a, 0xd2, 0xb0, 0x02, 0xbc, 0x85, 0xc4, 0x53, 0x0a, 
+    0x16, 0x59, 0x53, 0x08, 0x96, 0x59, 0xac, 0xf7, 0x69, 0xa6, 
+    0xe9, 0x7e, 0xd1, 0x6c, 0x31, 0x53, 0x3f, 0xf9, 0xdd, 0x63, 
+    0x8c, 0x31, 0x2f, 0xb4, 0xaf, 0xe2, 0x85, 0x43, 0x81, 0x3e, 
+    0xd5, 0x66, 0x29, 0x6a, 0xd5, 0x66, 0x29, 0x6a, 0x2a, 0x99, 
+    0xd6, 0x95, 0x01, 0xea, 0x7a, 0x41, 0x90, 0x79, 0xdd, 0x28, 
+    0x88, 0x12, 0x58, 0x28, 0x7f, 0x3c, 0x30, 0x76, 0xdd, 0x22, 
+    0x71, 0x46, 0x5a, 0x22, 0xb3, 0x74, 0x5e, 0x22, 0x31, 0x64, 
+    0xa1, 0xdd, 0xce, 0x9b, 0xc9, 0x34, 0x49, 0xe9, 0x23, 0x17, 
+    0xa7, 0xe5, 0xe2, 0x03, 0x60, 0xb3, 0x65, 0x24, 0x24, 0xf8, 
+    0x79, 0xeb, 0xa9, 0xf0, 0x08, 0xe3, 0x05, 0xfa, 0x48, 0xe3, 
+    0x05, 0xf8, 0xb7, 0x1c, 0xfa, 0x07, 0xb7, 0x5f, 0x5f, 0x26, 
+    0x52, 0xde, 0x36, 0x18, 0x87, 0x3d, 0x4d, 0xc4, 0x84, 0x61, 
+    0x2e, 0x5b, 0xfa, 0x02, 0x43, 0xae, 0x1e, 0xa7, 0x0a, 0xa6, 
+    0x9e, 0xa7, 0x0a, 0xa4, 0x61, 0x58, 0xf5, 0x5b, 0xce, 0x40, 
+    0xf0, 0xe3, 0x30, 0xb1, 0x0d, 0xca, 0xdc, 0x63, 0x69, 0x1a, 
+    0xf1, 0x3c, 0x88, 0x5a, 0xd1, 0xd1, 0xde, 0x33, 0xa0, 0x71, 
+    0xac, 0x82, 0xb0, 0x71, 0x8c, 0x82, 0x4f, 0x8e, 0x73, 0x7d, 
+    0x15, 0x7e, 0x14, 0x61, 0x32, 0x62, 0x14, 0x08, 0x5b, 0x91, 
+    0x64, 0x3b, 0xdc, 0x7c, 0x64, 0x6c, 0x09, 0x92, 0xf0, 0xda, 
+    0x2d, 0x56, 0x64, 0xfc, 0x0d, 0x56, 0x66, 0xfc, 0xf2, 0xa9, 
+    0x9b, 0x03, 0x1e, 0x93, 0xf3, 0x4c, 0xd4, 0x2c, 0x98, 0xa2, 
+    0x8f, 0xfd, 0x9a, 0xd3, 0xc8, 0xe7, 0x9f, 0x9e, 0xb1, 0x14, 
+    0x6b, 0x98, 0x21, 0x14, 0x0b, 0x08, 0x31, 0x14, 0x0b, 0x08, 
+    0xce, 0xeb, 0xf4, 0xf7, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0xb2, 0x73, 0xc5, 0x06, 
+    0xe6, 0xce, 0xd2, 0xe7, 0xd5, 0x7d, 0xf1, 0x55, 0xcc, 0xd9, 
+    0xf2, 0x91, 0xeb, 0x6e, 0x04, 0x02, 0x80, 0x64, 0xf3, 0x52, 
+    0xc0, 0x64, 0xf3, 0x12, 0x3f, 0x9b, 0x0c, 0xed, 0x03, 0x29, 
+    0xfb, 0x5d, 0xa6, 0x30, 0xd7, 0x96, 0xb2, 0xa5, 0xfa, 0x9d, 
+    0x80, 0x2c, 0xef, 0xed, 0xee, 0x93, 0x9c, 0x89, 0xa6, 0xf7, 
+    0x95, 0x9b, 0xa6, 0xb7, 0x95, 0x8b, 0x59, 0x48, 0x6a, 0x74, 
+    0x89, 0x28, 0x26, 0xa3, 0x73, 0xd3, 0xf4, 0xd6, 0x55, 0xce, 
+    0x4b, 0xcb, 0x6a, 0x2e, 0x64, 0x64, 0x16, 0x4e, 0xb1, 0xd0, 
+    0xb6, 0xed, 0xba, 0xdc, 0xb6, 0xee, 0xba, 0xdc, 0x49, 0x11, 
+    0x45, 0x23, 0x4c, 0x0c, 0x8e, 0x70, 0x70, 0x4a, 0x53, 0x5e, 
+    0x52, 0xeb, 0x38, 0xf2, 0xb9, 0x4f, 0x91, 0xee, 0x4a, 0xff, 
+    0x4a, 0xa8, 0x8c, 0xee, 0x80, 0xea, 0x0c, 0xee, 0x80, 0xea, 
+    0xf3, 0x11, 0x7f, 0x15, 0xbd, 0x66, 0xe5, 0x4a, 0xf7, 0x6d, 
+    0xa2, 0xbe, 0xfe, 0x87, 0xc1, 0xc9, 0x24, 0xa4, 0x7e, 0xb3, 
+    0x68, 0x06, 0xda, 0xad, 0xa0, 0x17, 0x52, 0xa9, 0xa4, 0x06, 
+    0x52, 0xa9, 0x5b, 0xf9, 0xad, 0x56, 0xad, 0xe7, 0x3e, 0x5e, 
+    0x0b, 0x9c, 0xcf, 0x74, 0xb4, 0x32, 0x8e, 0x88, 0xb6, 0x76, 
+    0x89, 0xf5, 0x85, 0x7f, 0x0a, 0x0f, 0x14, 0xb2, 0x29, 0x5f, 
+    0x14, 0xb6, 0x29, 0x5f, 0xeb, 0x49, 0xd6, 0xa0, 0x1f, 0x48, 
+    0x4b, 0xac, 0x49, 0x5b, 0xcb, 0x12, 0x6d, 0xc2, 0xda, 0x96, 
+    0x7c, 0xda, 0xfb, 0xbe, 0xd6, 0xf8, 0xaa, 0xfe, 0x53, 0x7a, 
+    0xe2, 0xfe, 0x5b, 0xfa, 0xea, 0xfe, 0xa4, 0x05, 0x15, 0x01, 
+    0x9a, 0xc6, 0xf6, 0x49, 0x9c, 0x36, 0x7c, 0xe7, 0xc6, 0x33, 
+    0x73, 0xd3, 0xf4, 0x03, 0xe3, 0x48, 0x56, 0x5d, 0x55, 0x1f, 
+    0x54, 0x13, 0x8f, 0xbe, 0x54, 0x13, 0x87, 0x9e, 0xab, 0xec, 
+    0x78, 0x61, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x40, 0x6a, 0xf8, 0x06, 0xc9, 0xa9, 
+    0xdd, 0x4b, 0x2f, 0x50, 0x9c, 0x1a, 0xeb, 0xb6, 0x24, 0xa9, 
+    0x73, 0xa2, 0xb0, 0x68, 0xbd, 0x82, 0xbc, 0x2c, 0xbf, 0x82, 
+    0xbc, 0x2c, 0x40, 0x7d, 0x43, 0xd3, 0xac, 0xfb, 0x7d, 0x7c, 
+    0x66, 0x6e, 0xe5, 0x0c, 0x97, 0xc8, 0x05, 0x2b, 0xcd, 0x02, 
+    0xdf, 0xf0, 0xcc, 0x43, 0x82, 0xbc, 0x89, 0x4a, 0xe7, 0xb1, 
+    0x8d, 0x4a, 0xc7, 0xb1, 0x72, 0xb5, 0x38, 0x4e, 0x08, 0x40, 
+    0x79, 0x74, 0x29, 0x9f, 0xbc, 0xb2, 0x4f, 0x81, 0x63, 0x49, 
+    0x1f, 0xb6, 0x70, 0x5b, 0x88, 0xe2, 0x47, 0x8e, 0xe8, 0xe0, 
+    0x04, 0xa2, 0xec, 0xe0, 0x04, 0xa2, 0x13, 0x1f, 0xfb, 0x5d, 
+    0x5e, 0xe2, 0x0d, 0x6e, 0x8f, 0x54, 0x00, 0x39, 0x79, 0x88, 
+    0x5f, 0x96, 0xa3, 0x30, 0xaa, 0x8b, 0x11, 0x56, 0xba, 0xf5, 
+    0x20, 0xe6, 0x3a, 0x4e, 0x20, 0xf6, 0xba, 0x4f, 0xdf, 0x09, 
+    0x45, 0xb0, 0x61, 0x3a, 0x89, 0x85, 0x99, 0x15, 0x41, 0x4b, 
+    0xea, 0x2f, 0xa1, 0x35, 0xba, 0x23, 0xe2, 0xd8, 0x0a, 0x43, 
+    0xa9, 0xc6, 0x68, 0x21, 0xa9, 0xd4, 0x6a, 0x23, 0xa9, 0xd0, 
+    0x95, 0xdc, 0x56, 0x2f, 0x22, 0x82, 0x8d, 0xa1, 0x5f, 0xd8, 
+    0xc0, 0x9f, 0x61, 0x63, 0x1b, 0xe4, 0x7c, 0x40, 0xa1, 0x32, 
+    0xbf, 0x0f, 0x79, 0x56, 0x3f, 0x4e, 0x19, 0x22, 0x3f, 0x4e, 
+    0x19, 0x26, 0xc0, 0xb1, 0xe6, 0xd9, 0x73, 0x1c, 0x55, 0xa0, 
+    0x61, 0xdd, 0xe3, 0x4b, 0x97, 0xb1, 0xe1, 0x31, 0x26, 0xd5, 
+    0xfb, 0xaa, 0x86, 0x65, 0x41, 0xab, 0x82, 0xcc, 0xc2, 0x3b, 
+    0x82, 0xcc, 0x83, 0x3b, 0x7d, 0x33, 0x3c, 0xc4, 0x81, 0xa5, 
+    0xf5, 0xe1, 0x9a, 0x9d, 0xba, 0x62, 0x10, 0xf7, 0xcd, 0x1f, 
+    0x63, 0x15, 0x35, 0x79, 0x23, 0x09, 0xf8, 0x18, 0x69, 0x59, 
+    0xfc, 0x58, 0x63, 0x59, 0xfc, 0x58, 0x9c, 0xa6, 0x03, 0xa7, 
+    0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x04, 0x3b, 0x3b, 0xd4, 0x29, 0xf6, 0x00, 0x55, 
+    0x52, 0xb9, 0x1d, 0x70, 0x46, 0x8d, 0x66, 0xdd, 0x55, 0xab, 
+    0xa6, 0x7f, 0x50, 0xb1, 0xee, 0xff, 0x10, 0xa1, 0xee, 0xff, 
+    0xef, 0x5e, 0x11, 0x00, 0x76, 0x2a, 0x8b, 0x86, 0x12, 0x2c, 
+    0xb8, 0x77, 0x7a, 0xc5, 0x6d, 0x83, 0x79, 0x58, 0x39, 0x84, 
+    0x86, 0xc4, 0x99, 0x4e, 0x3c, 0xb4, 0xd4, 0x40, 0x3c, 0xb4, 
+    0x94, 0x40, 0xc3, 0x4b, 0x6b, 0xbf, 0x01, 0xe4, 0x14, 0x40, 
+    0x64, 0xa1, 0x03, 0x58, 0xa7, 0x78, 0x65, 0x9b, 0x20, 0xd0, 
+    0x0e, 0x64, 0xd5, 0xdf, 0x5f, 0x88, 0x65, 0xfe, 0x6b, 0xa5, 
+    0x65, 0xfe, 0x6f, 0xa5, 0x9a, 0x01, 0x90, 0x5a, 0x44, 0x37, 
+    0xeb, 0xcf, 0xf0, 0x23, 0x0a, 0x0c, 0x55, 0xce, 0x91, 0x22, 
+    0xf7, 0xdd, 0xc8, 0x4e, 0xa7, 0xfa, 0x08, 0xaf, 0x9f, 0xd5, 
+    0x48, 0xd0, 0xb7, 0xdd, 0x48, 0xdd, 0x48, 0x22, 0xb7, 0x22, 
+    0xf6, 0x4d, 0x6d, 0x15, 0x05, 0xdb, 0x34, 0xb2, 0xa6, 0x73, 
+    0xa1, 0x01, 0x0c, 0x6e, 0x35, 0xd6, 0x58, 0x18, 0x80, 0x9c, 
+    0x4a, 0x08, 0x08, 0xb4, 0x48, 0x08, 0x08, 0xb4, 0xb7, 0xf7, 
+    0xf7, 0x4b, 0xd8, 0x1a, 0xc3, 0x7b, 0xd2, 0x74, 0x64, 0x21, 
+    0x0b, 0xc6, 0xd6, 0xb0, 0x68, 0x2a, 0xf3, 0xab, 0x74, 0x17, 
+    0x59, 0x69, 0x7c, 0x12, 0x88, 0xeb, 0x7c, 0x12, 0x98, 0xeb, 
+    0x83, 0xed, 0x67, 0x14, 0xd8, 0x4e, 0xad, 0x85, 0x03, 0xda, 
+    0x33, 0x63, 0xf3, 0x94, 0xdb, 0x23, 0xe1, 0x5d, 0x5f, 0x21, 
+    0x7b, 0xfd, 0x27, 0x1d, 0xec, 0x95, 0x1f, 0x53, 0xf9, 0x95, 
+    0x1f, 0x43, 0x06, 0x6a, 0xe0, 0xbc, 0x41, 0x9e, 0x3d, 0x77, 
+    0xc5, 0x9e, 0x5d, 0xba, 0x2b, 0x53, 0x0d, 0xae, 0xda, 0x92, 
+    0x81, 0x9c, 0x8a, 0xd6, 0x4b, 0xcb, 0xc6, 0x99, 0x46, 0xac, 
+    0xc6, 0x9d, 0x47, 0xac, 0x39, 0x62, 0xb8, 0x53, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x82, 0x87, 0x45, 0x45, 0x86, 0xdc, 0xb9, 0xe2, 0xca, 0x9b, 
+    0x84, 0xbb, 0x09, 0x78, 0x1b, 0xcf, 0x21, 0x2d, 0x2f, 0x53, 
+    0x85, 0xbe, 0x3e, 0xfb, 0x81, 0xbd, 0x3e, 0xfb, 0x7e, 0x42, 
+    0xc1, 0x04, 0xbb, 0x16, 0xd2, 0x57, 0x65, 0xb9, 0xde, 0x74, 
+    0x57, 0x17, 0xea, 0x24, 0x58, 0xf8, 0xb5, 0x62, 0x82, 0x5b, 
+    0x05, 0xbd, 0x97, 0x47, 0x64, 0x21, 0x93, 0x53, 0xe0, 0x21, 
+    0x6c, 0xac, 0x1f, 0xde, 0xee, 0x00, 0x50, 0x14, 0x63, 0xc9, 
+    0x80, 0xb5, 0xb0, 0x6c, 0x69, 0xfb, 0xcb, 0x54, 0xb3, 0x52, 
+    0x28, 0xa1, 0xde, 0x66, 0x32, 0x71, 0xa0, 0x52, 0x22, 0x71, 
+    0xa0, 0x52, 0xdd, 0x8e, 0x5f, 0xad, 0xae, 0x05, 0xf3, 0x7e, 
+    0x6e, 0x99, 0x69, 0xfa, 0x86, 0x04, 0xe6, 0x7f, 0xe1, 0xd1, 
+    0xdc, 0x00, 0x97, 0xd8, 0x6d, 0x5f, 0xc5, 0x5f, 0x48, 0x5f, 
+    0x85, 0x5d, 0x48, 0x5f, 0x7a, 0xa2, 0xb7, 0xa0, 0xb9, 0x99, 
+    0xd9, 0x4d, 0xbe, 0x56, 0xd4, 0xca, 0x76, 0xad, 0xf8, 0x9e, 
+    0xc1, 0xdd, 0x99, 0x0f, 0x7d, 0xb7, 0x3b, 0xa6, 0x75, 0x3f, 
+    0x7e, 0xb7, 0x75, 0xbf, 0x7b, 0xb7, 0x8a, 0x40, 0x84, 0x48, 
+    0x6f, 0x69, 0x1f, 0x7b, 0x05, 0xa6, 0xe5, 0xe7, 0xff, 0x7d, 
+    0x1a, 0x65, 0x1b, 0x25, 0x26, 0xee, 0xce, 0xb0, 0x18, 0x5a, 
+    0xde, 0xb1, 0x39, 0xc5, 0xde, 0xb1, 0x1b, 0xcc, 0x21, 0x4e, 
+    0xe4, 0x33, 0xcd, 0xac, 0x03, 0xb1, 0xef, 0xa5, 0xb5, 0x90, 
+    0x7b, 0xfb, 0x12, 0xf4, 0x7b, 0x35, 0xf6, 0x33, 0x97, 0xb3, 
+    0x55, 0x29, 0x61, 0xb3, 0x55, 0x31, 0x61, 0xb1, 0x55, 0x31, 
+    0x9e, 0x4e, 0xaa, 0xce, 0xf5, 0x1b, 0xc8, 0xed, 0x32, 0x8b, 
+    0xa8, 0x53, 0x22, 0x96, 0x56, 0x3a, 0x40, 0x49, 0x7e, 0x5e, 
+    0xb5, 0x20, 0x0f, 0x18, 0x86, 0x50, 0x2a, 0x2a, 0x86, 0x50, 
+    0x2a, 0x3a, 0x79, 0xaf, 0xd5, 0xc5, 0x00, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x9d, 0xe1, 
+    0x7a, 0x57, 0x02, 0xc3, 0x96, 0x40, 0xd8, 0x31, 0x97, 0x0e, 
+    0x76, 0x69, 0xd4, 0xa8, 0x12, 0x56, 0x5b, 0xad, 0x30, 0x74, 
+    0x1c, 0x8a, 0x30, 0x74, 0x1d, 0x88, 0xcf, 0x8b, 0xe2, 0x77, 
+    0x77, 0xa3, 0x56, 0x23, 0x21, 0x03, 0xc4, 0xa9, 0x27, 0xdd, 
+    0x3d, 0x7e, 0x29, 0xa5, 0xca, 0xc6, 0x84, 0x4a, 0xda, 0x5a, 
+    0xa4, 0xe4, 0xd6, 0xca, 0xa4, 0xe4, 0xd6, 0xca, 0x5b, 0x1b, 
+    0x29, 0x35, 0xd3, 0x77, 0x5a, 0x14, 0x6a, 0x59, 0xdc, 0xaf, 
+    0x98, 0xda, 0x97, 0x10, 0x8b, 0xa1, 0x8c, 0xf0, 0xe1, 0x38, 
+    0xaf, 0xa4, 0x81, 0xa7, 0xce, 0xe4, 0x81, 0xa6, 0x8e, 0xe4, 
+    0x7e, 0x59, 0x71, 0x1b, 0x03, 0x43, 0x26, 0x3e, 0x29, 0xd6, 
+    0xff, 0x60, 0x8b, 0x94, 0x72, 0x6c, 0x6c, 0xbe, 0x62, 0xfc, 
+    0x51, 0xa6, 0x3d, 0x62, 0x48, 0xab, 0x2d, 0xfe, 0x48, 0xab, 
+    0x2d, 0xfe, 0xb7, 0x54, 0xd2, 0x01, 0x2d, 0xa6, 0xdd, 0xf0, 
+    0x2a, 0x9a, 0xe3, 0x81, 0x8e, 0xe1, 0xf4, 0x42, 0x21, 0x10, 
+    0x66, 0x6b, 0xdd, 0xb5, 0x8e, 0x4c, 0x0e, 0x51, 0xcf, 0x72, 
+    0x0c, 0x71, 0xce, 0x72, 0xf3, 0x8e, 0x31, 0x8d, 0x71, 0x03, 
+    0xba, 0xc5, 0xc6, 0x42, 0xfe, 0xd4, 0xcc, 0x04, 0xb4, 0xd3, 
+    0x12, 0x99, 0xc5, 0x99, 0x58, 0x45, 0x3f, 0xe8, 0x99, 0xd9, 
+    0x1e, 0x59, 0x99, 0xcd, 0x9e, 0xd9, 0x66, 0x32, 0x61, 0x26, 
+    0x55, 0x01, 0xf8, 0xf1, 0xe3, 0x4e, 0xf6, 0x6d, 0x4e, 0x27, 
+    0xdf, 0x75, 0xb9, 0x69, 0x06, 0x99, 0x70, 0x84, 0x4a, 0x6a, 
+    0x36, 0xa6, 0x43, 0x6c, 0x34, 0x84, 0x42, 0x6c, 0xcb, 0x7b, 
+    0xbd, 0x93, 0x9e, 0x8f, 0xcb, 0xd2, 0xf9, 0x3a, 0xe5, 0x16, 
+    0x75, 0x34, 0x7b, 0xd0, 0x46, 0x4e, 0xfc, 0x82, 0xea, 0xab, 
+    0x23, 0x01, 0x4a, 0x39, 0x06, 0x02, 0x4a, 0x39, 0x26, 0x02, 
+    0xb5, 0xc6, 0xd9, 0xfd, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x55, 0xfe, 0x07, 0xbc, 
+    0x1a, 0xd1, 0x1d, 0x62, 0xb7, 0xcb, 0xfb, 0xf2, 0x3a, 0xc4, 
+    0xa8, 0xb2, 0x90, 0x25, 0x12, 0x25, 0x9e, 0xb4, 0x92, 0x69, 
+    0x9e, 0xa5, 0x92, 0x69, 0x61, 0x5a, 0x6d, 0x96, 0x85, 0x2c, 
+    0xb7, 0x43, 0xbb, 0x58, 0x0f, 0xf6, 0x49, 0x10, 0x36, 0x28, 
+    0xc0, 0xde, 0xec, 0x29, 0x48, 0x46, 0xe7, 0x3f, 0x0c, 0x57, 
+    0xa6, 0x3b, 0x08, 0x57, 0xa6, 0x3b, 0xf7, 0xa8, 0x59, 0xc4, 
+    0x56, 0x88, 0x07, 0x77, 0x0d, 0x50, 0x41, 0x93, 0xa3, 0x6b, 
+    0x26, 0x85, 0x6d, 0xd8, 0xdd, 0x7b, 0x00, 0xd0, 0xa7, 0x19, 
+    0x8c, 0xa8, 0xc7, 0x89, 0x88, 0xa8, 0xc7, 0x89, 0x77, 0x57, 
+    0x38, 0x76, 0x7a, 0x99, 0xee, 0x29, 0x21, 0x04, 0x71, 0x8f, 
+    0x8a, 0x10, 0x3f, 0x6b, 0x7a, 0xe4, 0x8d, 0x74, 0x4f, 0x99, 
+    0xa1, 0x71, 0x37, 0xcc, 0x1d, 0x60, 0x37, 0xcc, 0x9d, 0x60, 
+    0xc8, 0x33, 0x62, 0x9f, 0x8c, 0x6c, 0xed, 0xb5, 0xa1, 0xb0, 
+    0x6a, 0x75, 0xb2, 0x79, 0xac, 0xba, 0x4a, 0x50, 0x9c, 0x08, 
+    0x34, 0x95, 0x21, 0x0c, 0x70, 0x99, 0xed, 0x8b, 0x70, 0x99, 
+    0xed, 0x8b, 0x8f, 0x66, 0x12, 0x74, 0xd1, 0x4f, 0x29, 0xd6, 
+    0x91, 0x18, 0xfb, 0x98, 0xc8, 0xfa, 0x08, 0x4c, 0xd9, 0x5f, 
+    0x17, 0x7e, 0x2d, 0x81, 0x9b, 0x9c, 0x9c, 0xa3, 0x13, 0x3d, 
+    0x9c, 0xa3, 0x13, 0x3d, 0x63, 0x5c, 0xec, 0xc2, 0x7c, 0x6f, 
+    0x6c, 0x95, 0xa4, 0xcf, 0xec, 0x02, 0x94, 0xa6, 0x86, 0xa3, 
+    0x3f, 0x69, 0x96, 0x3d, 0x1a, 0xdd, 0xd0, 0xf1, 0x10, 0x84, 
+    0xce, 0x21, 0x10, 0xc5, 0xce, 0x21, 0xef, 0x3a, 0x31, 0xde, 
+    0x3f, 0xda, 0x87, 0x1f, 0x1e, 0xb6, 0x69, 0x34, 0x95, 0x1c, 
+    0xd3, 0x67, 0xd6, 0xa0, 0x25, 0x98, 0x26, 0x1a, 0x4c, 0xb2, 
+    0x02, 0x2a, 0x1f, 0xae, 0x06, 0x2a, 0x1d, 0xaa, 0xf9, 0xd5, 
+    0xe2, 0x55, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x89, 0x46, 0x8b, 0x02, 0xaf, 0x78, 
+    0x3d, 0x9f, 0x52, 0xf3, 0xa6, 0x4a, 0xbd, 0xc3, 0x39, 0x7f, 
+    0xc6, 0x6c, 0xd2, 0x95, 0xc8, 0x66, 0xf6, 0xfd, 0xc8, 0x66, 
+    0xf6, 0xfd, 0x37, 0x99, 0x09, 0x02, 0xbe, 0x4b, 0xf3, 0xd7, 
+    0x43, 0x58, 0x80, 0xf3, 0x0a, 0xeb, 0xd9, 0x24, 0x0e, 0x40, 
+    0x2b, 0x68, 0x93, 0x12, 0x7f, 0xf6, 0x96, 0x19, 0x2f, 0xd5, 
+    0x96, 0x13, 0x2d, 0xf6, 0x69, 0xec, 0xd2, 0x09, 0xea, 0xb8, 
+    0x07, 0x8e, 0x3d, 0xea, 0x71, 0x6b, 0x94, 0x2e, 0xf6, 0x1c, 
+    0xaf, 0x90, 0xd5, 0xd7, 0x15, 0x78, 0x1c, 0x9d, 0xa5, 0x18, 
+    0x1e, 0xc9, 0xa5, 0x18, 0x1c, 0xcd, 0x5a, 0xe7, 0xe3, 0x32, 
+    0xaa, 0x55, 0x62, 0x05, 0x0d, 0x14, 0xa6, 0x2a, 0x55, 0xbb, 
+    0x99, 0x0e, 0x82, 0x86, 0xe7, 0x4a, 0xf7, 0x92, 0xe9, 0x4b, 
+    0xa7, 0x71, 0xa9, 0x4b, 0xb7, 0x90, 0xa9, 0x4b, 0x48, 0x6f, 
+    0x56, 0xb4, 0xe2, 0x3a, 0x02, 0x50, 0xb7, 0xed, 0x73, 0xf0, 
+    0xee, 0x20, 0x1f, 0xf7, 0x41, 0x9b, 0xbb, 0x16, 0x96, 0x81, 
+    0xe1, 0x5b, 0xe3, 0x82, 0x82, 0x5e, 0x43, 0x82, 0x83, 0x5e, 
+    0xbc, 0x7d, 0x7c, 0xa1, 0xb3, 0x34, 0xda, 0x07, 0x6c, 0xdf, 
+    0xc0, 0x69, 0x43, 0x40, 0xf8, 0x39, 0x54, 0xd2, 0x27, 0xaf, 
+    0x11, 0x38, 0x3c, 0x06, 0xe0, 0x18, 0xb4, 0xae, 0xc0, 0x18, 
+    0x34, 0xae, 0x3f, 0xe7, 0xcb, 0x51, 0xe7, 0xca, 0x17, 0x92, 
+    0xd6, 0x51, 0x00, 0x27, 0xef, 0x54, 0x52, 0x97, 0xc3, 0xac, 
+    0x98, 0x5b, 0xf0, 0x1f, 0xb9, 0x88, 0xfd, 0x76, 0xd1, 0xa3, 
+    0xdd, 0x56, 0xd1, 0xa3, 0x22, 0xa9, 0x2e, 0x5c, 0xe6, 0x3b, 
+    0x8a, 0xeb, 0x1e, 0x1b, 0x31, 0xf4, 0xd1, 0x84, 0x2b, 0x09, 
+    0xbf, 0x71, 0xab, 0x4c, 0x86, 0x04, 0x64, 0x1b, 0x96, 0x30, 
+    0xe5, 0x28, 0x86, 0x30, 0xe1, 0x28, 0x79, 0xcf, 0x1e, 0xd7, 
+    0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0xad, 0x70, 0x07, 0x15, 0x41, 0xe2, 0xda, 0xf8, 
+    0x5f, 0xfa, 0x29, 0x0b, 0x74, 0xd5, 0x70, 0x2e, 0x25, 0xac, 
+    0xc1, 0xbe, 0x43, 0x9b, 0xc0, 0x84, 0x47, 0x99, 0xc0, 0x84, 
+    0xb8, 0x66, 0x3f, 0x7b, 0x86, 0xe6, 0x84, 0x37, 0x13, 0xf4, 
+    0x73, 0x2b, 0x32, 0x37, 0xa9, 0x86, 0x87, 0x2f, 0x4e, 0xd1, 
+    0x53, 0x3d, 0xc6, 0x78, 0x53, 0xb7, 0x77, 0x7a, 0x53, 0x37, 
+    0x77, 0x78, 0xac, 0xc8, 0x88, 0x87, 0x2c, 0x1e, 0xe1, 0xfa, 
+    0xb6, 0xbb, 0x93, 0x32, 0x33, 0xe2, 0x1a, 0xe0, 0x14, 0x80, 
+    0xe5, 0xa4, 0x1d, 0x92, 0x2a, 0x11, 0x5c, 0xba, 0x49, 0x29, 
+    0x5c, 0xb2, 0x49, 0x29, 0xa3, 0x4d, 0xb6, 0xd6, 0x04, 0xd3, 
+    0x44, 0x0e, 0x03, 0x98, 0x01, 0x01, 0xec, 0xf0, 0xf4, 0xb8, 
+    0xc3, 0x81, 0x00, 0x66, 0xbc, 0x34, 0x46, 0xcf, 0x51, 0x01, 
+    0x17, 0x0f, 0xd1, 0x21, 0x16, 0x0f, 0x2e, 0xde, 0xe9, 0xf0, 
+    0x8e, 0x46, 0x50, 0x4f, 0x38, 0x65, 0x3c, 0xf8, 0x30, 0xc5, 
+    0x6d, 0x04, 0x18, 0x23, 0x95, 0xb4, 0xb8, 0x9d, 0x9e, 0x9c, 
+    0x7a, 0xad, 0x94, 0x5b, 0x7a, 0xad, 0x94, 0x99, 0x85, 0x52, 
+    0x6b, 0x66, 0x5d, 0x7d, 0xbe, 0x84, 0xdf, 0xa9, 0xa8, 0xef, 
+    0xd6, 0x05, 0x06, 0x34, 0xcc, 0x56, 0x9f, 0x5c, 0x76, 0x96, 
+    0x25, 0x99, 0x3e, 0x5d, 0xb7, 0x94, 0x3e, 0x5c, 0x37, 0x9c, 
+    0xc1, 0xa3, 0xc8, 0x63, 0x84, 0x7c, 0x30, 0x1e, 0x2d, 0x81, 
+    0x43, 0x76, 0xb9, 0xdf, 0x6a, 0x4b, 0x53, 0xd7, 0xb4, 0x83, 
+    0x11, 0xfb, 0x1b, 0xe5, 0x0b, 0xe3, 0xe3, 0x61, 0x0b, 0xe3, 
+    0x73, 0x65, 0xf4, 0x1c, 0x8c, 0x9a, 0x42, 0xb0, 0x81, 0xf5, 
+    0x3f, 0xd6, 0x5e, 0x2b, 0x09, 0xd1, 0x68, 0x78, 0xf6, 0x7c, 
+    0x34, 0x97, 0xb3, 0x72, 0xf5, 0x7b, 0x82, 0xb8, 0x7c, 0x1d, 
+    0x82, 0xb8, 0x7c, 0x5d, 0x7d, 0x47, 0x83, 0xa2, 0x00, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x42, 0x78, 0x38, 0x30, 0xe9, 0x08, 0x72, 0x8b, 0x90, 0xd1, 
+    0x9b, 0xa8, 0x8d, 0xe8, 0x21, 0xd0, 0x7a, 0xf5, 0x97, 0xe7, 
+    0xfc, 0xf1, 0x37, 0x39, 0xfc, 0xf1, 0x37, 0x39, 0x03, 0x0e, 
+    0xc8, 0xc6, 0xa8, 0x3a, 0x07, 0x1f, 0xc9, 0x89, 0x53, 0xa3, 
+    0x67, 0x07, 0xe8, 0xd0, 0x18, 0x6d, 0x74, 0x14, 0xcf, 0xbd, 
+    0xd8, 0x4a, 0x27, 0x2e, 0xd8, 0x0e, 0x27, 0x2e, 0xd8, 0x0e, 
+    0xd8, 0xd1, 0x27, 0xf1, 0x20, 0x4d, 0x61, 0xe8, 0xb8, 0x12, 
+    0x70, 0xe7, 0x17, 0x55, 0x50, 0x4b, 0x5c, 0x9c, 0x4e, 0xd7, 
+    0xe5, 0x82, 0x8e, 0x6c, 0xce, 0x22, 0x5e, 0x13, 0xc4, 0x02, 
+    0x5e, 0x53, 0x3b, 0xfd, 0xa1, 0xac, 0x83, 0xaf, 0x89, 0xf6, 
+    0x1d, 0x2f, 0x00, 0x69, 0xe9, 0xc2, 0x12, 0xbe, 0xca, 0x32, 
+    0xbc, 0xd8, 0xcc, 0xb0, 0x18, 0x2f, 0x64, 0xb3, 0x19, 0xa5, 
+    0x6c, 0xb2, 0x19, 0xa5, 0x93, 0x4d, 0xe6, 0x5a, 0xbf, 0x77, 
+    0xce, 0x27, 0xb7, 0xd3, 0x92, 0xf1, 0x5a, 0x95, 0x4e, 0x05, 
+    0x60, 0x79, 0x74, 0xf9, 0x90, 0xf8, 0xb5, 0x6a, 0x16, 0xd9, 
+    0x35, 0x65, 0x16, 0xd9, 0xb5, 0x65, 0xe9, 0x26, 0x4a, 0x9a, 
+    0xa8, 0xac, 0x85, 0xb5, 0xe3, 0x29, 0x9a, 0xb9, 0xbc, 0x19, 
+    0xa6, 0xe1, 0x7e, 0xd0, 0xe6, 0x76, 0xad, 0x7b, 0xeb, 0x09, 
+    0x5d, 0x51, 0xe3, 0x0e, 0xdd, 0x51, 0xe3, 0x0c, 0x22, 0xae, 
+    0x1c, 0xf3, 0x3a, 0x2a, 0xb9, 0x43, 0xfd, 0x01, 0x22, 0x4a, 
+    0x5c, 0xc7, 0xff, 0x62, 0xe6, 0x8c, 0xdd, 0x14, 0x57, 0x22, 
+    0x1c, 0xff, 0x74, 0x07, 0x0d, 0xf1, 0x74, 0x27, 0x1d, 0xf7, 
+    0x8b, 0xd8, 0xe2, 0x08, 0x4a, 0x4e, 0x03, 0xa9, 0xa5, 0xff, 
+    0x80, 0xfa, 0x45, 0x34, 0x7e, 0xde, 0x4c, 0x65, 0x0e, 0x86, 
+    0xae, 0xc1, 0xe9, 0x00, 0x61, 0xc0, 0x42, 0xc2, 0xe1, 0xe0, 
+    0x42, 0xc2, 0x1e, 0x1f, 0xbd, 0x3d, 0x01, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x2d, 0xcf, 
+    0xaf, 0x90, 0x1a, 0xc5, 0x29, 0x00, 0x03, 0x62, 0xfb, 0xdf, 
+    0x4d, 0x4b, 0xdf, 0x21, 0x43, 0xe1, 0x2b, 0x0a, 0x86, 0x49, 
+    0xfe, 0x8c, 0x87, 0x49, 0xfe, 0x8c, 0x78, 0xb6, 0x01, 0x73, 
+    0xf6, 0x3f, 0x6f, 0xd2, 0xe2, 0xf7, 0x78, 0x40, 0xb9, 0xd6, 
+    0xe0, 0xb4, 0x21, 0x29, 0x0e, 0x8c, 0x83, 0xcc, 0xcf, 0xb2, 
+    0xa3, 0xd3, 0x4f, 0xb8, 0xa3, 0xd3, 0x4f, 0xb8, 0x5c, 0x2c, 
+    0xb0, 0x47, 0x42, 0xaf, 0x60, 0x7a, 0x4a, 0x01, 0xf7, 0xea, 
+    0xdb, 0x4f, 0xf9, 0xd6, 0xd9, 0x75, 0xe1, 0x68, 0xcc, 0x03, 
+    0xf5, 0x67, 0xb8, 0x95, 0x34, 0x67, 0xd8, 0x15, 0xf4, 0x67, 
+    0x27, 0xea, 0x0b, 0x98, 0x5c, 0x21, 0x3f, 0x10, 0x7c, 0xf6, 
+    0x39, 0x60, 0x9c, 0x3c, 0xb2, 0x6b, 0x01, 0x2f, 0x0b, 0x8b, 
+    0xd4, 0x24, 0xbf, 0x0b, 0x96, 0x0c, 0xab, 0x03, 0x96, 0x0c, 
+    0xbb, 0x8b, 0x69, 0xf3, 0x44, 0x74, 0xff, 0xec, 0xa7, 0x4e, 
+    0x06, 0x09, 0x49, 0x8a, 0x3f, 0xc6, 0x0f, 0x2e, 0x55, 0x98, 
+    0xd2, 0x44, 0x04, 0xae, 0x2e, 0xa0, 0xd4, 0x86, 0x6e, 0xa8, 
+    0x54, 0x8e, 0x6e, 0xa8, 0xab, 0x71, 0x91, 0x57, 0xeb, 0x8d, 
+    0xe3, 0x92, 0x9d, 0x42, 0xa5, 0x88, 0x53, 0xd1, 0xfe, 0x8b, 
+    0x98, 0xef, 0xe3, 0x0b, 0xf6, 0x41, 0x1f, 0xed, 0x50, 0x4b, 
+    0x9f, 0x5f, 0x50, 0x4b, 0x9f, 0x4d, 0xaf, 0xb4, 0x60, 0xb2, 
+    0x95, 0xa2, 0x74, 0x96, 0xa2, 0x13, 0xc0, 0x3a, 0x16, 0x37, 
+    0x67, 0xb5, 0xa6, 0xde, 0x74, 0x1e, 0x26, 0xba, 0x8d, 0x3c, 
+    0xa4, 0x5f, 0x9f, 0x34, 0xa4, 0x5f, 0x9f, 0x34, 0x5b, 0xa0, 
+    0x60, 0xcb, 0xd6, 0xd7, 0x42, 0xf8, 0x12, 0x3c, 0x4b, 0xe4, 
+    0xca, 0xf0, 0xd2, 0x55, 0x6f, 0xfe, 0xa5, 0x6b, 0x64, 0xfd, 
+    0xc3, 0x1b, 0xe9, 0x77, 0xce, 0x1b, 0xe9, 0x77, 0xce, 0x1b, 
+    0x16, 0x88, 0x31, 0xe4, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x5e, 0xcd, 0xe5, 0x27, 
+    0x6b, 0x60, 0xe6, 0x30, 0xc4, 0xd6, 0xbb, 0xd8, 0xbe, 0x1d, 
+    0x9b, 0xa8, 0xde, 0x19, 0xec, 0x31, 0xfe, 0x9c, 0xe4, 0xa3, 
+    0xfe, 0x1c, 0xe4, 0xa0, 0x01, 0xe3, 0x1b, 0x5f, 0x4e, 0x58, 
+    0xd8, 0x86, 0x9a, 0xa5, 0xc2, 0x00, 0x86, 0xee, 0x79, 0x65, 
+    0x24, 0xa4, 0x73, 0xcd, 0xad, 0xba, 0x27, 0x03, 0x06, 0x39, 
+    0x6b, 0x81, 0x06, 0x39, 0x63, 0x81, 0xf9, 0xc6, 0x9c, 0x7e, 
+    0x9d, 0xc6, 0xb2, 0xe3, 0xa3, 0xcf, 0xc4, 0x58, 0xd8, 0x61, 
+    0x76, 0xaa, 0x78, 0x03, 0x07, 0xef, 0xa1, 0x9b, 0xff, 0xf0, 
+    0xb1, 0x2b, 0x5f, 0xd8, 0xb1, 0x2b, 0x5f, 0xc8, 0x4e, 0xd4, 
+    0xa0, 0x37, 0x77, 0x6a, 0x53, 0x55, 0x2a, 0xbb, 0xf6, 0x31, 
+    0x0e, 0x70, 0x60, 0x94, 0x58, 0x33, 0x4d, 0x49, 0xef, 0xea, 
+    0x71, 0xdf, 0x64, 0x10, 0x12, 0x4b, 0x6e, 0x10, 0x10, 0x4b, 
+    0x91, 0xef, 0xef, 0xb4, 0xb7, 0xdc, 0xd2, 0x98, 0x57, 0x81, 
+    0x3c, 0xd2, 0xe2, 0x43, 0xb8, 0xa2, 0xad, 0xa4, 0xf5, 0x08, 
+    0x19, 0xb7, 0xb2, 0x6d, 0x6b, 0xbc, 0xbb, 0x0e, 0x69, 0xb4, 
+    0xbb, 0x0e, 0x96, 0x4b, 0x44, 0xf1, 0x45, 0x3c, 0x05, 0x69, 
+    0x15, 0x18, 0xf9, 0xec, 0x04, 0x36, 0x72, 0x17, 0x43, 0x4d, 
+    0x2c, 0x88, 0x54, 0x74, 0xdb, 0xf3, 0x60, 0xf4, 0xbd, 0xba, 
+    0x40, 0x74, 0xbd, 0xaa, 0xbf, 0x8b, 0x42, 0x55, 0xff, 0xf8, 
+    0x65, 0xd2, 0x7b, 0x20, 0x2d, 0x3b, 0xa6, 0x4b, 0x66, 0x56, 
+    0xe3, 0xe3, 0x23, 0x87, 0xe0, 0x30, 0x2d, 0xbe, 0x7c, 0x61, 
+    0x2d, 0xcb, 0x7c, 0x61, 0x2d, 0x8b, 0x83, 0x9e, 0xd2, 0x74, 
+    0xb6, 0xc4, 0xb4, 0xf7, 0xf0, 0x9c, 0xe5, 0x40, 0x60, 0x9c, 
+    0x9c, 0x1b, 0xee, 0xe4, 0x8f, 0x88, 0x10, 0xe2, 0x91, 0x31, 
+    0x50, 0xd5, 0x9c, 0x50, 0x50, 0xd5, 0x9c, 0x10, 0xaf, 0x2a, 
+    0x63, 0xef, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0xd0, 0x71, 0xc5, 0x4a, 0x13, 0x27, 
+    0xe6, 0x0e, 0x3d, 0x53, 0x17, 0x33, 0x86, 0xf7, 0x72, 0xd8, 
+    0x96, 0x12, 0xc1, 0x58, 0x52, 0x5a, 0xd1, 0x98, 0x52, 0x5b, 
+    0xd1, 0x98, 0xad, 0xa4, 0x2e, 0x67, 0xf3, 0xcd, 0x40, 0x44, 
+    0xbd, 0x28, 0x21, 0x89, 0xa6, 0xb4, 0x9a, 0xb8, 0x30, 0xae, 
+    0xe8, 0xee, 0x2d, 0x39, 0xe2, 0x2e, 0x6f, 0x7c, 0xf1, 0x5c, 
+    0x6f, 0x78, 0xe3, 0x6e, 0x90, 0x87, 0x1c, 0x91, 0xc9, 0x8d, 
+    0xc2, 0x64, 0x2b, 0x03, 0x86, 0x5c, 0x94, 0xa8, 0xe6, 0x7a, 
+    0xf7, 0x40, 0x86, 0xbe, 0xbb, 0x61, 0x1d, 0xe0, 0x99, 0xf9, 
+    0xbd, 0xb0, 0x99, 0xf9, 0xbd, 0xb0, 0x66, 0x06, 0x42, 0x4f, 
+    0xc6, 0xb5, 0xef, 0xf2, 0xcb, 0x0c, 0xae, 0x1b, 0x09, 0x8a, 
+    0x03, 0xcd, 0xb3, 0x7b, 0x22, 0x02, 0x76, 0xe8, 0xab, 0x32, 
+    0x74, 0xcc, 0x22, 0x26, 0x76, 0xcc, 0x22, 0x22, 0x89, 0x33, 
+    0xdd, 0xdd, 0xcb, 0xca, 0xe4, 0x24, 0x0f, 0x9b, 0x44, 0xb0, 
+    0x0c, 0xec, 0x1d, 0x24, 0xe5, 0xc1, 0x47, 0xb3, 0xbc, 0x60, 
+    0xad, 0x51, 0xe2, 0xe2, 0x2d, 0x5d, 0xe0, 0xe2, 0x2d, 0x55, 
+    0x1f, 0x1d, 0xd2, 0xaa, 0x78, 0x65, 0x42, 0x20, 0xd9, 0xc6, 
+    0x7f, 0x58, 0x02, 0x8b, 0x8b, 0x54, 0x64, 0xea, 0xc5, 0x56, 
+    0x16, 0x04, 0x75, 0xf0, 0x74, 0x9c, 0xcc, 0x80, 0x54, 0x8e, 
+    0xc5, 0xf0, 0xab, 0x71, 0x3a, 0x0f, 0x31, 0x9e, 0x75, 0x1e, 
+    0x41, 0x8f, 0x48, 0x84, 0x25, 0x09, 0xd0, 0x30, 0xc0, 0x2f, 
+    0xda, 0xba, 0x0e, 0x33, 0xb8, 0xea, 0x46, 0x2d, 0xf4, 0x73, 
+    0x46, 0x2f, 0xf4, 0xf3, 0xb9, 0xd0, 0x0b, 0x0c, 0x2a, 0x2c, 
+    0xab, 0x38, 0x2c, 0xa1, 0x4e, 0xac, 0x04, 0x56, 0xf1, 0x8a, 
+    0xca, 0xc3, 0x14, 0xb6, 0x15, 0x83, 0x73, 0x6d, 0xb7, 0x91, 
+    0x18, 0x47, 0x97, 0x93, 0x10, 0x47, 0x68, 0x6c, 0xef, 0xb8, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x5d, 0x80, 0xb1, 0x27, 0xb4, 0xb2, 0x4b, 0x0c, 
+    0x6a, 0xaa, 0x6d, 0x8a, 0xf7, 0xae, 0xb4, 0x61, 0xb2, 0x81, 
+    0x47, 0x95, 0xf4, 0x8f, 0xc0, 0x83, 0xf4, 0x8b, 0xc0, 0x87, 
+    0x0b, 0x74, 0x3f, 0x78, 0x8b, 0xc9, 0x6d, 0x3d, 0x54, 0x1e, 
+    0x54, 0xdd, 0x68, 0x2b, 0xf7, 0xb2, 0x86, 0x79, 0xdb, 0x0a, 
+    0xdd, 0x1b, 0x4a, 0xb4, 0xe0, 0x06, 0xa7, 0x14, 0xe0, 0x0e, 
+    0xa7, 0x14, 0x1f, 0xf1, 0x58, 0xeb, 0x66, 0xf8, 0x4f, 0x13, 
+    0x43, 0x34, 0x4c, 0x27, 0x1d, 0xee, 0xe6, 0x19, 0x14, 0xc8, 
+    0x9d, 0x41, 0x68, 0x71, 0xba, 0xe5, 0x7f, 0x15, 0xac, 0xa9, 
+    0x7d, 0x35, 0xae, 0xe9, 0x82, 0xca, 0x51, 0x16, 0xe0, 0xaf, 
+    0x12, 0x83, 0xac, 0x0c, 0x63, 0x31, 0x0f, 0xb9, 0x00, 0x51, 
+    0x0e, 0x27, 0xac, 0x84, 0x12, 0x1e, 0x02, 0x4e, 0x0c, 0x2f, 
+    0x61, 0xce, 0x0c, 0x2f, 0x61, 0x4e, 0xf3, 0xd0, 0x9e, 0xb1, 
+    0x7c, 0x5b, 0xcb, 0xde, 0x06, 0x2c, 0x19, 0xeb, 0x19, 0x2e, 
+    0x71, 0x96, 0xc4, 0xac, 0x44, 0x03, 0x5d, 0x1b, 0xb6, 0xae, 
+    0x45, 0xbe, 0x97, 0x8f, 0x45, 0x3e, 0x97, 0x8f, 0xba, 0xc1, 
+    0x68, 0x70, 0x46, 0x67, 0xfa, 0x4b, 0x16, 0xd1, 0x80, 0x61, 
+    0x87, 0xdd, 0x7c, 0x75, 0x0b, 0xe0, 0x48, 0xf4, 0xa6, 0x0b, 
+    0xe1, 0x02, 0x0b, 0x8a, 0x61, 0x40, 0x8f, 0x8b, 0x61, 0x40, 
+    0x70, 0x74, 0x9e, 0xbf, 0x7b, 0xfb, 0xcb, 0x91, 0xf4, 0x67, 
+    0x6c, 0x57, 0xf6, 0x16, 0x4f, 0x75, 0x42, 0x28, 0xbf, 0x0e, 
+    0x62, 0x69, 0x70, 0xba, 0x4a, 0x2a, 0x60, 0xa0, 0x4a, 0x28, 
+    0x60, 0xa2, 0xb5, 0xd7, 0x9f, 0x5d, 0xde, 0x3d, 0x13, 0x11, 
+    0xb6, 0xad, 0x07, 0xf7, 0x65, 0xbf, 0xb0, 0xb6, 0xd6, 0x10, 
+    0x28, 0xf7, 0x9b, 0x62, 0x0b, 0x48, 0x53, 0xb4, 0x2b, 0x33, 
+    0xd3, 0xb4, 0x2b, 0x73, 0x2c, 0x4b, 0xd4, 0x8c, 0x00, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0xbb, 0x06, 0xf8, 0xf4, 0xe2, 0x7f, 0x58, 0xb0, 0x4c, 0x95, 
+    0xc4, 0x15, 0x18, 0xd4, 0x3c, 0xd2, 0x2f, 0xe7, 0x0d, 0x13, 
+    0x19, 0x67, 0x0f, 0x50, 0x39, 0x67, 0x0c, 0x50, 0xc6, 0x98, 
+    0xf3, 0xaf, 0xca, 0x03, 0x23, 0x31, 0x22, 0xf6, 0xe4, 0xc6, 
+    0x43, 0x06, 0xd3, 0x04, 0x83, 0xf9, 0x2b, 0xe0, 0x5a, 0x80, 
+    0x4c, 0x36, 0x12, 0xf0, 0xc1, 0x77, 0x12, 0xd1, 0xc9, 0x77, 
+    0xed, 0x2e, 0x36, 0x88, 0xe9, 0x74, 0x24, 0xbb, 0x6a, 0x18, 
+    0x27, 0x66, 0xfb, 0xab, 0x3c, 0x45, 0xef, 0xba, 0x9c, 0x13, 
+    0xe7, 0xe3, 0xe2, 0x8e, 0x62, 0x97, 0x76, 0x03, 0x62, 0xb3, 
+    0xf6, 0x03, 0x9d, 0x4c, 0x09, 0xfc, 0x3b, 0x18, 0x69, 0x09, 
+    0xc9, 0xb5, 0x86, 0x0b, 0xb5, 0xc5, 0xe9, 0x2f, 0xd8, 0x24, 
+    0x62, 0x33, 0x91, 0x04, 0x64, 0xf8, 0x90, 0x01, 0x14, 0xfa, 
+    0x90, 0x01, 0x20, 0xfa, 0x6f, 0xfe, 0xdf, 0x05, 0x8a, 0x3f, 
+    0x8b, 0x7e, 0xad, 0x86, 0xaf, 0x6e, 0x24, 0x8a, 0x72, 0xc7, 
+    0x72, 0xea, 0x7b, 0xf6, 0x20, 0x25, 0x44, 0xb7, 0xa2, 0xb0, 
+    0x76, 0xc7, 0xa2, 0xa0, 0x76, 0xd7, 0x5d, 0x5f, 0x89, 0x28, 
+    0x3e, 0xa2, 0x76, 0x94, 0x65, 0x98, 0x57, 0xe7, 0x89, 0xb6, 
+    0x26, 0x29, 0x9c, 0xfb, 0xb6, 0xa6, 0xa2, 0x7e, 0xfa, 0xe2, 
+    0xb0, 0xf4, 0xe9, 0xe6, 0xb0, 0xf4, 0xe9, 0xe6, 0x4f, 0x0b, 
+    0x16, 0x19, 0x4d, 0xa5, 0x55, 0xd0, 0x1f, 0x4d, 0x7b, 0x8d, 
+    0xda, 0x14, 0xd1, 0xa6, 0x68, 0x62, 0xbb, 0xe5, 0x22, 0xba, 
+    0x2c, 0xc5, 0xc8, 0x7e, 0x7c, 0x49, 0xc8, 0x7a, 0x7c, 0x4d, 
+    0x37, 0x85, 0x83, 0xb2, 0xc3, 0x5e, 0x01, 0xcd, 0x9c, 0x11, 
+    0xf3, 0x2d, 0x7a, 0xee, 0x11, 0x31, 0xdc, 0xab, 0x44, 0xc4, 
+    0x51, 0x14, 0xb3, 0x77, 0x48, 0xb4, 0x45, 0x01, 0x58, 0xb4, 
+    0x45, 0x41, 0xa7, 0x4b, 0xba, 0xbe, 0x01, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0xbe, 0xc5, 
+    0x05, 0xaf, 0x59, 0x37, 0xee, 0x47, 0xeb, 0x14, 0xa1, 0x64, 
+    0xd2, 0x71, 0x6a, 0x0a, 0x66, 0x20, 0x65, 0xdf, 0xc6, 0x29, 
+    0x75, 0x22, 0x46, 0x21, 0x75, 0x26, 0xb9, 0xde, 0x8a, 0xd9, 
+    0x70, 0x38, 0x0c, 0x91, 0x5f, 0x36, 0x4e, 0xfd, 0x6f, 0x1f, 
+    0x8b, 0x2d, 0x97, 0xe7, 0x26, 0x1e, 0x06, 0x4f, 0x7c, 0xbe, 
+    0x02, 0xe7, 0xb5, 0xae, 0x82, 0xe7, 0xb4, 0xbe, 0x7d, 0x18, 
+    0x4b, 0x41, 0xed, 0xf3, 0x96, 0x19, 0xef, 0x3d, 0x82, 0xea, 
+    0xd6, 0x41, 0xb4, 0x36, 0xd4, 0x62, 0x5b, 0xf6, 0x85, 0xaf, 
+    0x52, 0xf8, 0xae, 0x61, 0x5e, 0xa7, 0xac, 0x61, 0x5e, 0xa7, 
+    0x53, 0x9e, 0xa1, 0x58, 0x63, 0x0e, 0x62, 0xd3, 0x65, 0x9c, 
+    0x9e, 0x88, 0x23, 0xdd, 0xa4, 0xc0, 0x8c, 0x63, 0xdd, 0xd9, 
+    0x45, 0xde, 0x60, 0x0f, 0xd6, 0x46, 0x02, 0xa3, 0x54, 0x4e, 
+    0x02, 0xa3, 0xab, 0xb1, 0xfd, 0x5c, 0xef, 0x16, 0x95, 0xbd, 
+    0x36, 0x9d, 0x3a, 0xd2, 0x62, 0x0c, 0x58, 0x20, 0x86, 0x9f, 
+    0x1d, 0xb8, 0xbf, 0x75, 0x78, 0xa6, 0x17, 0x7c, 0x59, 0x21, 
+    0x97, 0x7c, 0x59, 0x25, 0x68, 0x83, 0xa6, 0xda, 0x5f, 0xd0, 
+    0x64, 0xf4, 0x18, 0x66, 0x4e, 0xbf, 0x5b, 0xfc, 0x3f, 0x39, 
+    0xbe, 0xf4, 0x88, 0xe1, 0xe4, 0x44, 0x5a, 0xd6, 0xf7, 0x67, 
+    0x1a, 0x12, 0xf7, 0x44, 0x1a, 0x12, 0x08, 0xbb, 0xe5, 0xed, 
+    0xbc, 0xf8, 0xe7, 0x64, 0xf8, 0xf0, 0xa5, 0xc5, 0x00, 0xc5, 
+    0xf5, 0x39, 0x05, 0x8a, 0xc1, 0x6d, 0x29, 0xfd, 0x8e, 0x4d, 
+    0x68, 0x94, 0xc7, 0x6d, 0x28, 0x94, 0xc7, 0x6d, 0xd7, 0x6b, 
+    0x38, 0x92, 0x5a, 0xad, 0x07, 0xc2, 0x9b, 0x86, 0xe7, 0x02, 
+    0x5f, 0x10, 0xef, 0x0d, 0xe7, 0xf2, 0x98, 0x78, 0x3d, 0xe8, 
+    0xe4, 0x0f, 0x65, 0xc8, 0xd0, 0x7b, 0x65, 0xc8, 0xd0, 0x5b, 
+    0x9a, 0x37, 0x2f, 0xa4, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0xd0, 0xce, 0xb3, 0x2b, 
+    0x6c, 0xd6, 0x8b, 0x4e, 0xf2, 0xab, 0xeb, 0x05, 0xb0, 0xe7, 
+    0xcc, 0xd2, 0x15, 0xfe, 0x97, 0x4d, 0x86, 0xbe, 0xdf, 0xcd, 
+    0x86, 0xbe, 0xcf, 0xcd, 0x79, 0x41, 0x30, 0x32, 0x9c, 0xa9, 
+    0x61, 0x00, 0x3e, 0x63, 0xb4, 0x5b, 0x38, 0xf2, 0x36, 0x88, 
+    0x07, 0xb6, 0x64, 0x45, 0x92, 0xe4, 0x4d, 0xa2, 0x00, 0xb7, 
+    0x4d, 0x2c, 0x00, 0xb7, 0x4d, 0xac, 0xff, 0x48, 0xb2, 0x53, 
+    0xcf, 0x72, 0x93, 0xb4, 0x91, 0xbc, 0x1b, 0x03, 0x01, 0x7a, 
+    0xf9, 0x37, 0x06, 0xcf, 0x5c, 0x8e, 0x2f, 0xc9, 0x76, 0xb3, 
+    0x18, 0x7a, 0x76, 0xfc, 0x18, 0xba, 0x76, 0xbe, 0xe7, 0x05, 
+    0x89, 0x41, 0x6a, 0x72, 0x3f, 0xb8, 0x16, 0xfd, 0xf3, 0xdc, 
+    0xe4, 0x97, 0xbc, 0xcc, 0x12, 0x9b, 0x07, 0x02, 0xaf, 0xd9, 
+    0xdb, 0x92, 0x8d, 0x59, 0xc7, 0x06, 0x8d, 0xd9, 0xc7, 0x06, 
+    0x72, 0x26, 0x38, 0xf9, 0xcd, 0x73, 0xe4, 0x35, 0x7a, 0x80, 
+    0xf9, 0xfb, 0x09, 0x53, 0xe6, 0x44, 0x54, 0x9c, 0x24, 0x13, 
+    0xc8, 0xe3, 0x82, 0xe1, 0x48, 0xca, 0x94, 0x63, 0x48, 0xca, 
+    0x84, 0x63, 0xb7, 0x35, 0x7b, 0x9c, 0x41, 0xbb, 0x1b, 0xf7, 
+    0x2c, 0xfe, 0x1e, 0x13, 0x65, 0x39, 0xde, 0x26, 0x46, 0x50, 
+    0xf2, 0xa1, 0x00, 0x0b, 0xd5, 0x79, 0x54, 0x43, 0x15, 0x2a, 
+    0x44, 0x43, 0x15, 0x6b, 0xbb, 0xbc, 0xea, 0x94, 0xa3, 0x44, 
+    0xb6, 0xc5, 0x53, 0xaa, 0xcd, 0xa0, 0xe3, 0x06, 0xeb, 0x20, 
+    0x73, 0xc4, 0x45, 0x0d, 0x1a, 0xba, 0xda, 0x20, 0xd3, 0xe2, 
+    0xba, 0x64, 0xd3, 0xe2, 0xba, 0x20, 0x2c, 0x1d, 0x45, 0xdf, 
+    0x34, 0x47, 0x05, 0x41, 0x1c, 0xb3, 0x04, 0x02, 0x19, 0xa1, 
+    0xa4, 0x04, 0x7f, 0x75, 0xc3, 0xc7, 0xf9, 0xaf, 0xac, 0x8f, 
+    0xd9, 0xf3, 0x4f, 0xb7, 0xd9, 0xf3, 0x4f, 0x97, 0x26, 0x0c, 
+    0xb0, 0x68, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x13, 0xb9, 0xd8, 0x0b, 0xa9, 0xaf, 
+    0x83, 0x19, 0xb5, 0xea, 0x75, 0x81, 0xb1, 0xd3, 0x52, 0x95, 
+    0x60, 0x4b, 0x18, 0xde, 0x74, 0x9f, 0x98, 0x9d, 0x74, 0xdf, 
+    0x18, 0x9d, 0x8b, 0x20, 0xe7, 0x62, 0x82, 0x65, 0xb3, 0x18, 
+    0x0c, 0xa4, 0x3b, 0xfb, 0x39, 0xff, 0xe7, 0x47, 0x7b, 0x1d, 
+    0xb6, 0xfe, 0x7b, 0xb3, 0xa6, 0xda, 0x53, 0xb5, 0x66, 0xdb, 
+    0x5b, 0xb5, 0x66, 0xdb, 0xa4, 0x4a, 0x99, 0x24, 0x01, 0x3f, 
+    0xf3, 0xc0, 0x60, 0x70, 0x14, 0xd2, 0xe0, 0xf8, 0xc4, 0x31, 
+    0xc9, 0x20, 0xc1, 0x69, 0x35, 0x30, 0x56, 0xc9, 0x84, 0x28, 
+    0x4d, 0x4a, 0x8c, 0x28, 0x4d, 0x4a, 0x73, 0xd7, 0xb2, 0xb5, 
+    0x3f, 0x6e, 0x45, 0xd9, 0xa0, 0x54, 0x23, 0x06, 0x23, 0xcd, 
+    0x8f, 0x76, 0x39, 0x7d, 0xa7, 0x4d, 0x25, 0x90, 0x1e, 0x6b, 
+    0x21, 0xdf, 0xf3, 0x29, 0x21, 0xd7, 0xfb, 0x69, 0xde, 0x28, 
+    0x04, 0x96, 0x87, 0x17, 0xd5, 0x68, 0x9c, 0x09, 0x04, 0xce, 
+    0x1c, 0xc2, 0x2e, 0xed, 0xe5, 0xe2, 0xf5, 0x61, 0x52, 0x34, 
+    0x4f, 0x27, 0x5a, 0x66, 0x43, 0x21, 0x7a, 0x66, 0x43, 0x21, 
+    0x85, 0x99, 0xbc, 0xde, 0xb3, 0x1a, 0x43, 0x87, 0x0a, 0xd3, 
+    0xda, 0xc4, 0xa9, 0xfd, 0x54, 0xd3, 0x60, 0xa0, 0x2c, 0x41, 
+    0x75, 0x74, 0x20, 0x25, 0x68, 0x68, 0x2c, 0x03, 0x68, 0x60, 
+    0x2c, 0x03, 0x97, 0x9f, 0xd3, 0xfc, 0xec, 0x2a, 0x7c, 0x2e, 
+    0x7b, 0x70, 0x1b, 0x98, 0xc6, 0xa7, 0xd2, 0x32, 0xbd, 0x35, 
+    0xff, 0x14, 0x27, 0x7d, 0x1a, 0x50, 0x8e, 0x61, 0x3e, 0x50, 
+    0x8f, 0x61, 0x3e, 0x50, 0x70, 0x9e, 0xc1, 0xaf, 0x06, 0x1f, 
+    0xe1, 0xf8, 0x97, 0xab, 0x39, 0x0c, 0x51, 0x48, 0x88, 0x2a, 
+    0xa4, 0xdf, 0x5c, 0x42, 0x5e, 0xee, 0x8a, 0xec, 0x5c, 0xef, 
+    0xcc, 0xdc, 0x5c, 0xef, 0xcc, 0xdc, 0xa3, 0x10, 0x33, 0x23, 
+    0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 
+    0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 
+    0x00, 0x00
+};
+
diff --git a/parallel_bare_tests/Dijkstra/Dijkstra.c b/parallel_bare_tests/Dijkstra/Dijkstra.c
index a6f6cab..8fe903d 100644
--- a/parallel_bare_tests/Dijkstra/Dijkstra.c
+++ b/parallel_bare_tests/Dijkstra/Dijkstra.c
@@ -96,8 +96,8 @@ int main()
 #endif
 
     if ( hc == 2 && id == 0 ) {
-      reset_timer();
-      start_timer();
+      reset_timer(rt_cluster_id());
+      start_timer(rt_cluster_id());
     }
 
     synch_barrier();
@@ -105,8 +105,8 @@ int main()
     dijkstra_distance(mind, ohd);
 
     if ( hc == 2 && id == 0 ) {
-      stop_timer();
-      time = get_time();
+      stop_timer(rt_cluster_id());
+      time = get_time(rt_cluster_id());
     }
 #ifdef PROFILE
     // stop performance counters
diff --git a/parallel_bare_tests/LU/LU.c b/parallel_bare_tests/LU/LU.c
index 3700837..d6ddc81 100644
--- a/parallel_bare_tests/LU/LU.c
+++ b/parallel_bare_tests/LU/LU.c
@@ -174,8 +174,8 @@ int main(int argc, char **argv)
   for(hc = 0; hc < 3; ++hc) {
 
     if ( hc == 2 && id == 0 ) {
-      reset_timer();
-      start_timer();
+      reset_timer(rt_cluster_id());
+      start_timer(rt_cluster_id());
     }
     perf_reset();
     perf_start();
@@ -185,8 +185,8 @@ int main(int argc, char **argv)
     factor(G, N, N, pivots);
 
     if ( hc == 2 && id == 0 ) {
-      stop_timer();
-      time = get_time();
+      stop_timer(rt_cluster_id());
+      time = get_time(rt_cluster_id());
     }
     perf_stop();
 
diff --git a/parallel_bare_tests/conv16/conv16.c b/parallel_bare_tests/conv16/conv16.c
index c4d95a2..38c3a92 100644
--- a/parallel_bare_tests/conv16/conv16.c
+++ b/parallel_bare_tests/conv16/conv16.c
@@ -89,10 +89,10 @@ int test_singlethread(void (*test)(int16_t *, int16_t *, int16_t *, int, int, in
    if(rt_core_id() == 0) {
       load();
 
-      reset_timer();
-      start_timer();
+      reset_timer(rt_cluster_id());
+      start_timer(rt_cluster_id());
       test(g_W, g_x, g_y, IH, IW, FH, FW, OH, OW, 1, 0, 0);
-      stop_timer();
+      stop_timer(rt_cluster_id());
 
       #ifdef CHECK_CHECKSUM
       errors = 0;
@@ -110,7 +110,7 @@ int test_singlethread(void (*test)(int16_t *, int16_t *, int16_t *, int, int, in
       #endif
 
       #ifndef PULP_SPI
-      printf("%s, errors=%d, time=%d\n", str, errors, get_time());
+      printf("%s, errors=%d, time=%d\n", str, errors, get_time(rt_cluster_id()));
       #endif
       
    }
@@ -129,12 +129,12 @@ int test_multithread(void (*test)(int16_t *, int16_t *, int16_t *, int, int, int
    synch_barrier();
 
    if(rt_core_id() == 0) {
-      reset_timer();
-      start_timer();
+      reset_timer(rt_cluster_id());
+      start_timer(rt_cluster_id());
    }
    test(g_W, g_x, g_y, IH, IW, FH, FW, OH, OW, 1, 0, 0);
    if(rt_core_id() == 0) {
-      stop_timer();
+      stop_timer(rt_cluster_id());
 
       #ifdef CHECK_CHECKSUM
       errors = 0;
@@ -152,7 +152,7 @@ int test_multithread(void (*test)(int16_t *, int16_t *, int16_t *, int, int, int
       #endif
 
       #ifndef PULP_SPI
-      printf("%s, errors=%d, time=%d\n", str, errors, get_time());
+      printf("%s, errors=%d, time=%d\n", str, errors, get_time(rt_cluster_id()));
       #endif
       
    }
diff --git a/parallel_bare_tests/parMatrixMul/parMatrixMul.c b/parallel_bare_tests/parMatrixMul/parMatrixMul.c
index 78a3239..4ddd6aa 100644
--- a/parallel_bare_tests/parMatrixMul/parMatrixMul.c
+++ b/parallel_bare_tests/parMatrixMul/parMatrixMul.c
@@ -57,9 +57,6 @@ testcase_t testcases[] = {
 
 int main() {
 
-  if (rt_cluster_id() != 0)
-    return bench_cluster_forward(0);
-
   int nbErrors = run_suite(testcases);
 
   synch_barrier();
@@ -85,9 +82,9 @@ void matrix_multiplication(testresult_t *result, void (*start)(), void (*stop)()
   lb = coreid * chunk;
   //upper bound
   ub = lb + chunk;  
-  
+
   synch_barrier();
-  
+
   /********************* Benchmark Execution *********************/
   if (coreid<numcores) {
     start();
diff --git a/parallel_bare_tests/parMatrixMul/test.ini~ b/parallel_bare_tests/parMatrixMul/test.ini~
deleted file mode 100644
index f0cdaac..0000000
--- a/parallel_bare_tests/parMatrixMul/test.ini~
+++ /dev/null
@@ -1,6 +0,0 @@
-[test:parMatrixMul]
-configs=bitcode=0_omp=0_ovx=0
-command.all=make -f Makefile.sdk systemConfig=%(config)s clean all run
-timeout=5
-probe[totalCycles].metric=cycles
-probe[totalCycles].regexp=execution time: (\d+)
diff --git a/parallel_bare_tests/parMatrixMul32/gen_stimuli.py b/parallel_bare_tests/parMatrixMul32/gen_stimuli.py
index d9d9697..aac24be 100755
--- a/parallel_bare_tests/parMatrixMul32/gen_stimuli.py
+++ b/parallel_bare_tests/parMatrixMul32/gen_stimuli.py
@@ -15,7 +15,7 @@ def write_arr(f, name, arr):
 f = open('parMatrixMul32_stimuli.h', 'w')
 
 
-SIZE = 32 
+SIZE = 32
 RANGE = 2**15/SIZE
 
 m_a   = []
diff --git a/parallel_bare_tests/parMatrixMul32/parMatrixMul32_stimuli.h b/parallel_bare_tests/parMatrixMul32/parMatrixMul32_stimuli.h
index dc0adbd..e313733 100644
--- a/parallel_bare_tests/parMatrixMul32/parMatrixMul32_stimuli.h
+++ b/parallel_bare_tests/parMatrixMul32/parMatrixMul32_stimuli.h
@@ -1,3101 +1,3082 @@
-/*
- * Copyright (C) 2018 ETH Zurich and University of Bologna
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/* 
- * Mantainer: Luca Valente, luca.valente2@unibo.it
- */
 const int m_a[] = {
-830,
--456,
--926,
--662,
-135,
--922,
--509,
-995,
-818,
-595,
-135,
--639,
--371,
--234,
-842,
--1020,
--130,
-949,
--174,
-137,
--916,
-179,
-576,
--253,
-133,
-7,
--121,
-920,
-775,
--22,
-7,
-138,
-445,
--308,
-283,
-657,
-33,
--482,
-948,
--352,
-471,
--40,
-498,
-728,
--399,
--636,
-50,
-796,
--35,
-1011,
--970,
--619,
-411,
--959,
--152,
-40,
-624,
-1016,
-10,
--631,
--833,
-621,
--795,
--66,
-222,
-872,
--149,
--952,
--71,
-404,
--373,
-811,
-267,
-763,
--662,
--953,
-663,
--381,
-724,
-35,
-141,
-969,
-741,
-204,
--653,
--920,
--661,
--770,
--237,
-309,
-877,
--969,
--372,
--663,
--460,
-711,
-663,
--517,
--241,
-704,
--557,
--791,
-852,
--248,
-640,
--454,
--995,
--362,
-99,
-465,
--974,
-546,
--192,
-411,
-399,
--653,
-367,
--964,
--890,
--617,
--322,
--916,
--181,
--781,
--892,
-496,
--376,
-996,
--446,
-490,
-397,
--31,
--233,
--293,
-568,
--819,
--668,
--352,
-917,
-920,
-581,
--533,
-437,
--243,
-465,
-230,
-729,
--509,
-259,
--885,
-206,
--854,
-520,
-240,
-707,
--652,
--342,
--574,
-467,
-524,
-698,
--660,
-947,
--1,
--908,
-220,
-988,
--803,
--532,
-474,
-168,
--124,
--210,
-454,
-886,
--893,
--285,
--1007,
--866,
-263,
-369,
-885,
--24,
--1015,
-496,
--283,
-222,
-199,
--698,
-279,
-68,
--229,
-941,
-677,
--806,
--913,
-558,
-652,
-629,
--338,
--959,
-468,
--336,
--351,
--789,
--665,
-716,
--582,
--508,
--489,
--119,
-717,
-269,
--322,
-68,
-400,
-880,
-4,
--33,
-384,
--537,
-720,
--653,
-269,
--492,
-281,
--873,
--224,
--289,
-466,
--202,
--105,
--223,
-137,
--283,
--796,
-105,
--528,
--107,
--640,
-507,
-481,
-572,
--411,
-720,
-163,
--537,
--698,
--154,
-88,
-765,
-701,
-925,
--634,
-751,
-304,
--837,
-2,
--691,
-307,
-626,
--717,
--858,
--441,
-79,
-515,
-783,
-250,
-946,
-934,
-151,
--385,
--788,
--102,
--818,
--422,
--728,
-739,
-822,
-254,
--330,
-836,
--200,
-733,
--537,
--724,
--824,
--472,
-322,
--995,
--763,
--214,
-774,
-713,
--266,
-301,
-699,
-458,
--1004,
-301,
-425,
-386,
-787,
-436,
-60,
-614,
--162,
--900,
-657,
--970,
-495,
-830,
--620,
--411,
--982,
-842,
-502,
-44,
--78,
-740,
-699,
--307,
-894,
-915,
-524,
-731,
-687,
-769,
--718,
-825,
-508,
-952,
--68,
-733,
-305,
--505,
-667,
-675,
-855,
-221,
--510,
--164,
-731,
--497,
-247,
-37,
--269,
-858,
--664,
--523,
--449,
-772,
-205,
--350,
--286,
--871,
-448,
--91,
-350,
--182,
-233,
--870,
-884,
-885,
-826,
-532,
-429,
-591,
--238,
--435,
-71,
-661,
-192,
-982,
--937,
-282,
-939,
-107,
--648,
--237,
-212,
-748,
-912,
--76,
--20,
--304,
--175,
--390,
--878,
--360,
--476,
-439,
-558,
-552,
--510,
--120,
--106,
--959,
--575,
-505,
--829,
-16,
-734,
--882,
-848,
-864,
--653,
-905,
-751,
--304,
-64,
--262,
-715,
-281,
-585,
-65,
-100,
--579,
-678,
-425,
-883,
--198,
-482,
--508,
-477,
-1010,
-783,
--36,
-737,
--702,
--171,
-59,
-773,
--348,
--573,
--275,
--415,
--967,
-468,
--544,
--655,
--200,
--255,
--598,
--950,
-229,
-790,
-892,
--795,
--71,
--693,
--615,
-562,
--326,
--545,
-226,
-255,
-840,
--202,
--160,
-891,
-791,
--669,
--195,
--733,
-372,
-937,
-208,
-970,
-814,
-45,
-57,
-943,
-439,
-763,
--132,
-83,
--324,
-857,
--846,
-810,
--832,
-251,
--866,
--129,
--118,
-574,
-102,
--141,
-342,
--478,
-733,
-611,
--843,
-975,
-270,
--638,
--692,
--618,
-388,
--434,
-488,
--1009,
--237,
-566,
-925,
--596,
-640,
-378,
--30,
--267,
-299,
--260,
--134,
-615,
--628,
--966,
--436,
--111,
-34,
--4,
-687,
-133,
--666,
-167,
--1012,
-422,
-361,
-604,
-836,
-190,
--939,
-564,
-137,
--563,
-397,
-692,
-934,
-907,
--182,
--539,
-974,
-176,
-988,
-24,
-12,
--739,
--236,
--851,
--509,
-814,
--523,
-654,
--350,
--320,
--395,
--863,
-130,
--725,
-117,
--788,
--822,
--909,
-812,
-534,
--232,
-1009,
--583,
--455,
--257,
-791,
-971,
-53,
-837,
-423,
-527,
-889,
--378,
--39,
--572,
-713,
--588,
-862,
--608,
-364,
-46,
-476,
-928,
-392,
+921,
+223,
+-983,
+-567,
+100,
+801,
+-520,
+-758,
+288,
 -819,
--805,
--377,
--684,
-234,
-263,
--245,
--461,
-23,
-226,
-504,
--167,
-74,
-794,
-1022,
-464,
-538,
--713,
-748,
-332,
-246,
--969,
--293,
-77,
-43,
--940,
--977,
--170,
--203,
-274,
--20,
--467,
--820,
-315,
-461,
-291,
-953,
-251,
--885,
-1001,
-413,
-591,
-684,
-698,
--701,
-553,
--808,
--799,
--42,
--148,
--512,
--71,
-742,
-193,
--778,
-38,
-14,
--360,
--172,
-461,
-94,
-873,
-777,
--516,
--236,
-487,
-923,
--148,
--611,
--555,
-457,
-600,
-867,
--135,
--487,
-221,
-563,
--847,
--636,
--237,
--735,
--994,
-281,
-862,
-506,
--996,
--613,
--551,
--606,
-733,
--338,
--643,
-536,
-513,
-285,
+627,
+-331,
+-324,
+154,
+389,
+975,
+789,
+542,
 497,
-99,
--439,
--258,
--609,
--712,
-740,
--200,
-910,
--845,
-833,
--242,
--637,
--780,
--137,
+-213,
+-192,
+216,
+575,
+-490,
+-826,
+-93,
+161,
+143,
+866,
+722,
+-911,
+-567,
+-484,
+907,
+593,
+-261,
+454,
+-630,
+710,
 -155,
-263,
--895,
-877,
--95,
-453,
-777,
--196,
-576,
+210,
+-388,
+-207,
+-82,
 93,
--556,
--71,
--241,
-987,
-1012,
+512,
+-745,
+460,
+-914,
+1011,
+310,
+740,
+-465,
+-971,
+-51,
+392,
+378,
+846,
+-75,
+312,
+891,
+15,
+165,
+-238,
+138,
+295,
+-426,
+-341,
+105,
 117,
--149,
--940,
--790,
-330,
--559,
--170,
-136,
--26,
--227,
-153,
--937,
-161,
--954,
--996,
--725,
+61,
+1022,
+-93,
+-491,
+-367,
+422,
+560,
+698,
+-271,
+-917,
+-537,
+610,
+-525,
+456,
+-753,
+966,
+112,
+569,
+791,
+269,
+486,
+763,
+-793,
+769,
+-446,
+1001,
+-639,
+-621,
+-228,
+-207,
+22,
+-582,
+-120,
+963,
 242,
--369,
-672,
-474,
--181,
-856,
-863,
--419,
--279,
--1009,
-8,
--695,
--834,
-437,
--611,
+-387,
+-601,
+-821,
+-403,
+-149,
+-339,
+-678,
+-151,
+-578,
+-916,
+-688,
+716,
+-433,
+942,
+662,
+-767,
+-175,
+-470,
+135,
+521,
+583,
+719,
+-200,
+649,
+740,
+-280,
+1001,
+673,
+668,
+358,
+-568,
+282,
+-558,
+-268,
+-224,
+-909,
+462,
+-757,
+-129,
+-162,
+116,
+683,
+645,
+-679,
+-414,
+688,
+526,
+-247,
+281,
+-174,
+785,
+-545,
+119,
+-475,
+-38,
+697,
+790,
+-230,
+738,
+344,
+-387,
+753,
+132,
+306,
+949,
+577,
+-201,
+198,
+-1021,
+-407,
+465,
 -34,
--97,
-637,
-17,
--208,
-288,
+493,
+510,
+-800,
+40,
+354,
+-793,
+1014,
+89,
+108,
+415,
+428,
+544,
+449,
+-117,
+416,
+-309,
+246,
+272,
+248,
+295,
+-835,
+941,
+-453,
+475,
+627,
+-931,
+69,
+429,
+931,
+576,
+-193,
+-311,
+-669,
+-588,
 -826,
--29,
--828,
-777,
--490,
-240,
-244,
--436,
--636,
--413,
--730,
--426,
--547,
--135,
-453,
-930,
-357,
--926,
--448,
-96,
--672,
--1003,
-644,
-172,
-938,
-189,
--892,
--628,
--806,
-673,
--430,
-684,
-291,
-625,
-161,
--935,
--275,
-840,
-944,
-886,
--28,
-348,
--947,
-431,
--96,
--638,
-500,
--163,
-1022,
--775,
--641,
--42,
--90,
--192,
-568,
--845,
-748,
--95,
--625,
--426,
-656,
--48,
-895,
+996,
+-771,
+-863,
+-97,
+-58,
+-676,
+-917,
+218,
+-514,
+-85,
+-964,
+-831,
+-280,
+1015,
+-891,
+916,
+-86,
+683,
+-725,
+766,
+87,
+-827,
+-189,
+751,
+139,
+-38,
+-349,
+534,
+-523,
+-938,
+-251,
 791,
+-285,
+13,
+-93,
+619,
+-771,
+487,
+166,
+-1008,
+-858,
+-722,
+483,
+779,
+676,
+-941,
+758,
+-234,
+-7,
+67,
+-782,
+944,
+-289,
 1009,
--830,
+27,
+-311,
+-1002,
+-917,
+-494,
+-419,
+-475,
+450,
+-179,
 -442,
-406,
-771,
-941,
-337,
-603,
--601,
-712,
-844,
--611,
--649,
+126,
+-244,
+-613,
+930,
+-1012,
+1016,
+-1011,
+432,
+765,
+-793,
+132,
+-147,
+510,
+-781,
+334,
+653,
+969,
+39,
+268,
+-748,
+-162,
+967,
+-732,
+-637,
+467,
+14,
+-320,
+607,
+-569,
+79,
+103,
+-549,
+-75,
+514,
+850,
+-401,
+-804,
+830,
+-1013,
+835,
+488,
+345,
+-92,
+-924,
+468,
+245,
+-34,
+-669,
+-79,
+787,
+515,
+773,
+418,
 -163,
--911,
-833,
-284,
--591,
--574,
--653,
--454,
--731,
--784,
--511,
--1005,
-831,
-900,
--878,
-544,
--592,
-717,
-722,
-460,
--76,
--572,
-726,
-306,
-990,
-931,
-320,
--513,
--920,
--578,
-149,
-115,
--961,
-584,
-796,
-577,
--850,
-973,
--661,
--935,
--249,
+718,
+355,
+472,
+730,
+808,
+-122,
+-461,
+878,
+95,
+-198,
+373,
+-385,
+-688,
+984,
+-492,
+338,
+-842,
+-913,
+-198,
+-704,
+80,
+693,
+-659,
+452,
+880,
+-1020,
+-357,
+185,
+384,
+-762,
+312,
+-182,
+-627,
+1020,
+-393,
+664,
+92,
+962,
+516,
+-239,
+-38,
+924,
+355,
+-568,
+-787,
+-794,
+805,
+-725,
+668,
+-55,
+213,
+766,
+1003,
+-594,
+-21,
+-852,
+950,
+250,
+214,
+-538,
+-5,
+-657,
+-918,
+251,
+-340,
+-15,
+893,
+806,
+455,
+-479,
+928,
+748,
 134,
--942,
--19,
-269,
-972,
--810,
--678,
-976,
+615,
+-735,
+-269,
+-984,
+-4,
+256,
+-37,
+910,
+12,
 991,
-488,
--300,
--295,
-573,
-403,
+-955,
+-593,
+-64,
+133,
+562,
+-297,
+-788,
+-542,
+-776,
+534,
+-1018,
+40,
+823,
+157,
+-538,
+78,
+-439,
+-951,
+581,
+45,
+645,
+76,
+-542,
+101,
+554,
+849,
+-164,
+-122,
+736,
+226,
+147,
+-131,
+-759,
+128,
+-148,
+738,
+-455,
 590,
--518,
-341,
-135,
-710,
--381,
-807,
-668,
--300,
-828,
--205,
--710,
-498,
+801,
+35,
+933,
+30,
+-369,
+-653,
+-268,
+603,
+-579,
+956,
+-664,
+358,
+92,
+935,
+241,
+-597,
+106,
+561,
+-819,
+-627,
+994,
+857,
+998,
+-975,
+-168,
+407,
+524,
+-713,
+-104,
+467,
+-223,
+870,
+-958,
+338,
+-440,
+688,
 -786,
--657,
+-799,
+550,
+924,
+418,
+-809,
+1011,
+-863,
+451,
+741,
+-163,
+-507,
+-744,
+-234,
+367,
+-223,
 152,
-413,
-97,
-261,
-784,
-289,
--936,
--858,
-119,
--72,
-702,
--2,
--437,
-147,
--968,
--932,
--463,
--880,
-938,
--574,
--177,
+107,
+73,
+409,
+582,
+-524,
+-109,
+522,
+-205,
+-163,
+904,
+-238,
+-746,
+-832,
+1007,
+-326,
+641,
+-1010,
+549,
+840,
+-412,
+175,
+253,
+-121,
+213,
+877,
+-708,
+-501,
+459,
+-294,
+-456,
+-48,
+481,
+-628,
+432,
+-578,
+-487,
+-835,
+322,
+205,
+-269,
+313,
+-351,
+-361,
+-323,
+-555,
+-622,
+299,
+393,
+-710,
+411,
+328,
+708,
+167,
+677,
+-873,
+-468,
+922,
+-462,
+-536,
+-632,
+-135,
+-498,
+402,
 -932,
+-851,
+-373,
+-201,
+924,
+587,
+-840,
+-814,
+188,
+805,
+528,
+35,
+-165,
+-943,
+494,
+-276,
+-816,
+597,
+967,
+-997,
+-595,
+433,
+-750,
+-988,
+774,
+-45,
+-761,
+881,
+-634,
+-113,
+977,
+-457,
+-235,
+18,
+-125,
+971,
+254,
+57,
+-520,
+120,
+-720,
+97,
+-752,
+1000,
+-911,
 521,
-333,
-713,
--855,
--549,
-585,
--672,
-29,
-739,
-115,
--875,
--140,
--812,
+218,
+-735,
+-555,
+196,
+635,
+-819,
+-492,
+-271,
+-378,
+325,
+458,
+-618,
+-123,
+-126,
+161,
+-742,
+-439,
+-297,
+737,
 -30,
--898,
--528,
--1012,
--534,
-585,
--419,
-273,
+1001,
+163,
+975,
+654,
+-902,
+-255,
+578,
+-593,
+658,
+-227,
+-854,
+89,
+-326,
+-274,
+513,
+161,
+244,
+129,
+826,
+39,
+-720,
+-56,
+72,
+883,
+187,
+597,
 -551,
--452,
-642,
-836,
-881,
-235,
--450,
--1009,
--480,
--508,
-844,
-755,
--851,
+205,
+134,
+-792,
+-267,
+205,
+9,
+-98,
+927,
+822,
+-906,
+165,
+-41,
+276,
+783,
+-866,
+-172,
+815,
+54,
+496,
+491,
+-365,
+-588,
+-353,
+-617,
+123,
+639,
+-401,
+475,
+2,
+-938,
+777,
+570,
+184,
+-583,
+-466,
+-66,
+-467,
+121,
+792,
+889,
+-1017,
+442,
+649,
+-794,
+203,
+456,
+738,
+-792,
+767,
+359,
+-110,
+1021,
+227,
+463,
+440,
+-1019,
+817,
+616,
+266,
+-826,
+730,
+343,
+498,
+874,
+567,
+-74,
+-286,
+-665,
+-30,
+-74,
+-434,
+-694,
+901,
+-956,
+339,
+483,
+325,
+-455,
+-91,
 -28,
--363,
--727,
-590,
--111,
--764,
--364,
--167,
--409,
-291,
--253,
--718,
--943,
--47,
-755,
+-116,
+-446,
+631,
+230,
+538,
+81,
+-139,
+-995,
+-577,
+-695,
+976,
+-866,
+-270,
+334,
+334,
+-756,
+-859,
+194,
+77,
+-89,
+-483,
+906,
+26,
+-507,
+-196,
+-195,
+-929,
+713,
+744,
+653,
+533,
+-531,
 316,
--1012,
--456,
-212,
-19,
-358,
--732,
-665,
-659,
-445,
-706,
--723,
--471,
-266,
-740,
--207,
--164,
-446,
-418,
-482,
-986,
--382,
-126,
--920,
-260,
--488,
-929,
-515,
-969,
--725,
--57,
--106,
--146,
--454,
-58,
-706,
--297,
--233,
-262,
--749,
--498,
-660,
--657,
--279,
-136,
-401,
--330,
-569,
--159,
-774,
-816,
-484,
--333,
-642,
--460,
-};
-
-const int m_b[] = {
-739,
--691,
--674,
--272,
-816,
+-821,
+-524,
+-311,
+381,
+546,
+-547,
 -147,
+815,
+-792,
+315,
+543,
+-87,
+-645,
+-641,
+-494,
+127,
+-430,
+50,
+594,
+-44,
+-727,
+-823,
+906,
+-661,
+-670,
+-993,
+-129,
+116,
+639,
+549,
+-251,
+-610,
+413,
+-575,
+40,
+976,
+-50,
+-553,
+-771,
+60,
+-671,
+-860,
+803,
+521,
+594,
+-526,
+-195,
+-82,
+316,
+934,
+425,
+-694,
+917,
+-25,
+749,
+139,
+-428,
+-445,
+510,
+-144,
+115,
+-774,
+779,
+-171,
+-475,
+208,
 760,
--56,
-581,
--550,
-763,
-539,
-251,
--2,
--701,
-579,
--53,
--241,
-699,
-831,
-752,
-505,
-715,
--131,
-24,
+27,
+365,
+442,
+264,
+-866,
+-40,
+-498,
+-1019,
+781,
+-187,
+-668,
+844,
+39,
+-118,
+340,
+128,
+-235,
+-584,
+-883,
+-857,
+-562,
+405,
+-413,
+800,
+-74,
+542,
+645,
+882,
+-416,
+1020,
+481,
+461,
+-108,
+-309,
+80,
+435,
+-522,
+-572,
+815,
+-450,
+612,
+-648,
+123,
+-754,
+37,
+941,
+-518,
+980,
+205,
+-596,
+411,
+55,
+167,
+0,
+-387,
 -268,
-778,
-921,
-368,
--430,
--711,
--781,
-678,
--746,
--755,
-222,
-477,
--568,
-140,
-201,
--696,
--992,
--201,
-964,
--784,
--532,
+176,
+635,
+-677,
+-408,
+-728,
+691,
+534,
+229,
+378,
+-407,
+-406,
+833,
+463,
+861,
+656,
+221,
+-247,
+-791,
+-152,
+-934,
+-548,
+822,
+-957,
+478,
+-471,
+310,
+-965,
+-197,
+269,
+636,
+185,
+82,
+-420,
+-945,
+949,
+-221,
+636,
+-20,
+869,
+-634,
+-150,
+820,
+-1001,
+973,
+594,
+654,
+324,
+958,
+363,
+-804,
+-244,
+-278,
 -863,
-1014,
--642,
-6,
-589,
-924,
-662,
-238,
-485,
-859,
-808,
+-927,
 841,
--192,
-91,
-42,
-45,
-355,
-293,
--182,
-34,
--883,
-915,
--414,
-736,
-613,
-418,
--202,
-437,
--229,
+947,
+513,
+-325,
+-875,
+-187,
+-827,
+-261,
 -564,
--250,
--125,
--233,
--397,
-331,
-447,
--41,
-975,
-853,
-573,
--475,
--79,
--932,
-37,
--367,
--244,
-300,
-5,
--188,
-595,
-934,
-122,
--76,
--446,
-703,
--68,
-214,
-963,
--160,
-529,
-98,
-616,
-161,
-476,
--808,
-262,
-49,
-646,
--647,
--835,
--706,
-697,
--26,
--858,
-496,
-452,
-875,
-479,
-73,
--387,
-512,
-85,
-58,
-420,
+367,
+-496,
+-427,
 -167,
-708,
--847,
-408,
--254,
-62,
-700,
--576,
--850,
-18,
-186,
--940,
-519,
-906,
-559,
-480,
--1022,
--393,
--726,
+-411,
+144,
+-445,
+-990,
+-1019,
+332,
+-595,
+334,
+-170,
+718,
+855,
 -219,
-204,
--231,
+70,
+-538,
+-1011,
+-353,
+-856,
+-320,
+-732,
+914,
+203,
+121,
+-111,
+-336,
 899,
--998,
-350,
--795,
-180,
--36,
-261,
-315,
--565,
--707,
-864,
--279,
-859,
--993,
-37,
--841,
-287,
+732,
+-446,
+145,
 -613,
-693,
-626,
-887,
--781,
--779,
-415,
-610,
-759,
--655,
--999,
-521,
-569,
-726,
-761,
-330,
--513,
-797,
-975,
--883,
--718,
--748,
--237,
--892,
-942,
-715,
-63,
--272,
--368,
-738,
-177,
-267,
--703,
-991,
-1002,
--916,
-143,
-879,
--233,
--478,
-66,
-39,
--516,
-803,
--341,
-997,
-770,
--810,
--577,
--244,
-1005,
-190,
-593,
-530,
--243,
--41,
--801,
-16,
-45,
--834,
--722,
--265,
--64,
-1011,
--397,
--88,
--228,
-179,
-22,
--295,
--296,
-549,
-373,
+799,
+-159,
+54,
+290,
+740,
+-33,
 -502,
-1022,
-29,
-824,
--523,
-969,
--973,
--402,
--955,
--119,
--294,
+-498,
+118,
+962,
+589,
+-483,
+325,
+-260,
+219,
+919,
+-699,
+571,
+-553,
+-1020,
+-541,
+770,
+589,
+837,
+-193,
+378,
+898,
+-869,
+644,
+-461,
+998,
+-942,
+};
+
+const int m_b[] = {
+716,
+-76,
+15,
+81,
+585,
+711,
+-519,
+-845,
+831,
+-931,
+-29,
+-942,
+645,
+-480,
+-77,
+-969,
+853,
+-987,
+944,
+828,
+-610,
+66,
+405,
+832,
+-130,
+-817,
+842,
+549,
+-47,
+981,
+-271,
+842,
+890,
+510,
+488,
+754,
+454,
+-114,
+-266,
+-720,
+-45,
+-79,
+-114,
+-394,
+-538,
+-229,
+-873,
+975,
+550,
+656,
+-879,
+881,
+-582,
+446,
+-332,
+734,
+545,
+349,
+703,
+462,
+-677,
+-934,
+700,
+298,
+-978,
 489,
--905,
-806,
--864,
--41,
--735,
--15,
--338,
-990,
--294,
-403,
-485,
-248,
-468,
-389,
--735,
+-182,
+-1014,
+139,
+-289,
+792,
+362,
+605,
+537,
+-598,
+-112,
+-576,
+404,
+138,
+659,
+888,
+-183,
+-788,
+98,
+-1023,
+694,
+950,
+-969,
+637,
+799,
+-68,
+631,
+-599,
+405,
+591,
+825,
+-425,
+869,
+-89,
+114,
+-37,
+373,
+535,
+444,
+-295,
+-447,
+475,
+-122,
+-411,
+665,
+418,
+-791,
+883,
+-382,
+641,
+418,
+292,
+784,
 999,
--848,
--263,
-715,
-550,
--750,
--207,
--407,
--77,
--833,
--939,
+-823,
+304,
+293,
+997,
+-159,
+-982,
+-372,
+360,
+469,
+295,
+-449,
+-688,
+-937,
+-80,
+-456,
+620,
+-347,
+-425,
+868,
+-15,
+470,
 117,
-624,
--891,
-13,
-785,
-547,
--898,
-622,
-141,
-95,
--133,
--341,
-880,
-829,
-172,
-28,
-702,
-527,
--186,
--61,
--721,
--121,
--862,
--990,
--94,
--657,
--131,
-429,
-760,
--779,
-504,
-1012,
--17,
-800,
-892,
--163,
-309,
--700,
-20,
-808,
--198,
-0,
-331,
-735,
-259,
-479,
-820,
--450,
-3,
--736,
-391,
-1002,
-81,
--377,
--387,
--160,
--909,
-1014,
-881,
--955,
--206,
--369,
--672,
-2,
-842,
--835,
-427,
--392,
+581,
+-244,
+210,
+-70,
+16,
+-638,
+428,
+-828,
+394,
+-203,
+957,
+-289,
+-647,
+-210,
+1000,
+-820,
+555,
+-478,
+49,
+361,
 -917,
-31,
-712,
--570,
-586,
--59,
-538,
--432,
--470,
-499,
--528,
--969,
-1011,
--101,
+458,
+694,
+-217,
+459,
+-904,
+-126,
+-33,
+189,
+762,
+619,
+870,
+-799,
+620,
+-177,
+-1005,
+-807,
+-422,
+490,
+795,
+1013,
+-641,
+479,
+337,
+-92,
+-260,
+956,
+-329,
+864,
+-673,
+-98,
 -331,
--961,
-515,
--301,
-119,
-116,
--570,
-919,
-961,
-707,
--705,
-998,
-969,
-205,
-113,
-303,
-52,
--952,
--268,
--743,
-282,
-359,
--826,
+425,
+-878,
+863,
+-835,
+-834,
+-250,
+301,
+-4,
+-587,
+85,
+491,
+-313,
+-430,
+-54,
+173,
+-666,
+-794,
+114,
+849,
+24,
 140,
--936,
--900,
--682,
-400,
-379,
-624,
--72,
-264,
--887,
-29,
--1005,
--97,
--168,
--769,
-719,
-450,
-178,
--37,
-783,
+-590,
+568,
+-785,
+732,
+843,
+-73,
+10,
+894,
+-195,
+107,
+482,
+778,
+-708,
+582,
+-889,
+-461,
+571,
+918,
+416,
+-518,
+-24,
+-585,
+-38,
+-535,
+-167,
+-308,
+-793,
+-502,
+267,
+989,
+-1007,
+-802,
+-262,
+-106,
+-270,
+744,
+461,
+-683,
+-12,
+989,
+-591,
+-966,
+-300,
+-342,
+201,
+786,
+481,
+913,
+126,
+877,
+864,
+-622,
+-265,
+951,
+-322,
+417,
+417,
+-103,
+-941,
+-428,
+-273,
+-807,
+-649,
+266,
+658,
+-348,
+35,
+1013,
+1,
+968,
+206,
+-721,
+566,
+-468,
+-653,
+-81,
+-652,
+65,
+-595,
+329,
+759,
+386,
+-225,
 -857,
--237,
--409,
-845,
-914,
--877,
-175,
--717,
-191,
--713,
--365,
--672,
-775,
-516,
--656,
--506,
-989,
-573,
--59,
-523,
--392,
--374,
-891,
-408,
--115,
-488,
--3,
+22,
+-267,
+895,
+973,
 -401,
--2,
--102,
-809,
--284,
-833,
-237,
-60,
--77,
-294,
-431,
-940,
--462,
--309,
--988,
-331,
--197,
-614,
--829,
-1012,
-545,
--708,
-598,
-216,
-297,
-643,
--605,
-975,
+-934,
+-159,
+-637,
+516,
+-245,
+650,
+440,
+-245,
+-325,
+162,
+-372,
+457,
+-65,
+122,
+-851,
+-481,
+-960,
+-409,
+-35,
+-701,
+658,
+637,
+-639,
+270,
+274,
+748,
+-46,
+211,
+-311,
 -878,
-259,
-384,
-550,
--98,
-672,
--452,
-982,
--586,
--301,
-919,
--984,
--646,
+743,
+453,
+-625,
+493,
+977,
+-777,
+160,
+-638,
+759,
+-624,
 -973,
-124,
-790,
-648,
--683,
-330,
--724,
-77,
-647,
-442,
-11,
--790,
-585,
-517,
-434,
--795,
--316,
-50,
-235,
--182,
--355,
-505,
--292,
--298,
--659,
-391,
-571,
--848,
--870,
--869,
--987,
-105,
--786,
--791,
--550,
-448,
-565,
--610,
--910,
-694,
-144,
--50,
-891,
-795,
-691,
-626,
--461,
--354,
--484,
--844,
-582,
-626,
--82,
-646,
-762,
-36,
--671,
-120,
--874,
-685,
-1017,
-722,
-11,
--477,
--19,
-283,
--980,
-976,
--341,
-59,
--495,
--679,
-887,
--890,
--167,
-633,
-412,
--225,
-114,
+-681,
+166,
+-530,
+-677,
+657,
+-104,
+-398,
+-482,
+-669,
+1014,
+707,
 238,
-754,
-685,
--522,
-565,
-983,
-159,
--678,
-202,
--811,
--902,
-950,
-560,
--849,
--588,
-0,
--939,
-199,
--214,
--575,
--703,
--689,
-1019,
-799,
--478,
-435,
-499,
-686,
--1002,
-343,
--786,
-858,
-65,
-273,
-562,
--645,
-218,
+951,
+353,
+-116,
 -302,
-68,
--700,
--760,
--512,
--899,
--6,
-661,
--507,
-1016,
--679,
-325,
-314,
-601,
--280,
--22,
--761,
-238,
-779,
+-833,
+354,
+-504,
+758,
+-157,
 -260,
-118,
--836,
-801,
-62,
--382,
--187,
-233,
-285,
-890,
-978,
-888,
+-714,
+-594,
+-906,
+734,
+-897,
+99,
+440,
+-146,
+-449,
+584,
+-1007,
+368,
+-508,
+-870,
 -551,
-901,
--31,
-170,
-44,
-850,
--288,
-107,
-882,
--786,
+-361,
+646,
+29,
+486,
+742,
+-962,
 -718,
-936,
-761,
--796,
-374,
--142,
-868,
--410,
-579,
--122,
--14,
--1018,
--858,
-419,
--393,
--508,
--396,
-454,
--857,
--576,
--147,
--222,
--90,
--263,
--358,
--810,
-178,
-737,
--966,
-40,
--246,
-683,
-916,
--843,
--633,
-849,
--101,
-744,
--375,
-181,
--314,
-372,
-58,
-404,
--428,
--894,
--743,
-694,
--293,
--128,
--604,
--734,
--228,
-325,
--1018,
--150,
-202,
--808,
--610,
-741,
--978,
-191,
--57,
-458,
--29,
-169,
-193,
--43,
-1019,
-963,
--302,
-318,
--229,
--900,
--543,
--16,
--512,
-497,
--455,
--970,
-435,
--186,
-819,
--860,
-790,
-44,
-642,
-398,
--730,
--109,
--42,
-651,
--278,
--744,
-411,
--69,
+-22,
+548,
+718,
+-905,
+922,
+345,
+-159,
+495,
+429,
+525,
+937,
+-731,
+267,
+-878,
+773,
+835,
+27,
+855,
+643,
+-405,
+-505,
+648,
+405,
+535,
+-268,
+582,
+89,
+576,
+-268,
+-919,
+280,
+-216,
+-77,
+-476,
+-718,
+220,
+-608,
 -276,
--955,
--36,
-660,
--396,
--471,
-892,
--46,
-147,
--312,
-167,
--46,
--456,
-827,
--468,
-314,
--244,
-690,
+-424,
+859,
+97,
+256,
+-880,
+482,
+-817,
+929,
+-593,
+-1023,
+944,
+814,
+-502,
+200,
+182,
+990,
+-246,
+421,
+419,
+-867,
+262,
+-826,
+110,
+760,
+-18,
+5,
+-232,
+493,
+950,
+-930,
+-758,
+-805,
+546,
+-1001,
+811,
+-73,
+-679,
+92,
+-245,
+-447,
+-326,
 47,
+-646,
+411,
+-587,
+875,
+853,
+739,
+168,
+-754,
+666,
+124,
+-943,
+-214,
+-441,
+-739,
+30,
+-454,
+-254,
+808,
+610,
+-782,
+-212,
+381,
+-668,
+-1017,
+-927,
+-47,
+990,
 569,
-979,
--1005,
-689,
--748,
-629,
--896,
-13,
+88,
+931,
+865,
+-366,
+792,
+677,
+-523,
+469,
+-486,
+-995,
+687,
+692,
+908,
+663,
+645,
+-734,
+963,
+1008,
+-255,
+494,
+190,
+423,
+-895,
+-830,
+-319,
+296,
+261,
+-276,
+-259,
+244,
+-82,
+-1006,
+-732,
+-779,
+873,
+-144,
+144,
+253,
+319,
+-967,
+-527,
+396,
+-949,
+567,
+630,
+-559,
+-203,
 674,
--814,
--520,
-479,
--691,
-704,
--153,
--553,
-146,
--757,
-547,
--133,
-597,
-730,
-394,
-368,
--544,
--222,
-884,
-860,
--459,
-502,
--648,
-740,
--522,
-912,
-637,
-672,
-711,
-171,
-912,
-657,
-98,
--96,
--53,
--495,
--265,
--962,
-484,
--138,
--618,
-316,
--119,
-489,
-951,
-45,
--372,
-987,
--90,
-741,
-29,
--756,
--147,
-640,
+738,
+-286,
+-1002,
+967,
+173,
+302,
+-967,
+-637,
+472,
+-440,
+-932,
+-221,
+-326,
+-152,
+474,
+-830,
+57,
+-178,
+972,
+-667,
+815,
+538,
+84,
+-362,
+-870,
+-92,
+-208,
+-825,
+300,
+-173,
+354,
+604,
+102,
+-422,
+578,
 -539,
-376,
--875,
-680,
--261,
-988,
--371,
-724,
-999,
--101,
--617,
-397,
--1020,
-983,
-204,
-74,
--615,
-626,
-113,
--776,
-35,
-321,
-488,
--614,
--847,
--306,
--237,
-15,
-393,
--148,
-603,
-692,
--713,
-483,
--332,
--51,
--226,
-444,
--564,
--161,
-411,
-598,
--431,
--583,
-938,
--607,
-448,
-614,
--337,
--11,
--962,
-127,
-425,
--818,
--587,
-710,
--698,
-570,
--355,
--900,
-165,
--358,
-229,
-52,
-802,
--1022,
-52,
--604,
-86,
+795,
+714,
+-671,
+555,
+-373,
+1,
+-827,
+1007,
+-650,
+-914,
+-1003,
 328,
-353,
--940,
--256,
--578,
--912,
+253,
+-287,
+-62,
+-398,
+-37,
+289,
+141,
+-866,
+590,
+1,
+-464,
+789,
+246,
+665,
+-565,
+72,
+-1023,
+-149,
+-859,
+-109,
+-679,
+-972,
+552,
+-592,
+81,
+-621,
+62,
+322,
+163,
+-970,
+112,
+1009,
+-189,
+-54,
+-924,
+1001,
+-218,
+-211,
+721,
+71,
+956,
+818,
+-719,
+383,
+-367,
+989,
+-268,
+823,
+-104,
+386,
+1010,
+930,
+-841,
+-774,
+-307,
+-644,
+-836,
+-537,
+215,
+50,
+-901,
+869,
+-441,
+776,
+155,
+984,
+933,
+-618,
+363,
+-820,
+-111,
+-577,
 463,
+-794,
+-955,
+-786,
+-541,
+728,
+-827,
+1003,
+-787,
+-310,
+76,
+813,
+-187,
+-448,
+-627,
+-993,
+759,
+339,
+-426,
+-464,
+330,
 -1006,
-777,
-255,
+241,
+991,
+-193,
+696,
+-645,
+-557,
+-117,
+-323,
+896,
+-120,
+-562,
+-569,
+454,
+-317,
+836,
+-641,
+549,
+655,
+643,
+412,
+921,
+-680,
+736,
+-296,
+533,
+196,
+-885,
+-233,
+-490,
+-50,
+-470,
+-356,
+-232,
+402,
+-735,
+532,
+-983,
+-925,
+-796,
+-728,
+-969,
+-98,
+-684,
 515,
--899,
--34,
-975,
--30,
--97,
--815,
-316,
--923,
--324,
-728,
-722,
--99,
--86,
-687,
--636,
--535,
-761,
--759,
+172,
+-736,
+-912,
+814,
+484,
+254,
+-260,
+230,
+-1008,
+128,
+89,
+1023,
 -723,
--850,
-23,
--16,
--608,
-242,
-736,
-368,
-916,
-892,
-215,
--785,
-597,
+531,
+-1011,
+450,
+779,
+21,
+-100,
+-554,
+884,
+-603,
+-89,
+457,
+560,
+518,
+170,
+-715,
+700,
+811,
+-37,
+-893,
+530,
+-1010,
+-973,
+150,
+304,
+356,
+378,
+452,
+949,
+-333,
+922,
+-103,
+-480,
+900,
+-841,
+-329,
+-112,
+508,
+-384,
+730,
+229,
+190,
+150,
+982,
+-1003,
+37,
+-211,
+-924,
+358,
+544,
+-758,
+849,
+410,
+843,
+-378,
+898,
+-505,
+90,
+-43,
+-888,
+-731,
+-83,
+528,
+-613,
+-243,
+1007,
+981,
 587,
--33,
-382,
--147,
+-968,
+681,
+237,
+523,
+893,
+592,
+-583,
+384,
 -352,
--537,
--738,
-472,
--844,
-367,
-408,
+247,
+-771,
+-902,
+-744,
+874,
+30,
+134,
+-63,
+128,
+-530,
+-55,
+1016,
+-863,
+63,
+-538,
+790,
+306,
+-73,
+-328,
+-800,
+729,
+387,
+669,
+961,
+-881,
+-95,
+544,
+0,
+-19,
+-58,
+556,
+-399,
+-948,
+-1008,
+28,
+273,
+803,
 -742,
-282,
-451,
-427,
--696,
-227,
-374,
-753,
--971,
-735,
--768,
-941,
-888,
--854,
-166,
--590,
--809,
--187,
-633,
--607,
-241,
-215,
--516,
--345,
--135,
--51,
-215,
--690,
--807,
--595,
-607,
--574,
--323,
--765,
--3,
--970,
--316,
-557,
-985,
-983,
--368,
--763,
--316,
-943,
-284,
--341,
--24,
+653,
+51,
+-19,
+839,
 892,
-335,
-498,
--241,
-982,
--619,
-560,
-541,
-267,
--458,
--264,
--661,
-476,
-695,
-142,
--306,
-157,
-451,
-180,
-408,
-576,
-179,
--56,
-546,
-409,
+656,
+659,
+-791,
+-662,
+-490,
+855,
+-2,
+-525,
+-648,
+-211,
+788,
+1013,
+-126,
+874,
+940,
+-804,
+-1002,
+31,
+244,
+-827,
+-254,
+-220,
+-399,
+201,
+-369,
+257,
+-427,
+485,
+373,
+-373,
+-207,
+393,
+284,
+-382,
+985,
+-328,
+76,
 965,
--73,
--140,
-903,
-984,
+-237,
+-380,
+861,
+239,
+-263,
+-727,
+-896,
+-217,
+-745,
+-832,
+-347,
+-961,
+-615,
+-888,
+-756,
+337,
+-795,
+-275,
+-123,
+945,
+365,
+-483,
+901,
+-177,
+487,
+301,
+338,
+932,
+-545,
+574,
 146,
-890,
-461,
--173,
-21,
--286,
-561,
-949,
--653,
-1019,
--813,
-666,
--5,
+10,
+-610,
+-424,
+16,
+-545,
+140,
+741,
+389,
+75,
+-54,
+795,
+-283,
+705,
+-743,
+320,
+168,
+-493,
+202,
+708,
+753,
+-363,
+-33,
+-585,
+-567,
+850,
+310,
+-920,
+-606,
+-406,
+-141,
+-127,
+-30,
+905,
+-29,
+996,
+704,
+14,
+831,
+-502,
+-427,
+-518,
+-543,
+923,
+-725,
+-6,
+724,
+-613,
+-820,
+365,
+1,
+-234,
+694,
+452,
+-597,
+-812,
+945,
+160,
+494,
+-664,
+218,
+-714,
+-913,
+600,
+70,
+289,
+-175,
+-808,
+241,
+863,
+-479,
 412,
--55,
-438,
--113,
-426,
+-195,
+-879,
+-830,
+550,
+-746,
+-948,
+233,
+-515,
+-32,
+-664,
+72,
+914,
+-539,
+624,
+-428,
+121,
+209,
+-630,
+-414,
+-225,
+-459,
+-169,
 };
 
 const int m_exp[] = {
-162957,
-806616,
--3843440,
-425896,
-2155437,
--822680,
--787915,
-621747,
-2923873,
-1244247,
--97532,
-610424,
--202711,
-3997637,
--1726865,
-619432,
-907569,
--1548516,
--94406,
-1429714,
-728504,
--1334534,
--1711394,
--60696,
--1270495,
-42411,
--393126,
--1426839,
-766595,
-196776,
--1218293,
-805161,
-1753338,
-19573,
-1743399,
-1305905,
-2305550,
-691613,
-2936488,
--2499956,
-2011437,
--576528,
-4857673,
--522113,
-1643246,
-2107051,
-1243612,
-1226992,
--579490,
--2007336,
-735201,
-215903,
--2371734,
--604993,
--2228234,
--2776681,
--2711032,
--1272536,
--2272696,
--421571,
-1355942,
-88773,
-2385693,
--574961,
--2571513,
-636093,
--458404,
-546496,
-2480672,
-155697,
-101546,
--1189865,
-1932128,
-3516193,
--155025,
-758029,
--2133330,
-292469,
-259193,
-1783456,
-3153657,
--4306725,
--1756271,
-5321880,
-3387043,
--2499824,
--1354163,
-3589105,
--822661,
--787877,
--1296774,
--53917,
-1000889,
--2479239,
-522736,
-1610045,
--1888070,
-1262549,
-2391084,
--2993663,
-880432,
-2253879,
--738882,
-2578797,
--91861,
-3546872,
-1784236,
--1334715,
--370233,
-4225505,
-523924,
--1638264,
-496623,
--4977506,
--932716,
-2456108,
--1528737,
-619156,
--1996122,
--559761,
--3729495,
-2583519,
--688975,
-2327479,
-538691,
--216394,
--624983,
-3287485,
--948798,
-938455,
-1079709,
-2579776,
-560856,
-71743,
-2473123,
--1855915,
-982427,
--322892,
--1051648,
-279832,
-928945,
--563243,
-1512085,
-2329787,
--291655,
-219971,
-2918255,
--418247,
-1076304,
--260668,
-1590893,
-96628,
-459386,
--535583,
-20579,
--2682940,
-250239,
--1121829,
-4796976,
--1449568,
--884030,
--1012825,
--557312,
--660713,
-2703126,
--1000609,
-4119713,
--2261740,
-568061,
--807901,
-571882,
-4117378,
--3168209,
-1856661,
--1360182,
--275958,
--1371570,
-1334388,
-1163735,
-902800,
-1090827,
-2127115,
-1805917,
-1394998,
--1897442,
--1552564,
-4831224,
-5873204,
-862717,
--1328233,
--481290,
--3881988,
-1967283,
--3210455,
-194601,
-3623555,
-4233139,
--5588664,
-460768,
--2153116,
-1640966,
--1574766,
-4489450,
-2035547,
-591125,
--112588,
--677732,
-3516302,
--1487398,
--1392413,
--549940,
--122305,
-1734974,
--1296448,
-3345459,
-3278967,
--403583,
--2476952,
-3103106,
-3668860,
-876310,
--1372508,
-1606576,
--1242432,
--17363,
--536767,
--957924,
--840288,
-1387793,
--1263453,
--1580935,
-1255033,
--3199921,
--582299,
--3536182,
-8014,
-694206,
--868893,
--733959,
-691991,
-1905970,
--2105517,
--611054,
--2008274,
-1869390,
-781905,
-3204537,
-846463,
-582367,
-1557785,
-970604,
--232359,
--2832418,
-362033,
--287796,
--295851,
-115014,
-2136015,
--355990,
--3623476,
--2897332,
-3725302,
-243172,
--1681724,
--385794,
-472207,
--2443791,
-873057,
--2272285,
--1444140,
-127766,
--1695914,
--2155654,
-1292115,
-376558,
--2959312,
--1679103,
--1846952,
--2678017,
--3209721,
-4578141,
-3381475,
--115092,
--277905,
--1622304,
-444339,
-31506,
--1835191,
--1095261,
-1184692,
--1320035,
--768825,
--24927,
--441171,
--2206653,
-3016395,
-1322957,
-2359340,
--2045739,
--2535559,
-2589403,
-488750,
--352073,
--1583926,
-2559165,
--236263,
--2305301,
--1143355,
--811472,
-157863,
-2734081,
--3540908,
--3266165,
-3356957,
--2546548,
-120541,
--614357,
--321236,
--4262586,
-3134062,
-69010,
-1673410,
-578416,
-1494750,
-1401048,
-1434942,
-2202613,
--1168868,
-4234494,
-1083346,
-1136258,
-2189311,
-2610004,
-981866,
--1297232,
-590461,
--1470399,
-182048,
--987329,
--815848,
-4511995,
-3218281,
-1891474,
-2139246,
--1953818,
-760442,
-4370650,
--155141,
-141023,
--3911135,
-704029,
-1270803,
--1160789,
-639262,
--140142,
-492202,
--3552999,
-726790,
-730943,
-294228,
-1626723,
--2909194,
--54567,
-1406030,
--511915,
-14996,
-2901103,
-202415,
--765411,
-3954449,
-1634496,
-765702,
--1670288,
--1746308,
--236974,
-402163,
-2225940,
--4092672,
--487529,
-458477,
--850359,
-2502080,
--1833536,
--2648383,
-469522,
--2065549,
--1110710,
--1161586,
--1598119,
-312213,
--3414626,
-2809732,
--1270950,
-231301,
-948652,
--5241164,
--142131,
--101551,
--924256,
--4661011,
-1373376,
--384970,
--1771026,
-2352833,
--1508015,
--1567875,
--77896,
--298333,
--2109926,
-1841301,
--1841526,
-463269,
--1579323,
-3378298,
--1686942,
-2286923,
--1725730,
-888978,
--132747,
-1196962,
-765073,
-2561682,
-3394037,
-686372,
-3609261,
-1739268,
--1632029,
-223041,
-7802,
-1613144,
--1505669,
-236211,
-2125044,
--287008,
-1433003,
--3433392,
-2365446,
--1739067,
--1742567,
-572370,
-387905,
-137241,
--205103,
--3092809,
-3403304,
--626777,
-1682114,
-208043,
--3166100,
--494447,
--1814690,
--1598424,
--5459778,
-2167916,
--1761705,
-425344,
-1806134,
--1823602,
--846377,
--1890539,
--462188,
--1485705,
-101051,
--3918737,
-1878057,
-527232,
--3992037,
-1295344,
-76046,
--3594473,
-165856,
--209224,
-2988651,
-1205027,
--128456,
--688247,
--1858378,
-1319809,
--654178,
-1197795,
-2203611,
-429127,
--153254,
--1002844,
--73001,
--777076,
-624879,
--749349,
-517398,
-1047706,
--483949,
-21723,
--2834312,
-575021,
-2162429,
-1157482,
-251420,
-165568,
-3589958,
-1166415,
-971492,
--791336,
-221991,
-1202426,
--486702,
--2021786,
--998812,
-1488325,
-875139,
-2095034,
--3432500,
--3200858,
-708188,
--1230553,
--1027252,
-3204443,
-1541304,
-1441524,
-2393495,
-240432,
-154630,
-711539,
-1546622,
--3606712,
-2329971,
-2037395,
-488837,
--1149213,
-1004476,
--5353576,
--1445258,
-1830702,
-2334502,
--1027586,
--786918,
--830800,
--5014319,
-603373,
--2356785,
-391822,
--559514,
-832315,
-854580,
--274568,
--852866,
--1829693,
--224252,
--442985,
--511368,
--2738731,
-785781,
-1213180,
--452741,
--1912617,
-2501858,
-1190107,
-374756,
-4063611,
--3167900,
-590894,
--843960,
--1329399,
-1982532,
--1501664,
--480615,
-978312,
--776372,
--758612,
--342051,
-1297314,
--1330216,
-2192151,
-514261,
-1924564,
--1450262,
--1368891,
-709611,
-910433,
-10007,
-379287,
--927492,
-2341527,
--614574,
-213605,
-1313660,
--511351,
-2179833,
--3229739,
-970187,
-24229,
--563469,
--290337,
--1413501,
--1110798,
-1687978,
--1070551,
-667748,
--378655,
-1343272,
--1179176,
-465076,
-1061328,
--725446,
--632171,
--3115122,
-126962,
-679313,
-468649,
--54748,
--1693703,
--1504854,
-1706614,
-1984152,
--147089,
--677246,
--2363197,
--958222,
--1058809,
-2137431,
--1457226,
-1613151,
--1096760,
--62862,
--1303936,
--470334,
--677938,
--3350678,
-686828,
--3218401,
--2491773,
-823333,
--3515802,
--521386,
--745281,
-1801059,
--364988,
-262475,
--1628907,
-4610600,
--1790296,
-1637070,
--111930,
-695554,
-631793,
--58009,
-879861,
--396942,
--3491554,
--2099070,
--1427409,
-274894,
-3138017,
--3984461,
--1428478,
-1281318,
-1326242,
--2425032,
-3375811,
-1883570,
-986169,
--2026937,
-851459,
--2948484,
-887635,
-1664977,
--787389,
-192262,
--199263,
-2417949,
-1357190,
--675828,
--2140148,
--1992713,
--812035,
--189832,
--1687171,
--1455287,
--1968142,
--95663,
-1491237,
--97044,
--1478691,
--818723,
-540642,
--1935376,
--2011417,
--1496601,
--2432622,
--1711594,
-1082040,
-4280682,
--245468,
-2632647,
--2126102,
-4408049,
-1347238,
-1881351,
-2741419,
-275662,
-39237,
--90511,
-293945,
-1132383,
--38055,
-1318991,
--285413,
--1197211,
-758713,
-5302556,
--910446,
--1192408,
-1951665,
-599939,
-1684692,
--905949,
-467600,
--1781331,
-2472257,
--4560636,
-2121801,
-7702,
--953875,
-810154,
-2235783,
--1721404,
-1466909,
-823131,
-257449,
--467541,
-2196156,
-1231438,
--1055257,
-1235031,
--2824591,
-1407798,
-4017064,
-496444,
--2144864,
-172928,
--780962,
--1351533,
--770027,
--2035068,
--10361,
-2252778,
--21797,
--3779047,
--3409395,
-494668,
--331048,
--331324,
--355901,
--429937,
--109578,
-2141215,
-3507397,
--475137,
-1741184,
-2711387,
-1735523,
--546046,
--104827,
-1803323,
--1088839,
-10836,
--1512582,
--1126855,
-1130557,
--3863013,
--1429454,
--2311546,
-988010,
--34163,
--2855425,
-668343,
--504504,
--875488,
-29458,
--86216,
-288365,
--434627,
-3493242,
--130938,
--1988278,
-3075370,
-953334,
--4261726,
-691990,
--344124,
-1873383,
--2948777,
--37293,
--6201594,
-194461,
--3063704,
-3296851,
--361236,
-2148972,
--1689659,
--1587584,
--1973610,
-136604,
-864708,
--208814,
--5210513,
--2382198,
--938078,
--623312,
-2134892,
-20690,
--328389,
-425817,
-1494259,
-341657,
-2041202,
-215726,
-1999270,
-4556161,
--1534970,
-846236,
--1779277,
-487163,
--2503709,
--881316,
-641969,
--1533604,
--1307702,
--28391,
-2386501,
-1889399,
--2798701,
--1452890,
--2458612,
-1135665,
--822540,
--1224665,
--3168295,
-3202581,
--1833843,
-2028127,
-316317,
-2771444,
--1969294,
--1862180,
--1846963,
-3288593,
--1058949,
--4301372,
--2270312,
--1814302,
--2507256,
-2304887,
--1357358,
--398110,
-2224556,
--3014825,
--1386326,
-3381979,
-1996942,
-1199292,
-2726576,
--2409907,
--651850,
-619670,
--1475982,
-2412631,
--1348814,
--1226204,
-39236,
-1327167,
--1771649,
-547246,
--147406,
-2114780,
--861050,
--369180,
--1802507,
--1044212,
-1027808,
--268747,
-898922,
--1231614,
--2479622,
-3128666,
-2129144,
--2150987,
-743,
-1381784,
-1709971,
-3159903,
--1724033,
-1467305,
--1549225,
--1149491,
--1241667,
-1923850,
--1031013,
--1854871,
-5195468,
--1312499,
--1683183,
--2852635,
-1375570,
--2626830,
-252623,
--2799957,
-416997,
--768714,
-521710,
--1525373,
-1399518,
-781524,
-1078645,
-481978,
--169425,
--595560,
--4627513,
-1999980,
--1414190,
--912128,
--464019,
--1908668,
--1299171,
-2822564,
--3299799,
-1134772,
--717124,
--1175570,
-1450585,
--1338075,
-886999,
-352573,
-4223081,
-247130,
-1087502,
--1634036,
--1292452,
-595147,
-2611183,
--1018214,
-3045670,
-25600,
-527454,
--410755,
--1575051,
--4673725,
-2022228,
-342427,
--411410,
--3969320,
--1582114,
--196149,
-417835,
--2069373,
-1140594,
--1982957,
--1168881,
--1680247,
-3341440,
--2163154,
-644414,
-899481,
--3068320,
-1730507,
-2120763,
-800244,
--1950761,
-2797422,
-999558,
-4224011,
-267319,
--2575349,
-1303602,
-3822345,
--1179231,
--83832,
-1094373,
--2890364,
--2847299,
-1097183,
-270678,
--1041859,
--243475,
--2372503,
--983943,
-2643567,
--1140781,
--1564791,
-921293,
--2492580,
-412995,
-3308919,
--69183,
--1264313,
-292080,
--1901711,
--2160028,
--1433855,
--1297772,
-1209393,
--760271,
--1505265,
--2822187,
--750104,
-1326341,
--2406512,
--2811558,
-634457,
-2153554,
-1165807,
--1251857,
--2639503,
--2455102,
--397631,
-988547,
--2515259,
-4482079,
-763623,
-559995,
--1070705,
--2788379,
-545963,
--1318607,
--896632,
+4483102,
+-3837116,
+2010720,
+1534171,
+1759792,
+1381371,
+-3260944,
+-5398276,
+508357,
+615916,
+-595144,
+-1125435,
+116369,
+-3983003,
+622692,
+-104608,
+-2819471,
+349802,
+1320318,
+-2792252,
+235582,
+2875734,
+-2702444,
+588279,
+28913,
+-2727311,
+-2253437,
+2340518,
+4453365,
+450661,
+51465,
+1552749,
+278332,
+-2761260,
+-2945973,
+-602713,
+934446,
+-1807317,
+661383,
+2595744,
+938496,
+355527,
+-4539675,
+3451768,
+-3198419,
+-859721,
+439576,
+2530247,
+231425,
+1196852,
+-3068189,
+1806918,
+-1511788,
+-258993,
+-471602,
+3028081,
+184421,
+1760131,
+1900049,
+1894075,
+-2293124,
+1379361,
+3920518,
+-104505,
+-1537037,
+795601,
+-1678136,
+1156068,
+72605,
+-3464777,
+-3893561,
+1158218,
+1450934,
+2863262,
+522263,
+1510903,
+2791830,
+-1889894,
+-475635,
+745893,
+-2488967,
+1253464,
+-2450836,
+57133,
+-1007562,
+-602981,
+694793,
+3349495,
+-3611637,
+3738053,
+648859,
+-1421083,
+-2042791,
+348004,
+547002,
+-694910,
+1278153,
+1258721,
+-15823,
+958095,
+-3246096,
+-1176319,
+2696030,
+2353825,
+-851897,
+-305490,
+1144869,
+-489083,
+-771523,
+1010824,
+-823073,
+425529,
+-1866686,
+-964908,
+1199108,
+-859023,
+-1614558,
+-4792185,
+98458,
+-607409,
+551516,
+-2382173,
+548885,
+-3491850,
+-1445795,
+-1088426,
+-429760,
+-1591568,
+-229487,
+-2804269,
+-95359,
+-680286,
+2228343,
+-149962,
+-3782020,
+1139786,
+-832209,
+1322977,
+335698,
+3772007,
+550017,
+-3472461,
+-586406,
+1670398,
+30538,
+-282598,
+-1368441,
+830018,
+32500,
+1604523,
+-452374,
+1455260,
+1154312,
+726318,
+468334,
+2820494,
+-4420185,
+979034,
+1611313,
+498563,
+-13770,
+851039,
+-3280587,
+3868879,
+-103066,
+1718207,
+-1003760,
+948733,
+1821080,
+226516,
+153022,
+-2163314,
+-1801575,
+1033501,
+1049469,
+-4055909,
+2429031,
+-1896082,
+3772971,
+1794560,
+-1181046,
+1784208,
+1762035,
+1150961,
+-2353643,
+-3537267,
+4644670,
+-350704,
+-65156,
+-381320,
+503403,
+798969,
+-1050893,
+2272013,
+69652,
+-2143175,
+-2519168,
+181113,
+1345533,
+-1526855,
+-3214613,
+-1569966,
+1643995,
+3408010,
+2140174,
+2881951,
+-1757528,
+1421885,
+133420,
+26271,
+3763843,
+164641,
+2934726,
+1078553,
+-318986,
+-1270954,
+3115586,
+1701686,
+-110878,
+835257,
+-1885901,
+-2510356,
+-1454398,
+-909358,
+-1059566,
+672173,
+795074,
+-137536,
+250484,
+1597909,
+1086027,
+1172814,
+-819859,
+868662,
+1775692,
+-1335021,
+-1056456,
+-1069776,
+1901756,
+1119316,
+31858,
+2500976,
+-1272156,
+3567659,
+743992,
+-434535,
+-388297,
+2944552,
+1121393,
+-211286,
+1119305,
+-837106,
+-3291772,
+-644156,
+2234998,
+-1812166,
+-424086,
+-2579182,
+-4995925,
+1284777,
+-1104876,
+1042320,
+1483588,
+3210383,
+2799297,
+-950947,
+-1879848,
+-1211673,
+2034864,
+-1792252,
+1649777,
+-4130898,
+1649310,
+-4146430,
+3258963,
+2573110,
+-1215090,
+-3430412,
+-263614,
+-224321,
+2327869,
+-5401598,
+1214211,
+-4033005,
+-1731759,
+285306,
+141634,
+-1576674,
+-598491,
+-3747626,
+-2674724,
+-2964686,
+-970824,
+638559,
+2705879,
+-2155436,
+2188859,
+739422,
+-3369981,
+2369939,
+3084272,
+-416591,
+1505156,
+907810,
+316916,
+-3130875,
+2304606,
+-455439,
+713440,
+1125996,
+-233188,
+-558299,
+657160,
+-4046801,
+2829443,
+2340003,
+-1458632,
+248823,
+1757427,
+-748600,
+955486,
+295305,
+-3541031,
+2524547,
+-252008,
+1798917,
+-651016,
+317232,
+-166186,
+-2947741,
+-630069,
+-57348,
+-416785,
+-786579,
+1163624,
+-2816194,
+-1423888,
+-949604,
+1633977,
+2745062,
+1446168,
+1744305,
+-2731042,
+2067611,
+480904,
+317713,
+-7337,
+-2597976,
+-617019,
+509827,
+-1924859,
+-1690807,
+-2117204,
+-349188,
+4195303,
+539892,
+-729517,
+266502,
+-261572,
+-2455571,
+-1376324,
+218379,
+-803670,
+1434548,
+-2308471,
+1458581,
+-3021782,
+4822354,
+-1711275,
+2483161,
+-4126171,
+-1416011,
+2054421,
+1405379,
+1427968,
+1111079,
+415834,
+2082764,
+-2498692,
+5230909,
+3293478,
+-641370,
+650141,
+322583,
+756632,
+-173745,
+-3481500,
+-2660545,
+-11120,
+-463314,
+-320049,
+-444431,
+-230492,
+268636,
+-685754,
+127531,
+1545264,
+-194991,
+537373,
+-4125440,
+1332649,
+-3250371,
+744884,
+552588,
+-34707,
+2343452,
+2367038,
+-97324,
+639918,
+3476436,
+-545169,
+573332,
+1659803,
+1633795,
+813581,
+4605630,
+1099838,
+-2822247,
+-251728,
+1037532,
+1960620,
+84911,
+-1956257,
+-240695,
+2421524,
+-63612,
+-3347511,
+-857378,
+-1357842,
+359486,
+1415311,
+807379,
+1495270,
+1761577,
+1310276,
+-2769105,
+397395,
+-1840322,
+-1884252,
+-255958,
+-3081382,
+-2124826,
+1747341,
+-978451,
+1353749,
+-1101153,
+3971279,
+206472,
+2015201,
+2594079,
+-4123051,
+-891957,
+-29729,
+858096,
+-157999,
+-1592225,
+-342823,
+358660,
+1886511,
+240694,
+443098,
+2080523,
+6024770,
+-3700560,
+1297889,
+-2332463,
+4623788,
+5007792,
+551928,
+1178476,
+1577810,
+-693530,
+1354736,
+-1505662,
+-484525,
+1556410,
+-3354635,
+3291809,
+-2639983,
+3200194,
+-2593644,
+-2366448,
+-683796,
+-36714,
+1331052,
+-1496870,
+-3225114,
+-296918,
+1061488,
+3078786,
+-870052,
+-471043,
+-450798,
+2272153,
+-1273088,
+335248,
+1065448,
+3173343,
+-1399606,
+-2210999,
+1058461,
+908236,
+-1407584,
+404896,
+-4314166,
+-1677042,
+-529283,
+451139,
+2439430,
+-1969670,
+1996189,
+-1579974,
+905677,
+606252,
+-2345497,
+117665,
+-1449525,
+192620,
+-366900,
+-1746077,
+492086,
+484400,
+2604739,
+-1698451,
+1806720,
+-87307,
+1467434,
+711222,
+-1150222,
+765562,
+864307,
+3684554,
+2041083,
+814638,
+-2386832,
+-1146280,
+1680834,
+-700110,
+-1902486,
+357758,
+352621,
+-1794699,
+-1653822,
+-3480866,
+1426643,
+1112374,
+-215960,
+822012,
+40742,
+-2923668,
+-512511,
+60563,
+-56006,
+-13177,
+-983907,
+511449,
+1116084,
+-1402263,
+-3137436,
+-4694304,
+-1401688,
+-2354587,
+1718534,
+-632354,
+654037,
+-88483,
+-4299079,
+-1856729,
+-2314680,
+1964063,
+-796198,
+-3474061,
+820417,
+-737218,
+670097,
+-446965,
+633105,
+-3892853,
+3384774,
+426598,
+1088596,
+4790552,
+-2839492,
+-631535,
+3797532,
+395418,
+-264991,
+-572710,
+1830209,
+-1663436,
+1525072,
+833800,
+2468120,
+-1823599,
+2424168,
+-1128471,
+734423,
+70049,
+3244504,
+-396877,
+660320,
+-2949195,
+3406669,
+5059168,
+2653356,
+146574,
+2938392,
+1549690,
+-1983162,
+-5055829,
+-2544940,
+-3117725,
+469784,
+861349,
+1629417,
+3031088,
+1047359,
+2584842,
+-1784451,
+305406,
+3022787,
+-1451097,
+1436183,
+-762256,
+744395,
+-2908817,
+-1213801,
+-386293,
+-2302623,
+-653467,
+-982398,
+4623358,
+2153086,
+-1134981,
+1542818,
+901108,
+-1140854,
+2472719,
+1890985,
+-1137157,
+2055100,
+427363,
+-31301,
+3626328,
+1441498,
+805798,
+-987709,
+-384910,
+556339,
+896837,
+-2517725,
+723811,
+-1027352,
+-1617258,
+-2740782,
+-2127484,
+-492670,
+-1321177,
+-695931,
+577228,
+-1283787,
+-1505847,
+-1619433,
+360553,
+-1596086,
+61969,
+-288930,
+-1216163,
+1797143,
+-742194,
+600011,
+-241270,
+351273,
+-3012471,
+-308291,
+-541947,
+-1902029,
+-621432,
+-2305293,
+-1348202,
+1533325,
+2275364,
+2169230,
+1916682,
+1411401,
+-785542,
+-2382906,
+4840078,
+522582,
+392169,
+375358,
+2385174,
+-581586,
+193744,
+-345388,
+2125393,
+791936,
+-255410,
+-166805,
+4406885,
+305722,
+2511267,
+-849139,
+-2348174,
+3341503,
+68576,
+-2799415,
+964959,
+-1760400,
+-2160747,
+2514562,
+-270562,
+1664455,
+-3942140,
+-2168858,
+-323717,
+-947870,
+163100,
+-1825118,
+-1130352,
+-114785,
+-4236782,
+-1576945,
+2847746,
+-3145574,
+1388272,
+-66245,
+-1423067,
+2859547,
+2677480,
+-1185722,
+2365942,
+520958,
+2195243,
+-3450331,
+1877006,
+916363,
+778,
+-2206051,
+-466361,
+202139,
+-601938,
+51036,
+3647401,
+1668170,
+-1185629,
+166863,
+-1725512,
+1789148,
+-788015,
+978567,
+-1712453,
+-554712,
+-3372699,
+1803344,
+-897854,
+-3236016,
+-556372,
+-1457671,
+460324,
+1060707,
+1357333,
+1045365,
+-1930717,
+777059,
+-425937,
+-834420,
+-1189939,
+194507,
+2218312,
+-1094666,
+-3705973,
+906096,
+2777722,
+2231731,
+-3171764,
+-486828,
+1078622,
+861832,
+-959015,
+2611599,
+-135350,
+610641,
+-711138,
+3462805,
+-1165215,
+3647783,
+1621886,
+-1086517,
+60868,
+252964,
+136173,
+1301024,
+-850624,
+1666167,
+190080,
+-2965398,
+625535,
+234678,
+857337,
+1049677,
+-2208563,
+-1023464,
+1639382,
+954470,
+910791,
+-1502177,
+423774,
+3466346,
+2168473,
+-2370899,
+-1089579,
+-642893,
+-2052317,
+3526387,
+19090,
+-610922,
+-1375271,
+2011766,
+-1847092,
+-1502616,
+-317094,
+-351499,
+-261107,
+-528293,
+-503966,
+1436907,
+127400,
+-1450589,
+-66531,
+4939251,
+874909,
+-1049924,
+5164310,
+-1147856,
+1531778,
+-1614006,
+-2698637,
+722371,
+1119194,
+-3034501,
+-2697062,
+2495670,
+-930079,
+-1131546,
+829346,
+-1164813,
+-2168788,
+-1645737,
+273444,
+1073158,
+1225225,
+1448733,
+-1007394,
+265930,
+-3175050,
+218077,
+1056685,
+-2199129,
+-3176095,
+-2370525,
+-3462299,
+-2743628,
+-360088,
+-57644,
+3564156,
+1606706,
+3244375,
+-186857,
+1536695,
+-2418550,
+53892,
+-1853967,
+592569,
+3926891,
+-1493479,
+389078,
+-2642271,
+-44987,
+-943272,
+-481791,
+-795250,
+-672906,
+2136971,
+-1367609,
+407249,
+1883390,
+157911,
+777254,
+787184,
+1699794,
+-1377880,
+-467162,
+288764,
+-690701,
+2719792,
+-265032,
+267114,
+153648,
+2766982,
+-452454,
+2613652,
+-1277024,
+-1898807,
+424545,
+151097,
+-256482,
+1146113,
+2211639,
+-2626339,
+-625983,
+281070,
+-3647088,
+-706152,
+-143082,
+1952901,
+2632401,
+1517832,
+-1108696,
+3699051,
+345245,
+-891000,
+2471600,
+607863,
+1851266,
+-4486490,
+658378,
+-3525421,
+-2891857,
+2602450,
+-1166765,
+-7675379,
+802565,
+2692739,
+2706973,
+1088197,
+-6186397,
+2071083,
+-154701,
+2900460,
+1320338,
+1257382,
+1069736,
+-1175915,
+-2392666,
+-342978,
+572962,
+-183757,
+3250507,
+1769902,
+-1889788,
+1288066,
+-1338865,
+2089714,
+460913,
+1644860,
+-3713037,
+1239140,
+57169,
+-2306307,
+1193413,
+-2803012,
+2167117,
+-3619679,
+3239335,
+1189800,
+1966178,
+1222678,
+-752261,
+2193804,
+-1141191,
+3448382,
+264414,
+1019178,
+-75820,
+77692,
+364805,
+1902676,
+-2993417,
+-2451383,
+-836725,
+-869437,
+-1744519,
+564727,
+-112585,
+257397,
+428714,
+404777,
+-556998,
+-964791,
+2624446,
+-1787540,
+2258460,
+-3420895,
+-1616792,
+-423558,
+1590940,
+4888957,
+2445789,
+-456651,
+2452891,
+-1108754,
+-816766,
+-4738775,
+121791,
+2463584,
+4530841,
+-2526737,
+2192419,
+1145563,
+337989,
+-470142,
+-1691399,
+27807,
+1202786,
+3026122,
+-2617275,
+-5442474,
+5435819,
+-305284,
+2150722,
+-386769,
+3584133,
+-105374,
 };
 
 #define SIZE 32
diff --git a/pulp_cluster-mchan-tests.yaml b/pulp_cluster-mchan-tests.yaml
index c5a68ab..ab54a94 100644
--- a/pulp_cluster-mchan-tests.yaml
+++ b/pulp_cluster-mchan-tests.yaml
@@ -28,4 +28,4 @@ mchan_tests:
     command: make clean all run
   testMCHAN_basic_8cores:
     path: ./mchan_tests/testMCHAN_basic_8cores
-    command: make clean all run
+    command: make clean all run
\ No newline at end of file
diff --git a/reliability_tests/dmr_matmul/Makefile b/reliability_tests/dmr_matmul/Makefile
new file mode 100644
index 0000000..66ed60e
--- /dev/null
+++ b/reliability_tests/dmr_matmul/Makefile
@@ -0,0 +1,12 @@
+PULP_APP = test
+PULP_APP_SRCS = dmr_matmul.c
+
+PULP_CFLAGS = -O3
+PULP_LDFLAGS = -lm
+
+ifeq ($(fault_inject),1)
+	export FAULT_INJECTION=1
+	export FAULT_INJECTION_SCRIPT=$(CURDIR)/pulp_inject_fault.tcl
+endif
+
+include $(PULP_SDK_HOME)/install/rules/pulp.mk
diff --git a/reliability_tests/dmr_matmul/dmr_matmul.c b/reliability_tests/dmr_matmul/dmr_matmul.c
new file mode 100644
index 0000000..0cfcc61
--- /dev/null
+++ b/reliability_tests/dmr_matmul/dmr_matmul.c
@@ -0,0 +1,100 @@
+/*
+* @Author: Michael Rogenmoser
+* @Date:   2023-02-17 18:00:21
+* @Last Modified by:   Michael Rogenmoser
+* @Last Modified time: 2023-02-17 18:15:33
+*/
+#include <pulp.h>
+#include <stdio.h>
+#include "matmul.h"
+
+#define N_ITERS 1
+#define max(x,y) (x > y ? x : y)
+#define min(x,y) (x < y ? x : y)
+
+__attribute__ ((section(".heapsram"))) int A[SIZE][SIZE];
+__attribute__ ((section(".heapsram"))) int B[SIZE][SIZE];
+__attribute__ ((section(".heapsram"))) int C[SIZE][SIZE];
+
+void initialize_mat();
+
+void initialize_mat() {
+  int i,j;
+
+  for (i=0;i<SIZE;i++) {
+    for (j=0;j<SIZE;j++) {
+      A[i][j] = A_init[i][j];
+      B[i][j] = B_init[i][j];
+    }
+  }
+
+}
+
+void matrix_multiplication(testresult_t *result, void (*start)(), void (*stop)());
+
+testcase_t testcases[] = {
+  { .name = "Matrix Multiplication", .test = matrix_multiplication },
+  {0, 0}
+};
+
+int main() {
+  if (rt_cluster_id() != 0)
+    return bench_cluster_forward(0);
+
+  hmr_self_enable_dmr();
+  hmr_set_dmr_config_all(0    , // Core ID
+                         true , // Rapid recovery enabled
+                         true , // Setback enabled
+                         false); // Synch req
+  printf("Available Config: %x\n", hmr_get_available_config(rt_cluster_id()));
+  printf("after setup: %x\n", hmr_get_active_cores(rt_cluster_id()));
+
+  hmr_setup_barrier(hmr_get_active_cores(0));
+
+  int nbErrors = run_suite(testcases);
+
+  synch_barrier();
+
+  return nbErrors != 0;
+}
+
+void matrix_multiplication(testresult_t *result, void (*start)(), void (*stop)()) {
+  int coreid = rt_core_id();
+  int numcores = 4;
+  int *CHKSUM_RESULT;
+  short int i, iter, j, k;
+  int lb, ub, chunk;
+
+  if (coreid == 0){
+    printf("Start ParMatrixMul\n",0,0,0,0);
+    // initialize matrix A and B
+    initialize_mat();
+  }
+  //number of rows each core has to multiply
+  chunk = SIZE / numcores;
+  //lower bound
+  lb = coreid * chunk +  min(coreid, SIZE % numcores);
+  //upper bound
+  ub = (coreid + 1) * chunk + min(coreid + 1, SIZE % numcores);
+  if (coreid == numcores-1) {ub = SIZE;}
+
+  synch_barrier();
+
+  /********************* Benchmark Execution *********************/
+  if (coreid<numcores) {
+    start();
+    for (iter = 0; iter < N_ITERS; iter++) {
+      for (i = lb; i < ub; i++) {
+        for (k = 0; k < SIZE; k++) {
+          C[i][k] = 0;
+          for (j = 0; j < SIZE; j++)
+            C[i][k] += A[i][j] * B[j][k];
+        }
+      }
+    }
+  }
+  synch_barrier();
+  stop();
+
+  /********************* Benchmark Execution *********************/
+}
diff --git a/reliability_tests/dmr_matmul/matmul.h b/reliability_tests/dmr_matmul/matmul.h
new file mode 100644
index 0000000..eaed788
--- /dev/null
+++ b/reliability_tests/dmr_matmul/matmul.h
@@ -0,0 +1,56 @@
+#define SIZE           24
+#define MAT_SIZE       (SIZE*SIZE*sizeof(int))
+
+int A_init[SIZE][SIZE]  = {
+    {181,242,121,183,206,92,200,157,59,246,13,24,108,168,125,179,134,202,74,79,86,66,63,29},
+    {178,198,233,162,184,134,150,155,153,199,114,94,81,0,92,164,206,186,166,191,173,25,82,38},
+    {216,38,65,42,106,189,132,201,214,41,24,231,249,78,84,109,168,159,127,107,217,254,197,217},
+    {64,22,112,106,14,207,197,73,149,38,106,250,17,254,196,121,218,22,28,237,31,87,82,132},
+    {165,67,227,22,65,61,140,27,231,194,168,145,33,44,175,127,31,56,194,227,186,13,252,27},
+    {160,183,169,113,226,50,116,138,23,147,247,20,73,174,185,178,196,200,232,178,234,117,9,251},
+    {36,192,256,31,150,165,159,79,217,121,13,123,77,233,87,219,20,225,3,27,22,69,216,92},
+    {151,236,72,44,178,105,211,237,216,24,107,168,194,178,99,168,246,250,120,212,18,39,194,112},
+    {230,114,234,101,27,26,158,119,246,3,119,80,227,214,96,87,35,179,195,156,71,84,53,89},
+    {188,225,242,243,130,165,133,225,222,186,139,152,171,23,156,42,214,11,46,45,206,78,34,196},
+    {232,161,201,187,247,80,215,11,9,164,220,201,118,161,98,1,155,14,60,7,82,118,211,230},
+    {4,91,20,188,232,229,234,82,240,175,105,12,160,204,177,201,129,241,83,235,125,216,91,29},
+    {120,97,74,42,130,29,15,160,252,82,140,48,226,25,154,19,207,190,197,216,177,8,14,111},
+    {175,204,50,69,35,251,68,208,171,18,135,186,200,139,182,170,7,144,83,83,79,15,192,165},
+    {236,198,116,130,248,198,107,218,107,105,178,209,163,22,218,163,144,65,55,48,102,217,57,158},
+    {192,201,215,171,104,183,55,72,75,214,72,181,85,212,173,88,183,144,221,90,117,129,201,103},
+    {32,75,38,70,53,76,66,27,221,10,17,122,234,218,56,39,63,112,194,124,178,81,244,38},
+    {76,225,145,248,141,149,27,180,84,164,149,19,101,179,147,10,93,14,219,241,77,241,136,156},
+    {120,146,26,53,175,54,79,69,113,29,204,90,203,227,1,180,227,100,44,206,163,17,215,25},
+    {134,167,185,215,181,100,189,203,220,158,112,2,124,100,183,251,110,159,21,118,77,12,249,5},
+    {161,110,245,113,227,139,107,236,250,244,231,63,191,75,102,110,18,210,26,133,53,66,50,140},
+    {151,76,114,163,117,76,250,117,154,97,73,235,147,9,176,91,246,256,72,13,228,172,109,156},
+    {105,13,88,253,167,189,23,42,137,249,15,115,118,234,126,129,167,54,211,229,243,249,143,118},
+    {191,245,66,166,36,208,209,127,193,246,68,224,31,65,56,167,241,144,111,30,54,115,34,56}
+};
+
+int B_init[SIZE][SIZE]  = {
+    {0,0,0,1,0,0,1,1,0,0,0,0,1,0,0,1,0,0,0,1,0,0,1,1},
+    {0,1,0,1,0,0,1,1,0,1,0,1,0,1,0,0,1,1,0,1,1,0,0,1},
+    {1,1,0,1,1,1,1,0,1,0,0,0,0,0,0,0,0,0,0,1,0,0,1,1},
+    {0,0,0,1,1,0,0,0,1,0,1,1,0,1,1,1,0,1,0,0,1,1,1,1},
+    {1,1,0,0,1,1,0,0,1,1,0,1,0,1,1,1,1,1,1,1,1,1,0,0},
+    {1,0,1,0,1,1,0,1,0,0,1,1,0,1,1,0,0,0,0,0,1,1,1,0},
+    {1,1,1,0,1,1,1,0,1,1,0,0,1,0,1,1,1,1,0,0,0,1,0,0},
+    {1,1,1,1,1,1,0,1,1,0,1,1,0,0,0,0,1,0,1,0,1,0,0,0},
+    {1,1,1,1,1,1,0,1,0,1,0,0,0,0,0,1,0,1,0,0,1,0,0,1},
+    {1,1,0,1,0,0,0,0,1,1,1,0,0,1,0,0,1,1,0,1,1,0,0,0},
+    {1,1,0,1,1,0,1,1,0,1,1,0,1,0,1,1,1,1,0,0,1,1,0,0},
+    {0,0,0,0,1,1,0,1,0,0,1,1,0,0,0,1,0,0,1,0,0,0,1,1},
+    {0,0,1,0,1,1,0,1,0,0,1,1,1,1,1,0,1,0,1,0,0,0,1,1},
+    {1,0,0,0,0,1,1,0,0,1,0,1,1,1,0,0,1,0,1,1,1,1,0,0},
+    {1,1,1,1,1,1,0,0,1,1,1,0,0,1,1,1,0,1,0,1,0,0,0,0},
+    {1,1,0,1,1,0,1,1,0,0,0,1,0,1,0,1,0,0,1,1,0,1,1,0},
+    {0,1,1,1,0,0,1,0,1,1,0,0,0,1,1,0,1,1,1,0,0,0,0,1},
+    {1,1,0,1,1,0,0,0,1,0,0,0,1,1,1,1,0,0,0,0,1,0,0,1},
+    {0,0,0,0,1,0,0,1,0,0,1,1,0,0,1,1,0,0,0,0,1,0,0,1},
+    {0,0,1,1,1,0,0,1,0,0,0,1,1,0,1,0,1,1,1,1,0,1,1,0},
+    {1,0,1,0,1,0,1,0,0,1,1,1,1,0,1,1,0,1,0,1,0,0,0,1},
+    {0,0,1,0,0,0,0,0,1,1,1,1,1,1,0,0,1,1,1,0,0,1,1,0},
+    {1,0,1,1,0,1,1,0,0,0,0,1,1,1,0,1,1,0,1,0,0,1,1,1},
+    {0,0,0,1,0,1,0,0,0,0,1,1,1,1,0,1,0,1,0,0,1,0,0,1}
+};
\ No newline at end of file
diff --git a/reliability_tests/dmr_matmul/pulp_inject_fault.tcl b/reliability_tests/dmr_matmul/pulp_inject_fault.tcl
new file mode 100644
index 0000000..857d373
--- /dev/null
+++ b/reliability_tests/dmr_matmul/pulp_inject_fault.tcl
@@ -0,0 +1,51 @@
+# Copyright 2023 ETH Zurich and University of Bologna.
+# Solderpad Hardware License, Version 0.51, see LICENSE for details.
+# SPDX-License-Identifier: SHL-0.51
+#
+# Author: Michael Rogenmoser (michaero@iis.ee.ethz.ch)
+
+transcript quietly
+if {! [info exists ::env(VSIM_PATH)]} {error "Define VSIM_PATH"}
+set utils_base_path  [file join $::env(VSIM_PATH) scripts fault_injection_utils]
+set script_base_path [file join $::env(VSIM_PATH) fault_injection_sim scripts]
+
+set verbosity            2
+set log_injections       1
+# Easy way to generate a variable seed
+# set seed                 [clock seconds]
+# Default value
+set seed                 12345
+set print_statistics     1
+
+set inject_start_time 250856000000ps
+set inject_stop_time 413000000000ps
+set injection_clock "pulp_cluster_tb/cluster_i/clk_i"
+set injection_clock_trigger 0
+set fault_period 250
+set rand_initial_injection_phase 0
+# max_num set to 0 means until stop_time
+set max_num_fault_inject 0
+set signal_fault_duration 20ns
+set register_fault_duration 0ns
+
+set allow_multi_bit_upset 1
+set use_bitwidth_as_weight 0
+set check_core_output_modification 0
+set check_core_next_state_modification 0
+set reg_to_sig_ratio 1
+
+source [file join $utils_base_path pulp_extract_nets.tcl]
+
+set inject_signals_netlist []
+set inject_register_netlist []
+set output_netlist []
+set next_state_netlist []
+set assertion_disable_list []
+
+for {set idx 0} {$idx < 12} {incr idx} {
+    set inject_signals_netlist [list {*}$inject_signals_netlist {*}[get_all_core_nets $idx]]
+    set output_netlist [list {*}$output_netlist {*}[get_core_output_nets $idx]]
+}
+
+source [file join $script_base_path inject_fault.tcl]
+
diff --git a/reliability_tests/ecc_test/Makefile b/reliability_tests/ecc_test/Makefile
new file mode 100755
index 0000000..d19f286
--- /dev/null
+++ b/reliability_tests/ecc_test/Makefile
@@ -0,0 +1,15 @@
+PULP_APP = test
+PULP_APP_SRCS = ecc_test.c
+
+PULP_CFLAGS = -O3
+
+ifeq ($(fault_inject),1)
+	export FAULT_INJECTION=1
+	export FAULT_INJECTION_SCRIPT=$(CURDIR)/pulp_inject_fault.tcl
+endif
+
+ifeq ($(multi_bit_upset),1)
+	export MULTI_BIT_UPSET=1
+endif
+
+include $(PULP_SDK_HOME)/install/rules/pulp.mk
diff --git a/reliability_tests/ecc_test/ecc_test.c b/reliability_tests/ecc_test/ecc_test.c
new file mode 100644
index 0000000..99ee1cf
--- /dev/null
+++ b/reliability_tests/ecc_test/ecc_test.c
@@ -0,0 +1,69 @@
+/*
+ * Copyright (C) 2018 ETH Zurich and University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include <pulp.h>
+#include <stdint.h>
+#include <stdio.h>
+#include <stdlib.h>
+
+#define SIZE 0x400
+#define NUM_BANKS 16
+#define SCRUBBER_INTERVAL 2
+
+int main() {
+  // Collecting info about the core ID and the running cluster ID
+  unsigned int core_id = get_core_id();
+  unsigned int cluster_id = rt_cluster_id();
+
+  if (rt_cluster_id() != 0) return bench_cluster_forward(0);
+
+  if (core_id != 0) synch_barrier();
+
+  unsigned int *test_array = pi_l1_malloc(cluster_id, SIZE);
+
+  // Initializing the memory
+  for (int i = 0; i < SIZE; i++) {
+    pulp_write32(&test_array[i], i);
+  }
+
+  // Initialize the scrubbing interval for all memory banks
+  for (int i = 0; i < NUM_BANKS; i++)
+    tcdm_scrubber_set_interval(cluster_id, i, SCRUBBER_INTERVAL);
+
+  // Initialize the error-tracking variables
+  bool  mismatch = 0;
+  unsigned int error = 0;
+  for (int i = 0; i < SIZE; i++) {
+    mismatch = (pulp_read32(&test_array[i]) != i);
+    if (mismatch) {
+      error ++;
+      printf("Expected 0x%x, got 0x%x\n", i, pulp_read32(&test_array[i]));
+    }
+  }
+
+  unsigned int mismatch_cnt = 0;
+  unsigned int fix_cnt = 0;
+  unsigned int uncorrectable_cnt = 0;
+  for (int i = 0; i < 16; i++) {
+    mismatch_cnt += tcdm_scrubber_get_mismatch_count(cluster_id, i);
+    fix_cnt += tcdm_scrubber_get_fix_count(cluster_id, i);
+    uncorrectable_cnt += tcdm_scrubber_get_uncorrectable_count(cluster_id, i);
+  }
+
+  printf("mismatch_cnt: %d, fix_cnt: %d, uncorrectable_cnt: %d\n", mismatch_cnt, fix_cnt, uncorrectable_cnt);
+
+  return (error != 0) && (uncorrectable_cnt == 0);
+}
diff --git a/reliability_tests/ecc_test/pulp_inject_fault.tcl b/reliability_tests/ecc_test/pulp_inject_fault.tcl
new file mode 100644
index 0000000..45cef47
--- /dev/null
+++ b/reliability_tests/ecc_test/pulp_inject_fault.tcl
@@ -0,0 +1,53 @@
+# Copyright 2023 ETH Zurich and University of Bologna.
+# Solderpad Hardware License, Version 0.51, see LICENSE for details.
+# SPDX-License-Identifier: SHL-0.51
+#
+# Author: Michael Rogenmoser (michaero@iis.ee.ethz.ch)
+
+transcript quietly
+if {! [info exists ::env(VSIM_PATH)]} {error "Define VSIM_PATH"}
+set utils_base_path  [file join $::env(VSIM_PATH) scripts fault_injection_utils]
+set script_base_path [file join $::env(VSIM_PATH) fault_injection_sim scripts]
+
+set verbosity            2
+set log_injections       1
+# Easy way to generate a variable seed
+# set seed                 [clock seconds]
+# Default value
+set seed                 12345
+set print_statistics     1
+
+set inject_start_time 110584000000ps
+set inject_stop_time 203880000000ps
+set injection_clock "pulp_cluster_tb/cluster_i/clk_i"
+set injection_clock_trigger 0
+set fault_period 100
+set rand_initial_injection_phase 0
+# max_num set to 0 means until stop_time
+set max_num_fault_inject 0
+set signal_fault_duration 20ns
+set register_fault_duration 0ns
+
+set allow_multi_bit_upset $::env(MULTI_BIT_UPSET)
+set use_bitwidth_as_weight 0
+set check_core_output_modification 0
+set check_core_next_state_modification 0
+set reg_to_sig_ratio 1
+
+source [file join $utils_base_path pulp_extract_nets.tcl]
+
+set inject_signals_netlist []
+set inject_register_netlist []
+set output_netlist []
+set next_state_netlist []
+set assertion_disable_list []
+
+# for {set idx 0} {$idx < 12} {incr idx} {
+#     set inject_signals_netlist [list {*}$inject_signals_netlist {*}[get_all_core_nets $idx]]
+#     set output_netlist [list {*}$output_netlist {*}[get_core_output_nets $idx]]
+# }
+
+set inject_register_netlist [list {*}$inject_register_netlist {*}[get_memory_slice {0 16} {385 449}]]
+
+source [file join $script_base_path inject_fault.tcl]
+
diff --git a/reliability_tests/icache_fi_conv16/Makefile b/reliability_tests/icache_fi_conv16/Makefile
new file mode 100644
index 0000000..58620aa
--- /dev/null
+++ b/reliability_tests/icache_fi_conv16/Makefile
@@ -0,0 +1,11 @@
+PULP_APP = test
+PULP_APP_SRCS = icache_conv16.c
+
+PULP_CFLAGS = -O3
+
+ifeq ($(fault_inject),1)
+	export FAULT_INJECTION=1
+	export FAULT_INJECTION_SCRIPT=$(CURDIR)/pulp_inject_fault.tcl
+endif
+
+include $(PULP_SDK_HOME)/install/rules/pulp.mk
diff --git a/reliability_tests/icache_fi_conv16/conv16.h b/reliability_tests/icache_fi_conv16/conv16.h
new file mode 100644
index 0000000..b75c587
--- /dev/null
+++ b/reliability_tests/icache_fi_conv16/conv16.h
@@ -0,0 +1,397 @@
+/*
+ * Copyright (C) 2018 ETH Zurich and University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+/* 
+ * Mantainer: Luca Valente, luca.valente2@unibo.it
+ */
+/******************************************************************************
+ *                                                                            *
+ *             Multitherman Lab @ DEI - University of Bologna                 *
+ *                     Viale Risorgimento 2 40136                             *
+ *                     Bologna - phone 0512092759                             *
+ *                                                                            *
+ * Engineer:       Francesco Conti - f.conti@unibo.it                         *
+ *                                                                            *
+ * Project:        CConvNet                                                   *
+ * File:           conv16.h                                                   *
+ * Description:    16-bit fixed point convolution test                        *
+ *                                                                            *
+ ******************************************************************************/
+
+#ifndef _CONV16_H
+#define _CONV16_H
+
+// uncomment if doing test with SPI (disables printf's)
+// #define PULP_SPI
+
+// fractionary bits
+#define QF 13
+
+// uncomment if checking errors in detail
+#define CHECK_ERROR
+
+// uncomment if performing checks at all
+#define CHECK_CHECKSUM
+
+#define IMPRECISE_ASM5
+
+#define FIXED_MUL(a,b) ((a*b) >> QF);
+
+#define IH 32
+#define IW 32
+#define FH 5
+#define FW 5
+#define OH (IH-FH+1)
+#define OW (IW-FW+1)
+
+// right checksum
+#define RIGHT_CHECKSUM 0x009da8b0
+
+#define FIXED_MUL_ASM(W_ptr,x_ptr,conv) \
+__asm__ volatile \
+( \
+   "l.lhs  r28,0x0(%0);    " \
+   "l.addi %0,%0,0x4;      " \
+   "l.addi %1,%1,0xFFFC;   " \
+   "l.mul  r28,r28,r29;    " \
+   "l.srli r28,r28,0xD;    " \
+   "l.add  %2,%2,r28;      " \
+   : "=r"(W_ptr), "=r"(x_ptr), "=&r"(conv) \
+   : "r"(W_ptr), "r"(x_ptr)                \
+   : "r28", "r29", "cc"                    \
+)
+
+#define FIXED_MUL_ASM5_IMPRECISE(W_ptr,x_ptr,conv) \
+__asm__ volatile \
+( \
+   "l.ori  r25,r0,0xffff;  " \
+   "l.ori  r26,r0,16;      " \
+   "l.andi r27,r27,0x0000; " \
+   "l.lwz  r28,0x0000(%0); " \
+   "l.lwz  r29,0x0000(%1); " \
+   "l.ror  r30,r28,r26;    " \
+   "l.ror  r31,r29,r26;    " \
+   "l.and  r30,r30,r25;    " \
+   "l.and  r31,r31,r25;    " \
+   "l.mul  r30,r30,r31;    " \
+   "l.add  r27,r27,r30;    " \
+   "l.and  r28,r28,r25;    " \
+   "l.and  r29,r29,r25;    " \
+   "l.mul  r28,r28,r29;    " \
+   "l.add  r27,r27,r28;    " \
+   "l.lwz  r28,0x0004(%0); " \
+   "l.lwz  r29,0xfffc(%1); " \
+   "l.ror  r30,r28,r26;    " \
+   "l.ror  r31,r29,r26;    " \
+   "l.and  r30,r30,r25;    " \
+   "l.and  r31,r31,r25;    " \
+   "l.mul  r30,r30,r31;    " \
+   "l.add  r27,r27,r30;    " \
+   "l.and  r28,r28,r25;    " \
+   "l.and  r29,r29,r25;    " \
+   "l.mul  r28,r28,r29;    " \
+   "l.add  r27,r27,r28;    " \
+   "l.lwz  r28,0x0008(%0); " \
+   "l.lwz  r29,0xfff8(%1); " \
+   "l.ror  r31,r29,r26;    " \
+   "l.and  r28,r28,r25;    " \
+   "l.and  r31,r31,r25;    " \
+   "l.mul  r28,r28,r31;    " \
+   "l.add  r27,r27,r28;    " \
+   "l.srli r27,r27,0xD;    " \
+   "l.add  %2,%2,r27;      " \
+   : "=r"(W_ptr), "=r"(x_ptr), "=&r"(conv) \
+   : "r"(W_ptr), "r"(x_ptr), "r"(conv)     \
+   : "r27", "r28", "r29", "cc"             \
+)
+
+#define FIXED_MUL_ASM5_PRECISE(W_ptr,x_ptr,conv) \
+__asm__ volatile \
+( \
+   "l.lhs  r28,0x0000(%0); " \
+   "l.lhs  r29,0x0000(%1); " \
+   "l.mul  r28,r28,r29;    " \
+   "l.srli r28,r28,0xD;    " \
+   "l.add  %2,%2,r28;      " \
+   "l.lhs  r28,0x0002(%0); " \
+   "l.lhs  r29,0xfffe(%1); " \
+   "l.mul  r28,r28,r29;    " \
+   "l.srli r28,r28,0xD;    " \
+   "l.add  %2,%2,r28;      " \
+   "l.lhs  r28,0x0004(%0); " \
+   "l.lhs  r29,0xfffc(%1); " \
+   "l.mul  r28,r28,r29;    " \
+   "l.srli r28,r28,0xD;    " \
+   "l.add  %2,%2,r28;      " \
+   "l.lhs  r28,0x0006(%0); " \
+   "l.lhs  r29,0xfffa(%1); " \
+   "l.mul  r28,r28,r29;    " \
+   "l.srli r28,r28,0xD;    " \
+   "l.add  %2,%2,r28;      " \
+   "l.lhs  r28,0x0008(%0); " \
+   "l.lhs  r29,0xfff8(%1); " \
+   "l.mul  r28,r28,r29;    " \
+   "l.srli r28,r28,0xD;    " \
+   "l.add  %2,%2,r28;      " \
+   : "=r"(W_ptr), "=r"(x_ptr), "=&r"(conv) \
+   : "r"(W_ptr), "r"(x_ptr)                \
+   : "r27", "r28", "r29", "cc"             \
+)
+
+// #define FIXED_MUL_ASM25_PRECISE(W_ptr,x_ptr,conv) \
+// __asm__ volatile \
+// ( \
+//    "l.lwz  r28,0x0000(%0); " \
+//    "l.lwz  r29,0x0000(%1); " \
+//    "l.mul  r28,r28,r29;    " \
+//    "l.srli r28,r28,0xD;    " \
+//    "l.add  %2,%2,r28;      " \
+//    "l.lwz  r28,0x0004(%0); " \
+//    "l.lwz  r29,0xfffc(%1); " \
+//    "l.mul  r28,r28,r29;    " \
+//    "l.srli r28,r28,0xD;    " \
+//    "l.add  %2,%2,r28;      " \
+//    "l.lwz  r28,0x0008(%0); " \
+//    "l.lwz  r29,0xfff8(%1); " \
+//    "l.mul  r28,r28,r29;    " \
+//    "l.srli r28,r28,0xD;    " \
+//    "l.add  %2,%2,r28;      " \
+//    "l.lwz  r28,0x000c(%0); " \
+//    "l.lwz  r29,0xfff4(%1); " \
+//    "l.mul  r28,r28,r29;    " \
+//    "l.srli r28,r28,0xD;    " \
+//    "l.add  %2,%2,r28;      " \
+//    "l.lwz  r28,0x0010(%0); " \
+//    "l.lwz  r29,0xfff0(%1); " \
+//    "l.mul  r28,r28,r29;    " \
+//    "l.srli r28,r28,0xD;    " \
+//    "l.add  %2,%2,r28;      " \
+//    \
+//    "l.slli %1,r28,0x2;    " \
+//    \
+//    "l.lwz  r28,0x0014(%0); " \
+//    "l.lwz  r29,0x0000(%1); " \
+//    "l.mul  r28,r28,r29;    " \
+//    "l.srli r28,r28,0xD;    " \
+//    "l.add  %2,%2,r28;      " \
+//    "l.lwz  r28,0x0018(%0); " \
+//    "l.lwz  r29,0xfffc(%1); " \
+//    "l.mul  r28,r28,r29;    " \
+//    "l.srli r28,r28,0xD;    " \
+//    "l.add  %2,%2,r28;      " \
+//    "l.lwz  r28,0x001c(%0); " \
+//    "l.lwz  r29,0xfff8(%1); " \
+//    "l.mul  r28,r28,r29;    " \
+//    "l.srli r28,r28,0xD;    " \
+//    "l.add  %2,%2,r28;      " \
+//    "l.lwz  r28,0x0020(%0); " \
+//    "l.lwz  r29,0xfff4(%1); " \
+//    "l.mul  r28,r28,r29;    " \
+//    "l.srli r28,r28,0xD;    " \
+//    "l.add  %2,%2,r28;      " \
+//    "l.lwz  r28,0x0024(%0); " \
+//    "l.lwz  r29,0xfff0(%1); " \
+//    "l.mul  r28,r28,r29;    " \
+//    "l.srli r28,r28,0xD;    " \
+//    "l.add  %2,%2,r28;      " \
+//    // WIP
+//    : "=r"(W_ptr), "=r"(x_ptr), "=&r"(conv)   \
+//    : "r"(W_ptr), "r"(x_ptr), "r"(w), "r"(fw) \
+//    : "r27", "r28", "r29", "cc"               \
+// )
+
+#define XPTR_UPDATE(x_ptr,w,fw) \
+__asm__ volatile \
+( \
+   "l.sub  r28,%1,%2;      " \
+   "l.slli r28,r28,0x2;    " \
+   "l.sub  %0,%0,r28;      " \
+   : "=&r"(x_ptr) \
+   : "r"(x_ptr), "r"(w), "r"(fw) \
+   : "r28", "cc" \
+)
+
+int test_singlethread(void (*test)(int16_t *, int16_t *, int16_t *, int, int, int, int, int, int, int, int, int), char *str);
+int test_multithread(void (*test)(int16_t *, int16_t *, int16_t *, int, int, int, int, int, int, int, int, int), char *str);
+
+void conv16_gold(int16_t *__restrict__ W, int16_t *__restrict__ x, int16_t *__restrict__ y, int h, int w, int fh, int fw, int oh, int ow, int nif, int a, int b);
+void conv16_asm_mul(int16_t *__restrict__ W, int16_t *__restrict__ x, int16_t *__restrict__ y, int h, int w, int fh, int fw, int oh, int ow, int nif, int a, int b);
+// void conv16_asm_mac(int16_t *__restrict__ W, int16_t *__restrict__ x, int16_t *__restrict__ y, int h, int w, int fh, int fw, int oh, int ow, int nif, int a, int b);
+void conv16_asm_mul_unrolled_5x5(int16_t *__restrict__ W, int16_t *__restrict__ x, int16_t *__restrict__ y, int h, int w, int fh, int fw, int oh, int ow, int nif, int a, int b);
+void conv16_unrolled_5x5(int16_t *__restrict__ W, int16_t *__restrict__ x, int16_t *__restrict__ y, int h, int w, int fh, int fw, int oh, int ow, int nif, int a, int b);
+void conv16_unrolled_ptr_5x5(int16_t *__restrict__ W, int16_t *__restrict__ x, int16_t *__restrict__ y, int h, int w, int fh, int fw, int oh, int ow, int nif, int a, int b);
+// void conv16_gold_four_finest(int16_t *__restrict__ W, int16_t *__restrict__ x, int16_t *__restrict__ y, int h, int w, int fh, int fw, int oh, int ow, int nif, int a, int b, int32_t *shared_conv);
+// void conv16_gold_four_fine(int16_t *__restrict__ W, int16_t *__restrict__ x, int16_t *__restrict__ y, int h, int w, int fh, int fw, int oh, int ow, int nif, int a, int b);
+void conv16_gold_four_coarse(int16_t *__restrict__ W, int16_t *__restrict__ x, int16_t *__restrict__ y, int h, int w, int fh, int fw, int oh, int ow, int nif, int a, int b);
+void conv16_unrolled_5x5_four_coarse(int16_t *__restrict__ W, int16_t *__restrict__ x, int16_t *__restrict__ y, int h, int w, int fh, int fw, int oh, int ow, int nif, int a, int b);
+void conv16_unrolled_ptr_5x5_four_coarse(int16_t *__restrict__ W, int16_t *__restrict__ x, int16_t *__restrict__ y, int h, int w, int fh, int fw, int oh, int ow, int nif, int a, int b);
+void conv16_gold_four_coarsest(int16_t *__restrict__ W, int16_t *__restrict__ x, int16_t *__restrict__ y, int h, int w, int fh, int fw, int oh, int ow, int nif, int a, int b);
+void conv16_unrolled_5x5_four_coarsest(int16_t *__restrict__ W, int16_t *__restrict__ x, int16_t *__restrict__ y, int h, int w, int fh, int fw, int oh, int ow, int nif, int a, int b);
+void conv16_unrolled_ptr_5x5_four_coarsest(int16_t *__restrict__ W, int16_t *__restrict__ x, int16_t *__restrict__ y, int h, int w, int fh, int fw, int oh, int ow, int nif, int a, int b);
+void conv16_asm_mul_unrolled_5x5_four_coarsest(int16_t *__restrict__ W, int16_t *__restrict__ x, int16_t *__restrict__ y, int h, int w, int fh, int fw, int oh, int ow, int nif, int a, int b);
+
+void load();
+int check(int16_t *y);
+int checksum(int16_t *y);
+
+int16_t correct_yout[OH*OW] = {
+   0x0352,
+   0x036c,
+   0x0386,
+   0x03a0,
+   0x03ba,
+   0x03d4,
+   0x03ee,
+   0x0408,
+   0x0422,
+   0x043c,
+   0x0456,
+   0x0470,
+   0x04ee,
+   0x0508,
+   0x0522,
+   0x053c,
+   0x0556,
+   0x0570,
+   0x058a,
+   0x05a4,
+   0x05be,
+   0x05d8,
+   0x05f2,
+   0x060c,
+   0x068a,
+   0x06a4,
+   0x06be,
+   0x06d8,
+   0x06f2,
+   0x070c,
+   0x0726,
+   0x0740,
+   0x075a,
+   0x0774,
+   0x078e,
+   0x07a8,
+   0x0826,
+   0x0840,
+   0x085a,
+   0x0874,
+   0x088e,
+   0x08a8,
+   0x08c2,
+   0x08dc,
+   0x08f6,
+   0x0910,
+   0x092a,
+   0x0944,
+   0x09c2,
+   0x09dc,
+   0x09f6,
+   0x0a10,
+   0x0a2a,
+   0x0a44,
+   0x0a5e,
+   0x0a78,
+   0x0a92,
+   0x0aac,
+   0x0ac6,
+   0x0ae0,
+   0x0b5e,
+   0x0b78,
+   0x0b92,
+   0x0bac,
+   0x0bc6,
+   0x0be0,
+   0x0bfa,
+   0x0c14,
+   0x0c2e,
+   0x0c48,
+   0x0c62,
+   0x0c7c,
+   0x0cfa,
+   0x0d14,
+   0x0d2e,
+   0x0d48,
+   0x0d62,
+   0x0d7c,
+   0x0d96,
+   0x0db0,
+   0x0dca,
+   0x0de4,
+   0x0dfe,
+   0x0e18,
+   0x0e96,
+   0x0eb0,
+   0x0eca,
+   0x0ee4,
+   0x0efe,
+   0x0f18,
+   0x0f32,
+   0x0f4c,
+   0x0f66,
+   0x0f80,
+   0x0f9a,
+   0x0fb4,
+   0x1032,
+   0x104c,
+   0x1066,
+   0x1080,
+   0x109a,
+   0x10b4,
+   0x10ce,
+   0x10e8,
+   0x1102,
+   0x111c,
+   0x1136,
+   0x1150,
+   0x11ce,
+   0x11e8,
+   0x1202,
+   0x121c,
+   0x1236,
+   0x1250,
+   0x126a,
+   0x1284,
+   0x129e,
+   0x12b8,
+   0x12d2,
+   0x12ec,
+   0x136a,
+   0x1384,
+   0x139e,
+   0x13b8,
+   0x13d2,
+   0x13ec,
+   0x1406,
+   0x1420,
+   0x143a,
+   0x1454,
+   0x146e,
+   0x1488,
+   0x1506,
+   0x1520,
+   0x153a,
+   0x1554,
+   0x156e,
+   0x1588,
+   0x15a2,
+   0x15bc,
+   0x15d6,
+   0x15f0,
+   0x160a,
+   0x1624,
+};
+
+#endif
diff --git a/reliability_tests/icache_fi_conv16/icache_conv16.c b/reliability_tests/icache_fi_conv16/icache_conv16.c
new file mode 100644
index 0000000..e62c440
--- /dev/null
+++ b/reliability_tests/icache_fi_conv16/icache_conv16.c
@@ -0,0 +1,935 @@
+/*
+ * Copyright (C) 2018 ETH Zurich and University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+/* 
+ * Mantainer: Luca Valente, luca.valente2@unibo.it
+ */
+/******************************************************************************
+ *                                                                            *
+ *             Multitherman Lab @ DEI - University of Bologna                 *
+ *                     Viale Risorgimento 2 40136                             *
+ *                     Bologna - phone 0512092759                             *
+ *                                                                            *
+ * Engineer:       Francesco Conti - f.conti@unibo.it                         *
+ *                                                                            *
+ * Project:        CConvNet                                                   *
+ * File:           conv16.c                                                   *
+ * Description:    16-bit fixed point convolution test                        *
+ *                                                                            *
+ ******************************************************************************/
+
+#include <pulp.h>
+#include <stdint.h>
+#include "conv16.h"
+
+__attribute__((section(".heapsram"))) int16_t g_W[FH*FW];
+__attribute__((section(".heapsram"))) int16_t g_x[IH*IW];
+__attribute__((section(".heapsram"))) int16_t g_y[OH*OW];
+__attribute__((section(".heapsram"))) int16_t g_y_in[OH*OW];
+
+int main() {
+
+   if (rt_cluster_id() != 0)
+    return bench_cluster_forward(0);
+ 
+   int errors = 0;
+   int sum = 0;
+
+   // single-threaded "golden" by-the-book convolution
+   errors += test_singlethread(&conv16_gold, "sequential convolution");
+
+   // single-threaded loop-unrolled convolution
+   errors += test_singlethread(&conv16_unrolled_5x5, "sequential loop-unrolled convolution");
+
+   // single-threaded loop-unrolled pointer-optimized convolution
+   errors += test_singlethread(&conv16_unrolled_ptr_5x5, "sequential loop-unrolled pointer-optimized convolution");
+
+   // multi-threaded by-the-book convolution (1 thread per output pixel)
+   errors += test_multithread(&conv16_gold_four_coarse, "4-threaded convolution (1 thread per output pixel)");
+
+   // multi-threaded loop-unrolled convolution (1 thread per output pixel)
+   errors += test_multithread(&conv16_unrolled_5x5_four_coarse, "4-threaded loop-unrolled convolution (1 thread per output pixel)");
+
+   // multi-threaded loop-unrolled pointer-optimized convolution (1 thread per output pixel)
+   errors += test_multithread(&conv16_unrolled_ptr_5x5_four_coarse, "4-threaded loop-unrolled pointer-optimized convolution (1 thread per output pixel)");
+
+   // multi-threaded by-the-book convolution (1 thread per output row)
+   errors += test_multithread(&conv16_gold_four_coarsest, "4-threaded convolution (1 thread per output row)");
+
+   // multi-threaded loop-unrolled convolution (1 thread per output row)
+   errors += test_multithread(&conv16_unrolled_5x5_four_coarsest, "4-threaded loop-unrolled convolution (1 thread per output row)");
+
+   // multi-threaded loop-unrolled pointer-optimized convolution (1 thread per output row)
+   errors += test_multithread(&conv16_unrolled_ptr_5x5_four_coarsest, "4-threaded loop-unrolled pointer-optimized convolution (1 thread per output row)");
+
+   synch_barrier();
+
+   // TODO readout icache errors
+
+   return errors;
+}
+
+int test_singlethread(void (*test)(int16_t *, int16_t *, int16_t *, int, int, int, int, int, int, int, int, int), char *str) {
+   int errors = 0;
+   int sum = 0;
+
+   synch_barrier();
+
+   if(rt_core_id() == 0) {
+      load();
+
+      reset_timer(rt_cluster_id());
+      start_timer(rt_cluster_id());
+      test(g_W, g_x, g_y, IH, IW, FH, FW, OH, OW, 1, 0, 0);
+      stop_timer(rt_cluster_id());
+
+      #ifdef CHECK_CHECKSUM
+      errors = 0;
+      sum = checksum(g_y);
+      if(sum != RIGHT_CHECKSUM) {
+         #ifndef PULP_SPI
+ 	 printf("wrong checksum, 0x%08x instead of 0x%08x\n", sum, RIGHT_CHECKSUM);
+         #ifndef CHECK_ERROR
+         errors += 1;
+         #endif
+         #endif
+         #ifdef CHECK_ERROR
+         errors = check(g_y);
+         #endif
+      }
+      #else
+      errors = -1;
+      #endif
+
+      #ifndef PULP_SPI
+      printf("%s, errors=%d, time=%d\n", str, errors, get_time(rt_cluster_id()));
+      #endif
+      
+   }
+
+   return errors;
+}
+
+int test_multithread(void (*test)(int16_t *, int16_t *, int16_t *, int, int, int, int, int, int, int, int, int), char *str) {
+   int errors = 0;
+   int sum = 0;
+
+   if(rt_core_id() == 0) {
+      load();
+   }
+   
+   synch_barrier();
+
+   if(rt_core_id() == 0) {
+      reset_timer(rt_cluster_id());
+      start_timer(rt_cluster_id());
+   }
+   test(g_W, g_x, g_y, IH, IW, FH, FW, OH, OW, 1, 0, 0);
+   if(rt_core_id() == 0) {
+      stop_timer(rt_cluster_id());
+
+      #ifdef CHECK_CHECKSUM
+      errors = 0;
+      sum = checksum(g_y);
+      if(sum != RIGHT_CHECKSUM) {
+         #ifndef PULP_SPI
+         printf("wrong checksum, 0x%08x instead of 0x%08x\n", sum, RIGHT_CHECKSUM);
+         #ifndef CHECK_ERROR
+         errors += 1;
+         #endif
+         #endif
+         #ifdef CHECK_ERROR
+         errors = check(g_y);
+         #endif
+      }
+      #else
+      errors = -1;
+      #endif
+
+      #ifndef PULP_SPI
+      printf("%s, errors=%d, time=%d\n", str, errors, get_time(rt_cluster_id()));
+      #endif
+      
+   }
+
+   return errors;
+}
+
+void conv16_gold(int16_t *__restrict__ W, int16_t *__restrict__ x, int16_t *__restrict__ y, int h, int w, int fh, int fw, int oh, int ow, int nif, int a, int b) {
+   int i;
+   for (i=0; i<oh; i++) {
+      int j;
+      for (j=0; j<ow; j++) {
+         int32_t conv = 0;
+         int ui;
+         for (ui=0; ui<fh; ui++) {
+            int uj;
+            for (uj=0; uj<fw; uj++) {
+               int m;
+               int n;
+               m = i-ui+fh-1;
+               n = j-uj+fw-1;
+               #ifndef IMPRECISE_ASM5
+               conv += FIXED_MUL(W[((((a*nif)+b)*fh)+ui)*fw+uj] , x[(((b*h)+m)*w)+n]);
+               #else
+               conv += W[((((a*nif)+b)*fh)+ui)*fw+uj] * x[(((b*h)+m)*w)+n];
+               #endif
+            }
+         }
+         #ifndef IMPRECISE_ASM5
+         y[(a*oh+i)*ow+j] = (y[(a*oh+i)*ow+j] + conv);
+         #else
+         y[(a*oh+i)*ow+j] = (y[(a*oh+i)*ow+j] + (conv >> QF));
+         #endif
+      }
+   }
+}
+
+void conv16_unrolled_5x5(int16_t *__restrict__ W, int16_t *__restrict__ x, int16_t *__restrict__ y, int h, int w, int fh, int fw, int oh, int ow, int nif, int a, int b) {
+   int i;
+   for (i=0; i<oh; i++) {
+      int j;
+      for (j=0; j<ow; j++) {
+
+         int32_t conv = 0;
+
+         #ifndef IMPRECISE_ASM5
+
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+0)*5+0], x[(((b*h)+(i+4-0))*w)+(j+4-0)]);
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+0)*5+1], x[(((b*h)+(i+4-0))*w)+(j+4-1)]);
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+0)*5+2], x[(((b*h)+(i+4-0))*w)+(j+4-2)]);
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+0)*5+3], x[(((b*h)+(i+4-0))*w)+(j+4-3)]);
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+0)*5+4], x[(((b*h)+(i+4-0))*w)+(j+4-4)]);
+
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+1)*5+0], x[(((b*h)+(i+4-1))*w)+(j+4-0)]);
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+1)*5+1], x[(((b*h)+(i+4-1))*w)+(j+4-1)]);
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+1)*5+2], x[(((b*h)+(i+4-1))*w)+(j+4-2)]);
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+1)*5+3], x[(((b*h)+(i+4-1))*w)+(j+4-3)]);
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+1)*5+4], x[(((b*h)+(i+4-1))*w)+(j+4-4)]);
+
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+2)*5+0], x[(((b*h)+(i+4-2))*w)+(j+4-0)]);
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+2)*5+1], x[(((b*h)+(i+4-2))*w)+(j+4-1)]);
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+2)*5+2], x[(((b*h)+(i+4-2))*w)+(j+4-2)]);
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+2)*5+3], x[(((b*h)+(i+4-2))*w)+(j+4-3)]);
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+2)*5+4], x[(((b*h)+(i+4-2))*w)+(j+4-4)]);
+
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+3)*5+0], x[(((b*h)+(i+4-3))*w)+(j+4-0)]);
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+3)*5+1], x[(((b*h)+(i+4-3))*w)+(j+4-1)]);
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+3)*5+2], x[(((b*h)+(i+4-3))*w)+(j+4-2)]);
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+3)*5+3], x[(((b*h)+(i+4-3))*w)+(j+4-3)]);
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+3)*5+4], x[(((b*h)+(i+4-3))*w)+(j+4-4)]);
+
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+4)*5+0], x[(((b*h)+(i+4-4))*w)+(j+4-0)]);
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+4)*5+1], x[(((b*h)+(i+4-4))*w)+(j+4-1)]);
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+4)*5+2], x[(((b*h)+(i+4-4))*w)+(j+4-2)]);
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+4)*5+3], x[(((b*h)+(i+4-4))*w)+(j+4-3)]);
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+4)*5+4], x[(((b*h)+(i+4-4))*w)+(j+4-4)]);
+
+         y[(a*oh+i)*ow+j] += conv;
+
+         #else
+
+         conv += W[((((a*nif)+b)*5)+0)*5+0] * x[(((b*h)+(i+4-0))*w)+(j+4-0)];
+         conv += W[((((a*nif)+b)*5)+0)*5+1] * x[(((b*h)+(i+4-0))*w)+(j+4-1)];
+         conv += W[((((a*nif)+b)*5)+0)*5+2] * x[(((b*h)+(i+4-0))*w)+(j+4-2)];
+         conv += W[((((a*nif)+b)*5)+0)*5+3] * x[(((b*h)+(i+4-0))*w)+(j+4-3)];
+         conv += W[((((a*nif)+b)*5)+0)*5+4] * x[(((b*h)+(i+4-0))*w)+(j+4-4)];
+
+         conv += W[((((a*nif)+b)*5)+1)*5+0] * x[(((b*h)+(i+4-1))*w)+(j+4-0)];
+         conv += W[((((a*nif)+b)*5)+1)*5+1] * x[(((b*h)+(i+4-1))*w)+(j+4-1)];
+         conv += W[((((a*nif)+b)*5)+1)*5+2] * x[(((b*h)+(i+4-1))*w)+(j+4-2)];
+         conv += W[((((a*nif)+b)*5)+1)*5+3] * x[(((b*h)+(i+4-1))*w)+(j+4-3)];
+         conv += W[((((a*nif)+b)*5)+1)*5+4] * x[(((b*h)+(i+4-1))*w)+(j+4-4)];
+
+         conv += W[((((a*nif)+b)*5)+2)*5+0] * x[(((b*h)+(i+4-2))*w)+(j+4-0)];
+         conv += W[((((a*nif)+b)*5)+2)*5+1] * x[(((b*h)+(i+4-2))*w)+(j+4-1)];
+         conv += W[((((a*nif)+b)*5)+2)*5+2] * x[(((b*h)+(i+4-2))*w)+(j+4-2)];
+         conv += W[((((a*nif)+b)*5)+2)*5+3] * x[(((b*h)+(i+4-2))*w)+(j+4-3)];
+         conv += W[((((a*nif)+b)*5)+2)*5+4] * x[(((b*h)+(i+4-2))*w)+(j+4-4)];
+
+         conv += W[((((a*nif)+b)*5)+3)*5+0] * x[(((b*h)+(i+4-3))*w)+(j+4-0)];
+         conv += W[((((a*nif)+b)*5)+3)*5+1] * x[(((b*h)+(i+4-3))*w)+(j+4-1)];
+         conv += W[((((a*nif)+b)*5)+3)*5+2] * x[(((b*h)+(i+4-3))*w)+(j+4-2)];
+         conv += W[((((a*nif)+b)*5)+3)*5+3] * x[(((b*h)+(i+4-3))*w)+(j+4-3)];
+         conv += W[((((a*nif)+b)*5)+3)*5+4] * x[(((b*h)+(i+4-3))*w)+(j+4-4)];
+
+         conv += W[((((a*nif)+b)*5)+4)*5+0] * x[(((b*h)+(i+4-4))*w)+(j+4-0)];
+         conv += W[((((a*nif)+b)*5)+4)*5+1] * x[(((b*h)+(i+4-4))*w)+(j+4-1)];
+         conv += W[((((a*nif)+b)*5)+4)*5+2] * x[(((b*h)+(i+4-4))*w)+(j+4-2)];
+         conv += W[((((a*nif)+b)*5)+4)*5+3] * x[(((b*h)+(i+4-4))*w)+(j+4-3)];
+         conv += W[((((a*nif)+b)*5)+4)*5+4] * x[(((b*h)+(i+4-4))*w)+(j+4-4)];
+
+         y[(a*oh+i)*ow+j] += (conv >> QF);
+
+         #endif
+
+      }
+   }
+}
+
+void conv16_unrolled_ptr_5x5(int16_t *__restrict__ W, int16_t *__restrict__ x, int16_t *__restrict__ y, int h, int w, int fh, int fw, int oh, int ow, int nif, int a, int b) {
+   register int i;
+   register int j;
+
+   int16_t *y_ptr = y + a*oh*ow;
+   int16_t *x_base = x + b*h*w + (fh-1)*w + (fw-1);
+   int16_t *W_base = &W[((a*nif)+b)*fh*fw];
+   int16_t *W_ptr;
+   int16_t *x_ptr;
+
+   for (i=0; i<oh; i++) {
+      for (j=0; j<ow; j++) {
+
+         int conv = 0;
+
+         W_ptr = W_base;
+         x_ptr = x_base + i*w + j;
+
+         #ifndef IMPRECISE_ASM5
+
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         x_ptr -= w-5;
+
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         x_ptr -= w-5;
+
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         x_ptr -= w-5;
+
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         x_ptr -= w-5;
+
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         x_ptr -= w-5;
+
+         *(y_ptr + i*ow+j) += conv;
+
+         #else
+
+         conv += *W_ptr++ * *x_ptr--;
+         conv += *W_ptr++ * *x_ptr--;
+         conv += *W_ptr++ * *x_ptr--;
+         conv += *W_ptr++ * *x_ptr--;
+         conv += *W_ptr++ * *x_ptr--;
+         x_ptr -= w-5;
+
+         conv += *W_ptr++ * *x_ptr--;
+         conv += *W_ptr++ * *x_ptr--;
+         conv += *W_ptr++ * *x_ptr--;
+         conv += *W_ptr++ * *x_ptr--;
+         conv += *W_ptr++ * *x_ptr--;
+         x_ptr -= w-5;
+
+         conv += *W_ptr++ * *x_ptr--;
+         conv += *W_ptr++ * *x_ptr--;
+         conv += *W_ptr++ * *x_ptr--;
+         conv += *W_ptr++ * *x_ptr--;
+         conv += *W_ptr++ * *x_ptr--;
+         x_ptr -= w-5;
+
+         conv += *W_ptr++ * *x_ptr--;
+         conv += *W_ptr++ * *x_ptr--;
+         conv += *W_ptr++ * *x_ptr--;
+         conv += *W_ptr++ * *x_ptr--;
+         conv += *W_ptr++ * *x_ptr--;
+         x_ptr -= w-5;
+
+         conv += *W_ptr++ * *x_ptr--;
+         conv += *W_ptr++ * *x_ptr--;
+         conv += *W_ptr++ * *x_ptr--;
+         conv += *W_ptr++ * *x_ptr--;
+         conv += *W_ptr++ * *x_ptr--;
+         x_ptr -= w-5;
+
+         *(y_ptr + i*ow+j) += conv >> QF;
+
+         #endif
+
+      }
+   }
+
+}
+
+void conv16_gold_four_coarse(int16_t *__restrict__ W, int16_t *__restrict__ x, int16_t *__restrict__ y, int h, int w, int fh, int fw, int oh, int ow, int nif, int a, int b) {
+   int i;
+   register int myid      = rt_core_id();
+   register int num_cores = get_core_num();
+   // synch_barrier();
+   for (i=0; i<oh; i++) {
+      int j;
+      for (j=myid; j<ow; j+=num_cores) {
+         int ui;
+         int32_t conv = 0;
+         for (ui=0; ui<fh; ui++) {
+            int uj;
+            for (uj=0; uj<fw; uj++) {
+               int m;
+               int n;
+               m = i-ui+fh-1;
+               n = j-uj+fw-1;
+               #ifndef IMPRECISE_ASM5
+               conv += FIXED_MUL(W[((((a*nif)+b)*fh)+ui)*fw+uj] , x[(((b*h)+m)*w)+n]);
+               #else
+               conv += W[((((a*nif)+b)*fh)+ui)*fw+uj] * x[(((b*h)+m)*w)+n];
+               #endif
+            }
+         }
+         #ifndef IMPRECISE_ASM5
+         y[(a*oh+i)*ow+j] = (y[(a*oh+i)*ow+j] + conv);
+         #else
+         y[(a*oh+i)*ow+j] = (y[(a*oh+i)*ow+j] + (conv >> QF));
+         #endif
+      }
+   }
+   synch_barrier();
+}
+
+void conv16_gold_four_coarsest(int16_t *__restrict__ W, int16_t *__restrict__ x, int16_t *__restrict__ y, int h, int w, int fh, int fw, int oh, int ow, int nif, int a, int b) {
+   int i;
+   register int myid      = rt_core_id();
+   register int num_cores = get_core_num();
+   // synch_barrier();
+   for (i=myid; i<oh; i+=num_cores) {
+      int j;
+      for (j=0; j<ow; j++) {
+         int ui;
+         int32_t conv = 0;
+         for (ui=0; ui<fh; ui++) {
+            int uj;
+            for (uj=0; uj<fw; uj++) {
+               int m;
+               int n;
+               m = i-ui+fh-1;
+               n = j-uj+fw-1;
+               #ifndef IMPRECISE_ASM5
+               conv += FIXED_MUL(W[((((a*nif)+b)*fh)+ui)*fw+uj] , x[(((b*h)+m)*w)+n]);
+               #else
+               conv += W[((((a*nif)+b)*fh)+ui)*fw+uj] * x[(((b*h)+m)*w)+n];
+               #endif
+            }
+         }
+         #ifndef IMPRECISE_ASM5
+         y[(a*oh+i)*ow+j] = (y[(a*oh+i)*ow+j] + conv);
+         #else
+         y[(a*oh+i)*ow+j] = (y[(a*oh+i)*ow+j] + (conv >> QF));
+         #endif
+      }
+   }
+   synch_barrier();
+}
+
+void conv16_unrolled_5x5_four_coarse(int16_t *__restrict__ W, int16_t *__restrict__ x, int16_t *__restrict__ y, int h, int w, int fh, int fw, int oh, int ow, int nif, int a, int b) {
+   int i;
+   register int myid      = rt_core_id();
+   register int num_cores = get_core_num();
+   // synch_barrier();
+   for (i=0; i<oh; i++) {
+      int j;
+      for (j=myid; j<ow; j+=num_cores) {
+
+         int32_t conv = 0;
+
+         #ifndef IMPRECISE_ASM5
+
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+0)*5+0], x[(((b*h)+(i+4-0))*w)+(j+4-0)]);
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+0)*5+1], x[(((b*h)+(i+4-0))*w)+(j+4-1)]);
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+0)*5+2], x[(((b*h)+(i+4-0))*w)+(j+4-2)]);
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+0)*5+3], x[(((b*h)+(i+4-0))*w)+(j+4-3)]);
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+0)*5+4], x[(((b*h)+(i+4-0))*w)+(j+4-4)]);
+
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+1)*5+0], x[(((b*h)+(i+4-1))*w)+(j+4-0)]);
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+1)*5+1], x[(((b*h)+(i+4-1))*w)+(j+4-1)]);
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+1)*5+2], x[(((b*h)+(i+4-1))*w)+(j+4-2)]);
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+1)*5+3], x[(((b*h)+(i+4-1))*w)+(j+4-3)]);
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+1)*5+4], x[(((b*h)+(i+4-1))*w)+(j+4-4)]);
+
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+2)*5+0], x[(((b*h)+(i+4-2))*w)+(j+4-0)]);
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+2)*5+1], x[(((b*h)+(i+4-2))*w)+(j+4-1)]);
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+2)*5+2], x[(((b*h)+(i+4-2))*w)+(j+4-2)]);
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+2)*5+3], x[(((b*h)+(i+4-2))*w)+(j+4-3)]);
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+2)*5+4], x[(((b*h)+(i+4-2))*w)+(j+4-4)]);
+
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+3)*5+0], x[(((b*h)+(i+4-3))*w)+(j+4-0)]);
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+3)*5+1], x[(((b*h)+(i+4-3))*w)+(j+4-1)]);
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+3)*5+2], x[(((b*h)+(i+4-3))*w)+(j+4-2)]);
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+3)*5+3], x[(((b*h)+(i+4-3))*w)+(j+4-3)]);
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+3)*5+4], x[(((b*h)+(i+4-3))*w)+(j+4-4)]);
+
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+4)*5+0], x[(((b*h)+(i+4-4))*w)+(j+4-0)]);
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+4)*5+1], x[(((b*h)+(i+4-4))*w)+(j+4-1)]);
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+4)*5+2], x[(((b*h)+(i+4-4))*w)+(j+4-2)]);
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+4)*5+3], x[(((b*h)+(i+4-4))*w)+(j+4-3)]);
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+4)*5+4], x[(((b*h)+(i+4-4))*w)+(j+4-4)]);
+
+         y[(a*oh+i)*ow+j] += conv; // because i'm using 32-bit int
+
+         #else
+
+         conv += W[((((a*nif)+b)*5)+0)*5+0] * x[(((b*h)+(i+4-0))*w)+(j+4-0)];
+         conv += W[((((a*nif)+b)*5)+0)*5+1] * x[(((b*h)+(i+4-0))*w)+(j+4-1)];
+         conv += W[((((a*nif)+b)*5)+0)*5+2] * x[(((b*h)+(i+4-0))*w)+(j+4-2)];
+         conv += W[((((a*nif)+b)*5)+0)*5+3] * x[(((b*h)+(i+4-0))*w)+(j+4-3)];
+         conv += W[((((a*nif)+b)*5)+0)*5+4] * x[(((b*h)+(i+4-0))*w)+(j+4-4)];
+
+         conv += W[((((a*nif)+b)*5)+1)*5+0] * x[(((b*h)+(i+4-1))*w)+(j+4-0)];
+         conv += W[((((a*nif)+b)*5)+1)*5+1] * x[(((b*h)+(i+4-1))*w)+(j+4-1)];
+         conv += W[((((a*nif)+b)*5)+1)*5+2] * x[(((b*h)+(i+4-1))*w)+(j+4-2)];
+         conv += W[((((a*nif)+b)*5)+1)*5+3] * x[(((b*h)+(i+4-1))*w)+(j+4-3)];
+         conv += W[((((a*nif)+b)*5)+1)*5+4] * x[(((b*h)+(i+4-1))*w)+(j+4-4)];
+
+         conv += W[((((a*nif)+b)*5)+2)*5+0] * x[(((b*h)+(i+4-2))*w)+(j+4-0)];
+         conv += W[((((a*nif)+b)*5)+2)*5+1] * x[(((b*h)+(i+4-2))*w)+(j+4-1)];
+         conv += W[((((a*nif)+b)*5)+2)*5+2] * x[(((b*h)+(i+4-2))*w)+(j+4-2)];
+         conv += W[((((a*nif)+b)*5)+2)*5+3] * x[(((b*h)+(i+4-2))*w)+(j+4-3)];
+         conv += W[((((a*nif)+b)*5)+2)*5+4] * x[(((b*h)+(i+4-2))*w)+(j+4-4)];
+
+         conv += W[((((a*nif)+b)*5)+3)*5+0] * x[(((b*h)+(i+4-3))*w)+(j+4-0)];
+         conv += W[((((a*nif)+b)*5)+3)*5+1] * x[(((b*h)+(i+4-3))*w)+(j+4-1)];
+         conv += W[((((a*nif)+b)*5)+3)*5+2] * x[(((b*h)+(i+4-3))*w)+(j+4-2)];
+         conv += W[((((a*nif)+b)*5)+3)*5+3] * x[(((b*h)+(i+4-3))*w)+(j+4-3)];
+         conv += W[((((a*nif)+b)*5)+3)*5+4] * x[(((b*h)+(i+4-3))*w)+(j+4-4)];
+
+         conv += W[((((a*nif)+b)*5)+4)*5+0] * x[(((b*h)+(i+4-4))*w)+(j+4-0)];
+         conv += W[((((a*nif)+b)*5)+4)*5+1] * x[(((b*h)+(i+4-4))*w)+(j+4-1)];
+         conv += W[((((a*nif)+b)*5)+4)*5+2] * x[(((b*h)+(i+4-4))*w)+(j+4-2)];
+         conv += W[((((a*nif)+b)*5)+4)*5+3] * x[(((b*h)+(i+4-4))*w)+(j+4-3)];
+         conv += W[((((a*nif)+b)*5)+4)*5+4] * x[(((b*h)+(i+4-4))*w)+(j+4-4)];
+
+         y[(a*oh+i)*ow+j] += (conv >> QF); // because i'm using 32-bit int
+
+         #endif
+
+      }
+   }
+   synch_barrier();
+}
+
+void conv16_unrolled_5x5_four_coarsest(int16_t *__restrict__ W, int16_t *__restrict__ x, int16_t *__restrict__ y, int h, int w, int fh, int fw, int oh, int ow, int nif, int a, int b) {
+   int i;
+   register int myid      = rt_core_id();
+   register int num_cores = get_core_num();
+   // synch_barrier();
+   for (i=myid; i<oh; i+=num_cores) {
+      int j;
+      for (j=0; j<ow; j++) {
+
+         int32_t conv = 0;
+
+         #ifndef IMPRECISE_ASM5
+
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+0)*5+0], x[(((b*h)+(i+4-0))*w)+(j+4-0)]);
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+0)*5+1], x[(((b*h)+(i+4-0))*w)+(j+4-1)]);
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+0)*5+2], x[(((b*h)+(i+4-0))*w)+(j+4-2)]);
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+0)*5+3], x[(((b*h)+(i+4-0))*w)+(j+4-3)]);
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+0)*5+4], x[(((b*h)+(i+4-0))*w)+(j+4-4)]);
+
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+1)*5+0], x[(((b*h)+(i+4-1))*w)+(j+4-0)]);
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+1)*5+1], x[(((b*h)+(i+4-1))*w)+(j+4-1)]);
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+1)*5+2], x[(((b*h)+(i+4-1))*w)+(j+4-2)]);
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+1)*5+3], x[(((b*h)+(i+4-1))*w)+(j+4-3)]);
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+1)*5+4], x[(((b*h)+(i+4-1))*w)+(j+4-4)]);
+
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+2)*5+0], x[(((b*h)+(i+4-2))*w)+(j+4-0)]);
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+2)*5+1], x[(((b*h)+(i+4-2))*w)+(j+4-1)]);
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+2)*5+2], x[(((b*h)+(i+4-2))*w)+(j+4-2)]);
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+2)*5+3], x[(((b*h)+(i+4-2))*w)+(j+4-3)]);
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+2)*5+4], x[(((b*h)+(i+4-2))*w)+(j+4-4)]);
+
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+3)*5+0], x[(((b*h)+(i+4-3))*w)+(j+4-0)]);
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+3)*5+1], x[(((b*h)+(i+4-3))*w)+(j+4-1)]);
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+3)*5+2], x[(((b*h)+(i+4-3))*w)+(j+4-2)]);
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+3)*5+3], x[(((b*h)+(i+4-3))*w)+(j+4-3)]);
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+3)*5+4], x[(((b*h)+(i+4-3))*w)+(j+4-4)]);
+
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+4)*5+0], x[(((b*h)+(i+4-4))*w)+(j+4-0)]);
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+4)*5+1], x[(((b*h)+(i+4-4))*w)+(j+4-1)]);
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+4)*5+2], x[(((b*h)+(i+4-4))*w)+(j+4-2)]);
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+4)*5+3], x[(((b*h)+(i+4-4))*w)+(j+4-3)]);
+         conv += FIXED_MUL(W[((((a*nif)+b)*5)+4)*5+4], x[(((b*h)+(i+4-4))*w)+(j+4-4)]);
+
+         y[(a*oh+i)*ow+j] += conv; // because i'm using 32-bit int
+
+         #else
+
+         conv += W[((((a*nif)+b)*5)+0)*5+0] * x[(((b*h)+(i+4-0))*w)+(j+4-0)];
+         conv += W[((((a*nif)+b)*5)+0)*5+1] * x[(((b*h)+(i+4-0))*w)+(j+4-1)];
+         conv += W[((((a*nif)+b)*5)+0)*5+2] * x[(((b*h)+(i+4-0))*w)+(j+4-2)];
+         conv += W[((((a*nif)+b)*5)+0)*5+3] * x[(((b*h)+(i+4-0))*w)+(j+4-3)];
+         conv += W[((((a*nif)+b)*5)+0)*5+4] * x[(((b*h)+(i+4-0))*w)+(j+4-4)];
+
+         conv += W[((((a*nif)+b)*5)+1)*5+0] * x[(((b*h)+(i+4-1))*w)+(j+4-0)];
+         conv += W[((((a*nif)+b)*5)+1)*5+1] * x[(((b*h)+(i+4-1))*w)+(j+4-1)];
+         conv += W[((((a*nif)+b)*5)+1)*5+2] * x[(((b*h)+(i+4-1))*w)+(j+4-2)];
+         conv += W[((((a*nif)+b)*5)+1)*5+3] * x[(((b*h)+(i+4-1))*w)+(j+4-3)];
+         conv += W[((((a*nif)+b)*5)+1)*5+4] * x[(((b*h)+(i+4-1))*w)+(j+4-4)];
+
+         conv += W[((((a*nif)+b)*5)+2)*5+0] * x[(((b*h)+(i+4-2))*w)+(j+4-0)];
+         conv += W[((((a*nif)+b)*5)+2)*5+1] * x[(((b*h)+(i+4-2))*w)+(j+4-1)];
+         conv += W[((((a*nif)+b)*5)+2)*5+2] * x[(((b*h)+(i+4-2))*w)+(j+4-2)];
+         conv += W[((((a*nif)+b)*5)+2)*5+3] * x[(((b*h)+(i+4-2))*w)+(j+4-3)];
+         conv += W[((((a*nif)+b)*5)+2)*5+4] * x[(((b*h)+(i+4-2))*w)+(j+4-4)];
+
+         conv += W[((((a*nif)+b)*5)+3)*5+0] * x[(((b*h)+(i+4-3))*w)+(j+4-0)];
+         conv += W[((((a*nif)+b)*5)+3)*5+1] * x[(((b*h)+(i+4-3))*w)+(j+4-1)];
+         conv += W[((((a*nif)+b)*5)+3)*5+2] * x[(((b*h)+(i+4-3))*w)+(j+4-2)];
+         conv += W[((((a*nif)+b)*5)+3)*5+3] * x[(((b*h)+(i+4-3))*w)+(j+4-3)];
+         conv += W[((((a*nif)+b)*5)+3)*5+4] * x[(((b*h)+(i+4-3))*w)+(j+4-4)];
+
+         conv += W[((((a*nif)+b)*5)+4)*5+0] * x[(((b*h)+(i+4-4))*w)+(j+4-0)];
+         conv += W[((((a*nif)+b)*5)+4)*5+1] * x[(((b*h)+(i+4-4))*w)+(j+4-1)];
+         conv += W[((((a*nif)+b)*5)+4)*5+2] * x[(((b*h)+(i+4-4))*w)+(j+4-2)];
+         conv += W[((((a*nif)+b)*5)+4)*5+3] * x[(((b*h)+(i+4-4))*w)+(j+4-3)];
+         conv += W[((((a*nif)+b)*5)+4)*5+4] * x[(((b*h)+(i+4-4))*w)+(j+4-4)];
+
+         y[(a*oh+i)*ow+j] += (conv >> QF); // because i'm using 32-bit int
+
+         #endif
+
+      }
+   }
+   synch_barrier();
+}
+
+void conv16_unrolled_ptr_5x5_four_coarse(int16_t *__restrict__ W, int16_t *__restrict__ x, int16_t *__restrict__ y, int h, int w, int fh, int fw, int oh, int ow, int nif, int a, int b) {
+   register int i;
+   register int j;
+   register int myid      = rt_core_id();
+   register int num_cores = get_core_num();
+   int16_t *W_ptr;
+   int16_t *x_ptr;
+   int16_t *y_ptr = y + a*oh*ow;
+   int16_t *x_base = x + b*h*w + (fh-1)*w + (fw-1);
+   int16_t *W_base = &W[((a*nif)+b)*fh*fw];
+   
+   // synch_barrier();
+
+   for (i=0; i<oh; i++) {
+      for (j=myid; j<ow; j+=num_cores) {
+
+         int conv = 0;
+
+         W_ptr = W_base;
+         x_ptr = x_base + i*w + j;
+
+         #ifndef IMPRECISE_ASM5
+
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         x_ptr -= w-5;
+
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         x_ptr -= w-5;
+
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         x_ptr -= w-5;
+
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         x_ptr -= w-5;
+
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         x_ptr -= w-5;
+
+         *(y_ptr + i*ow+j) += conv;
+
+         #else
+
+         conv += *W_ptr++ * *x_ptr--;
+         conv += *W_ptr++ * *x_ptr--;
+         conv += *W_ptr++ * *x_ptr--;
+         conv += *W_ptr++ * *x_ptr--;
+         conv += *W_ptr++ * *x_ptr--;
+         x_ptr -= w-5;
+
+         conv += *W_ptr++ * *x_ptr--;
+         conv += *W_ptr++ * *x_ptr--;
+         conv += *W_ptr++ * *x_ptr--;
+         conv += *W_ptr++ * *x_ptr--;
+         conv += *W_ptr++ * *x_ptr--;
+         x_ptr -= w-5;
+
+         conv += *W_ptr++ * *x_ptr--;
+         conv += *W_ptr++ * *x_ptr--;
+         conv += *W_ptr++ * *x_ptr--;
+         conv += *W_ptr++ * *x_ptr--;
+         conv += *W_ptr++ * *x_ptr--;
+         x_ptr -= w-5;
+
+         conv += *W_ptr++ * *x_ptr--;
+         conv += *W_ptr++ * *x_ptr--;
+         conv += *W_ptr++ * *x_ptr--;
+         conv += *W_ptr++ * *x_ptr--;
+         conv += *W_ptr++ * *x_ptr--;
+         x_ptr -= w-5;
+
+         conv += *W_ptr++ * *x_ptr--;
+         conv += *W_ptr++ * *x_ptr--;
+         conv += *W_ptr++ * *x_ptr--;
+         conv += *W_ptr++ * *x_ptr--;
+         conv += *W_ptr++ * *x_ptr--;
+         x_ptr -= w-5;
+
+         *(y_ptr + i*ow+j) += conv >> QF;
+
+         #endif
+
+      }
+   }
+
+   synch_barrier();
+}
+
+void conv16_unrolled_ptr_5x5_four_coarsest(int16_t *__restrict__ W, int16_t *__restrict__ x, int16_t *__restrict__ y, int h, int w, int fh, int fw, int oh, int ow, int nif, int a, int b) {
+   // register int i;
+   // register int j;
+   // register int myid = rt_core_id();
+   // int16_t *W_ptr;
+   // int16_t *x_ptr;
+   int16_t *y_ptr = y + a*oh*ow;
+   int16_t *x_base = x + b*h*w + (fh-1)*w + (fw-1);
+   int16_t *W_base = &W[((a*nif)+b)*fh*fw];
+
+   register int myid      = rt_core_id();
+   register int num_cores = get_core_num();
+
+   register int i;
+   register int j;
+   int16_t *W_ptr;
+   int16_t *x_ptr;
+
+   for (i=myid; i<oh; i+=num_cores) {
+      for (j=0; j<ow; j++) {
+
+         int conv = 0;
+
+         W_ptr = W_base;
+         x_ptr = x_base + i*w + j;
+
+         #ifndef IMPRECISE_ASM5
+
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         x_ptr -= w-5;
+
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         x_ptr -= w-5;
+
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         x_ptr -= w-5;
+
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         x_ptr -= w-5;
+
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         conv += FIXED_MUL(*W_ptr++, *x_ptr--);
+         x_ptr -= w-5;
+
+         *(y_ptr + i*ow+j) += conv;
+
+         #else
+
+         conv += *W_ptr++ * *x_ptr--;
+         conv += *W_ptr++ * *x_ptr--;
+         conv += *W_ptr++ * *x_ptr--;
+         conv += *W_ptr++ * *x_ptr--;
+         conv += *W_ptr++ * *x_ptr--;
+         x_ptr -= w-5;
+
+         conv += *W_ptr++ * *x_ptr--;
+         conv += *W_ptr++ * *x_ptr--;
+         conv += *W_ptr++ * *x_ptr--;
+         conv += *W_ptr++ * *x_ptr--;
+         conv += *W_ptr++ * *x_ptr--;
+         x_ptr -= w-5;
+
+         conv += *W_ptr++ * *x_ptr--;
+         conv += *W_ptr++ * *x_ptr--;
+         conv += *W_ptr++ * *x_ptr--;
+         conv += *W_ptr++ * *x_ptr--;
+         conv += *W_ptr++ * *x_ptr--;
+         x_ptr -= w-5;
+
+         conv += *W_ptr++ * *x_ptr--;
+         conv += *W_ptr++ * *x_ptr--;
+         conv += *W_ptr++ * *x_ptr--;
+         conv += *W_ptr++ * *x_ptr--;
+         conv += *W_ptr++ * *x_ptr--;
+         x_ptr -= w-5;
+
+         conv += *W_ptr++ * *x_ptr--;
+         conv += *W_ptr++ * *x_ptr--;
+         conv += *W_ptr++ * *x_ptr--;
+         conv += *W_ptr++ * *x_ptr--;
+         conv += *W_ptr++ * *x_ptr--;
+         x_ptr -= w-5;
+
+         *(y_ptr + i*ow+j) += conv >> QF;
+
+         #endif
+
+      }
+   }
+
+   synch_barrier();
+
+}
+
+#ifndef __GCC__
+#ifndef __riscv__
+void conv16_asm_mul_unrolled_5x5_four_coarsest(int16_t *__restrict__ W, int16_t *__restrict__ x, int16_t *__restrict__ y, int h, int w, int fh, int fw, int oh, int ow, int nif, int a, int b) {
+   register int i;
+   register int j;
+   register int myid      = rt_core_id();
+   register int num_cores = get_core_num();
+   register int16_t *W_ptr;
+   register int16_t *x_ptr;
+   int16_t *y_ptr = y + a*oh*ow;
+   int16_t *x_base = x + b*h*w + (fh-1)*w + (fw-1);
+   int16_t *W_base = &W[((a*nif)+b)*fh*fw];
+   for (i=myid; i<oh; i+=num_cores) {
+      int j;
+      for (j=0; j<ow; j++) {
+         W_ptr = W_base;
+         x_ptr = x_base + i*w + j;
+         int32_t conv = 0;
+         int ui;
+         for (ui=0; ui<fh; ui++) {
+
+            #ifndef IMPRECISE_ASM5
+            FIXED_MUL_ASM5_PRECISE(W_ptr,x_ptr,conv);
+            #else
+            register int32_t conv_int = conv;
+            FIXED_MUL_ASM5_IMPRECISE(W_ptr,x_ptr,conv_int);
+            conv = conv_int;
+            #endif
+
+            x_ptr -= w-fw+5;
+            W_ptr += 5;
+
+         }
+         y[(a*oh+i)*ow+j] = (y[(a*oh+i)*ow+j] + conv) & 0x0000ffff; // because i'm using 32-bit int
+      }
+   }
+   synch_barrier();
+}
+#endif
+#endif
+
+void load() {
+   int i;
+   for(i=0; i<FH*FW; i++) {
+      g_W[i] = (int16_t) 0x2000;
+   }
+   for(i=0; i<IH*IW; i++) {
+      g_x[i] = (int16_t) (i % 0x2000);
+   }
+   for(i=0; i<OH*OW; i++) {
+      g_y[i] = (int16_t) (i % 0x2000);
+      g_y_in[i] = g_y[i];
+      // g_y_in[i] = (int16_t) (i % 0x2000);
+      // g_y[i] = 0;
+   }
+}
+
+void printW() {
+   int i;
+   for(i=0; i<FH*FW; i++)
+      printf("%04x\n", g_W[i]);
+}
+
+void printX() {
+   int i;
+   for(i=0; i<IH*IW; i++)
+      printf("%04x\n", g_x[i]);
+}
+
+int check(int16_t *y) {
+   int i;
+   int errors = 0;
+   for(i=0; i<OH*OW; i++) {
+      int err = ((y[i]!=correct_yout[i]) ? 1 : 0);
+      errors += err;
+      #ifndef PULP_SPI
+      if(err)
+         printf("error: y_out=%04x instead of %04x! y_in=%04x\n", y[i], correct_yout[i], g_y_in[i], 0);
+      #endif
+   }
+   return errors;
+}
+
+int checksum(int16_t *y) {
+   int sum = 0;
+   int i;
+   for(i=0; i<OH*OW; i++)
+      sum += y[i];
+   return sum;
+}
diff --git a/reliability_tests/icache_fi_conv16/pulp_inject_fault.tcl b/reliability_tests/icache_fi_conv16/pulp_inject_fault.tcl
new file mode 100644
index 0000000..5968d60
--- /dev/null
+++ b/reliability_tests/icache_fi_conv16/pulp_inject_fault.tcl
@@ -0,0 +1,53 @@
+# Copyright 2023 ETH Zurich and University of Bologna.
+# Solderpad Hardware License, Version 0.51, see LICENSE for details.
+# SPDX-License-Identifier: SHL-0.51
+#
+# Author: Michael Rogenmoser (michaero@iis.ee.ethz.ch)
+
+transcript quietly
+if {! [info exists ::env(VSIM_PATH)]} {error "Define VSIM_PATH"}
+set utils_base_path  [file join $::env(VSIM_PATH) scripts fault_injection_utils]
+set script_base_path [file join $::env(VSIM_PATH) fault_injection_sim scripts]
+
+set verbosity            2
+set log_injections       1
+# Easy way to generate a variable seed
+# set seed                 [clock seconds]
+# Default value
+set seed                 12345
+set print_statistics     1
+
+set inject_start_time 223512000000ps
+set inject_stop_time 4400000000000ps
+set injection_clock "pulp_cluster_tb/cluster_i/clk_i"
+set injection_clock_trigger 0
+set fault_period 1000
+set rand_initial_injection_phase 0
+# max_num set to 0 means until stop_time
+set max_num_fault_inject 0
+set signal_fault_duration 20ns
+set register_fault_duration 0ns
+
+set allow_multi_bit_upset 0
+set use_bitwidth_as_weight 0
+set check_core_output_modification 0
+set check_core_next_state_modification 0
+set reg_to_sig_ratio 1
+
+source [file join $utils_base_path pulp_extract_nets.tcl]
+
+set inject_signals_netlist []
+set inject_register_netlist []
+set output_netlist []
+set next_state_netlist []
+set assertion_disable_list []
+
+# for {set idx 0} {$idx < 12} {incr idx} {
+#     set inject_signals_netlist [list {*}$inject_signals_netlist {*}[get_all_core_nets $idx]]
+#     set output_netlist [list {*}$output_netlist {*}[get_core_output_nets $idx]]
+# }
+
+set inject_register_netlist [list {*}$inject_register_netlist {*}[get_all_icache_state]]
+
+source [file join $script_base_path inject_fault.tcl]
+
diff --git a/rt_nn_tests.yaml b/rt_nn_tests.yaml
new file mode 100644
index 0000000..96848d6
--- /dev/null
+++ b/rt_nn_tests.yaml
@@ -0,0 +1,16 @@
+parallel_bare_tests:
+  xpnn_conv:
+    path: ./rt_nn_tests/xpnn_conv
+    command: make clean all run
+  xpnnv2_conv:
+    path: ./rt_nn_tests/xpnnv2_conv
+    command: make clean all run
+  xpnn_maxpool_2b:
+    path: ./rt_nn_tests/xpnn_maxpool_2b
+    command: make clean all run
+  xptnn_linear:
+    path: ./rt_nn_tests/xptnn_linear
+    command: make clean all run
+  xptnn_maxpool:
+    path: ./rt_nn_tests/xptnn_maxpool
+    command: make clean all run
diff --git a/rt_nn_tests/xpnn_conv/Makefile b/rt_nn_tests/xpnn_conv/Makefile
new file mode 100644
index 0000000..a180c60
--- /dev/null
+++ b/rt_nn_tests/xpnn_conv/Makefile
@@ -0,0 +1,19 @@
+APP = test
+PULP_APP = test
+PULP_APP_SRCS = test.c
+PULP_APP_SRCS += xpulp_nn_conv_u4_u8_i2.c
+PULP_APP_SRCS += xpulp_nn_matmul_u4_u8_i2.c
+
+CORE=8
+
+
+PULP_CFLAGS += -DNUM_CORES=$(CORE) -I. -O3
+PULP_LDFLAGS += -lc -lm -lgcc -Wl,-print-memory-usage 
+
+PULP_APP_CFLAGS += -DNUM_CORES=$(CORE) -I. -O3
+PULP_APP_LDFLAGS += -lc -lm -lgcc -Wl,-print-memory-usage 
+
+PULP_CFLAGS += -DPULP_RUNTIME -DSINGLE_CORE_DMA -DALWAYS_BLOCK_DMA_TRANSFERS
+
+
+include $(PULP_SDK_HOME)/install/rules/pulp.mk
diff --git a/rt_nn_tests/xpnn_conv/data_statstest.h b/rt_nn_tests/xpnn_conv/data_statstest.h
new file mode 100644
index 0000000..d679d0f
--- /dev/null
+++ b/rt_nn_tests/xpnn_conv/data_statstest.h
@@ -0,0 +1,1894 @@
+// this file is generated automatically by the python script `generate_statstest.py`
+
+#ifndef _DATA_H
+#define _DATA_H
+
+
+
+#define IM2COL_DIM (72 * NUM_CORES * 2) // for NUM_CORES cores
+#define OUTPUT_DIM 512
+#define OUTPUT_DIM_FP (1)
+uint8_t pWeight_0 [] = {
+0x29,
+0xc5,
+0xac,
+0x95,
+0x91,
+0x84,
+0x79,
+0x41,
+0xf7,
+0xd6,
+0x8a,
+0x4e,
+0xa8,
+0x85,
+0x7f,
+0x76,
+0x57,
+0x20,
+0x83,
+0xa5,
+0xfa,
+0x8a,
+0xd4,
+0xf2,
+0x58,
+0xa0,
+0x7d,
+0x60,
+0x0a,
+0x58,
+0x02,
+0x04,
+0x86,
+0xfd,
+0xcc,
+0xfb,
+0x8c,
+0xc0,
+0xc0,
+0x88,
+0xde,
+0x5d,
+0x3b,
+0x3c,
+0x69,
+0x80,
+0xd0,
+0x38,
+0x83,
+0xe5,
+0x35,
+0x27,
+0x14,
+0x5d,
+0x2a,
+0xe8,
+0xd0,
+0xf9,
+0xcc,
+0x81,
+0x16,
+0x31,
+0xd3,
+0xc6,
+0xf1,
+0xfa,
+0xe2,
+0x5b,
+0x08,
+0x6d,
+0xa5,
+0x01,
+0xb2,
+0x28,
+0x54,
+0x06,
+0x78,
+0x97,
+0x40,
+0x73,
+0xd2,
+0x27,
+0x96,
+0xbd,
+0x1f,
+0xc0,
+0xfa,
+0xb8,
+0x49,
+0x68,
+0x75,
+0x2f,
+0xe5,
+0xa4,
+0xcb,
+0xcc,
+0x09,
+0x39,
+0x85,
+0x71,
+0xc3,
+0xfb,
+0x6f,
+0x9b,
+0x42,
+0x0f,
+0x91,
+0x1b,
+0x66,
+0xaf,
+0xfb,
+0xc8,
+0xf8,
+0xba,
+0x9d,
+0x0a,
+0xd2,
+0xa8,
+0x39,
+0x4a,
+0x2d,
+0xcd,
+0xf8,
+0xc4,
+0x8c,
+0x02,
+0xe2,
+0x4f,
+0x68,
+0xff,
+0x96,
+0x7f,
+0x96,
+0xad,
+0xa5,
+0xf2,
+0xb8,
+0xb8,
+0x5d,
+0x02,
+0x4c,
+0x19,
+0xb0,
+0x11,
+0x51,
+0x8d,
+0xb3,
+0xb1,
+0x3f,
+0xcd,
+0xaf,
+0x85,
+0x94,
+0x65,
+0x24,
+0xa1,
+0xbb,
+0x3e,
+0xa1,
+0xe2,
+0xab,
+0xb4,
+0x62,
+0x6c,
+0xf5,
+0x2f,
+0x5a,
+0xbe,
+0x68,
+0xba,
+0x8a,
+0xcf,
+0xa1,
+0x3f,
+0xfa,
+0xb8,
+0xe9,
+0xe8,
+0xfa,
+0x4e,
+0xdd,
+0x57,
+0x45,
+0x1a,
+0x5c,
+0xfd,
+0x58,
+0x77,
+0x52,
+0xd6,
+0x15,
+0x07,
+0x39,
+0xf2,
+0x34,
+0xf2,
+0x66,
+0x90,
+0xea,
+0x42,
+0x42,
+0x43,
+0x36,
+0xd7,
+0xc3,
+0xb8,
+0x2c,
+0x10,
+0xd0,
+0x90,
+0x2d,
+0xf9,
+0x6a,
+0xbc,
+0x57,
+0x19,
+0x69,
+0x60,
+0x65,
+0x7f,
+0xee,
+0x10,
+0x06,
+0x71,
+0xe4,
+0x49,
+0x5b,
+0x5a,
+0x5a,
+0x9c,
+0x6f,
+0x87,
+0x68,
+0x44,
+0xb4,
+0xf8,
+0xe8,
+0xae,
+0x75,
+0xe0,
+0x0f,
+0x43,
+0x51,
+0x71,
+0xde,
+0xce,
+0x56,
+0x13,
+0x3d,
+0xf9,
+0x80,
+0x1f,
+0x0a,
+0x9e,
+0x2b,
+0xa7,
+0x73,
+0x76,
+0xe2,
+0x36,
+0x16,
+0x42,
+0x0c,
+0x73,
+0x59,
+0xf6,
+0x53,
+0xaf,
+0x77,
+0x12,
+0x67,
+0x33,
+0xb7,
+0x90,
+0x41,
+0x55,
+0x90,
+0xbf,
+0x8c,
+0x31,
+0xfc,
+0xdd,
+0x9c,
+0x42,
+0x9a,
+0xfd,
+0x30,
+0xe5,
+0x12,
+0x12,
+0x6c,
+0xc3,
+0xa6,
+0xd8,
+0xb0,
+0x41,
+0x4c,
+0xd0,
+0x27,
+0x97,
+0x7f,
+0x8b,
+0x4c,
+0x44,
+0xb0,
+0x7f,
+0xba,
+0x18,
+0x37,
+0x46,
+0x9e,
+0xdc,
+0xbf,
+0xd8,
+0x97,
+0xce,
+0x3b,
+0x61,
+0x7e,
+0x9b,
+0x36,
+0xe7,
+0x0c,
+0x4e,
+0x57,
+0x21,
+0x36,
+0x18,
+0x4d,
+0x17,
+0x41,
+0x83,
+0xd4,
+0x8b,
+0xf7,
+0x53,
+0x4d,
+0xff,
+0x98,
+0xbb,
+0x79,
+0x38,
+0x2e,
+0x22,
+0x8f,
+0xe5,
+0x5e,
+0x81,
+0xdb,
+0xf0,
+0x2b,
+0x16,
+0xf6,
+0x69,
+0xb1,
+0xe5,
+0xd6,
+0xab,
+0xc6,
+0xb1,
+0xd0,
+0x62,
+0x57,
+0xa4,
+0x6c,
+0xe8,
+0x0e,
+0x65,
+0xb6,
+0x18,
+0x7a,
+0xe7,
+0x99,
+0x92,
+0x68,
+0x83,
+0x7e,
+0xe3,
+0xca,
+0x61,
+0xb0,
+0x5e,
+0x4d,
+0x25,
+0x92,
+0xe5,
+0x02,
+0xbb,
+0x81,
+0xfa,
+0x81,
+0x92,
+0x9d,
+0x80,
+0xd7,
+0xf4,
+0x73,
+0xe3,
+0x63,
+0x64,
+0x5a,
+0x26,
+0x98,
+0x55,
+0x32,
+0x65,
+0xf9,
+0x5c,
+0x5d,
+0xdc,
+0x2e,
+0x59,
+0xf1,
+0x90,
+0xf5,
+0x78,
+0x05,
+0xb1,
+0x5c,
+0xde,
+0xdd,
+0x93,
+0x3f,
+0xd8,
+0xc2,
+0x2a,
+0x94,
+0xfd,
+0x51,
+0x72,
+0x9c,
+0xda,
+0x58,
+0xc6,
+0x94,
+0xe5,
+0xed,
+0x3d,
+0xf3,
+0x7a,
+0xd8,
+0x91,
+0xbf,
+0xeb,
+0x31,
+0xda,
+0xe4,
+0x89,
+0x74,
+0x31,
+0x86,
+0x42,
+0xe2,
+0x1b,
+0x7e,
+0x28,
+0xaf,
+0xf8,
+0xfd,
+0x0b,
+0x9a,
+0xa7,
+0x39,
+0xd8,
+0x9d,
+0x25,
+0x7f,
+0xe5,
+0xc4,
+0x7f,
+0x98,
+0x9f,
+0x79,
+0x5f,
+0x4b,
+0x7a,
+0x05,
+0x07,
+0x6c,
+0xe9,
+0x1f,
+0xd0,
+0x60,
+0xfb,
+0x7f,
+0xf6,
+0x5c,
+0xa2,
+0x3d,
+0x2c,
+0x22,
+0x7b,
+0xbe,
+0x39,
+0x0a,
+0xfc,
+0x8f,
+0x8b,
+0x80,
+0x6f,
+0x6f,
+0x1b,
+0x28,
+0x9b,
+0x79,
+0x0e,
+0xb3,
+0x57,
+0x07,
+0x45,
+0x61,
+0x83,
+0x12,
+0x7a,
+0x53,
+0x3d,
+0x80,
+0xf1,
+0xe1,
+0xd1,
+0x40,
+0x26,
+0xfd,
+0xd8,
+0x81,
+0xab,
+0xbe,
+0x16,
+0xfa,
+0xf2,
+0x25,
+0x15,
+0x8f,
+0xcd,
+0xc7,
+0xbe,
+0xe1,
+0xd9,
+0x95,
+0x3b,
+0xbd,
+0x6b,
+0x81,
+0x59,
+0x26,
+0x27,
+0xe8,
+0x1b,
+0x74,
+0x7b,
+0x95,
+0xb4,
+0xe0,
+0x60,
+0x0b,
+0x44,
+0x67,
+0x7e,
+0xb9,
+0xfb,
+0xc2,
+0xc3,
+0xb2,
+0xd5,
+0x22,
+0x39,
+0xae,
+0xb2,
+0xd9,
+0x2e,
+0x63,
+0x19,
+0x04,
+0xe5,
+0x69,
+0xf5,
+0x5f,
+0x2a,
+0x9f,
+0x8f,
+0xe2,
+0xab,
+0x89,
+0x8d,
+0xd9,
+0x23,
+0x45,
+0x75,
+0xa2,
+0x2e,
+0x1d,
+0xc2,
+0xe4,
+0xf0,
+0xe0,
+0xbb,
+0x26,
+0x8b,
+0x7c,
+0xd4,
+0xaa,
+0x6d,
+0x74,
+0x0e,
+0xa7,
+0x73,
+0x44,
+0xc2,
+0xee,
+0x74,
+0xd4,
+0x07,
+0xf9,
+0xdd,
+0xd6,
+0x57,
+0x84,
+0xdd,
+0xee,
+0x67,
+0x82,
+0xd1,
+0x52,
+0xb3,
+0x35,
+0xb3,
+0xb4,
+0x3e,
+0x19,
+0xd7,
+0x55,
+0x22,
+0xf0,
+0xdd,
+0xe1,
+0x80,
+0x15,
+0x71,
+0xff,
+0x58,
+0x87,
+0xe2,
+0xf2,
+0xd4,
+0x05,
+0x68,
+0x46,
+0xf2,
+0x22,
+0xd0,
+0x46,
+0xa8,
+0xa4,
+0x16,
+0x81,
+0xc5,
+0x7d,
+0x1b,
+0x50,
+0x25,
+0x91,
+0xcd,
+0x29,
+0xa6,
+0x35,
+0x32,
+0xf5,
+0xff,
+0x53,
+0xd7,
+0x00,
+0xf6,
+0xab,
+0x39,
+0xac,
+0x85,
+0xe9,
+0x1a,
+0x80,
+0x1f,
+0x89,
+0x30,
+0x77,
+0x9a,
+0x5e,
+0xb2,
+0xdb,
+0x74,
+0x53,
+0x48,
+0x6a,
+0x91,
+0x11,
+0xf0,
+0xd8,
+0x94,
+0x63,
+0xbc,
+0x32,
+0x84,
+0x55,
+0x04,
+0x84,
+0x8a,
+0x29,
+0x38,
+0x9c,
+0x2c,
+0x14,
+0x23,
+0xfe,
+0xb9,
+0x88,
+0x6e,
+0xac,
+0xf4,
+0xcf,
+0xd6,
+0x4a,
+0xe5,
+0x19,
+0xad,
+0x80,
+0x4b,
+0x34,
+0x0f,
+0xbc,
+0x93,
+0x86,
+0x1c,
+0x77,
+0x1c,
+0x7e,
+0xda,
+0x39,
+0xf3,
+0xb0,
+0x7b,
+0xae,
+0x0a,
+0x4a,
+0x49,
+0xba,
+0x69,
+0xf5,
+0xa5,
+0x05,
+0x3b,
+0xab,
+0xfd,
+0xa1,
+0xc6,
+0xc7,
+0x8d,
+0x7e,
+0x1a,
+0x1d,
+0xa4,
+0x6d,
+0x54,
+0xe2,
+0x1b,
+0xb9,
+0x46,
+0x42,
+0xbe,
+0xc9,
+0x6a,
+0xc7,
+0x14,
+0xa1,
+0xd4,
+0x27,
+0x39,
+0x65,
+0xf8,
+0xb3,
+0xf4,
+0x1a,
+0x31,
+0x9a,
+0xc5,
+0x74,
+0x63,
+0xb3,
+0xa9,
+0x10,
+0xe2,
+0x22,
+0x1d,
+0xc6,
+0x20,
+0x00,
+0x58,
+0xa7,
+0xaa,
+0x82,
+0x25,
+0x71,
+0xf6,
+0xb0,
+0x5d,
+0xd8,
+0xeb,
+0xf5,
+0x17,
+0xd0,
+0x04,
+0x07,
+0xac,
+0xf5,
+0x63,
+0x7a,
+0x26,
+0x0a,
+0x68,
+0xe3,
+0x74,
+0xc5,
+0x44,
+0x44,
+0xd8,
+0xf3,
+0xa5,
+0x99,
+0x77,
+0xc4,
+0x48,
+0x47,
+0xc0,
+0x94,
+0xd6,
+0x27,
+0x79,
+0x3d,
+0x0a,
+0x83,
+0x46,
+0x0b,
+0x8f,
+0xd7,
+0x03,
+0x7c,
+0xa7,
+0x35,
+0xd5,
+0x70,
+0xb2,
+0x19,
+0x29,
+0x14,
+0xb6,
+0xbe,
+0x61,
+0xfb,
+0x5c,
+0x04,
+0xf4,
+0x7f,
+0xb4,
+0xbd,
+0xcb,
+0x4c,
+0xae,
+0xd0,
+0x20,
+0x3e,
+0xd9,
+0x25,
+0xcc,
+0xa9,
+0xd2,
+0x13,
+0xf7,
+0x66,
+0x4e,
+0xbe,
+0xeb,
+0xed,
+0x6b,
+0xb0,
+0x9c,
+0xd3,
+0x94,
+0x21,
+0x33,
+0x70,
+0x16,
+0x53,
+0xfe,
+0x21,
+0x08,
+0xf5,
+0xbc,
+0x1b,
+0xb1,
+0x92,
+0x3a,
+0xf4,
+0xd2,
+0xc7,
+0x40,
+0x96,
+0x01,
+0xbc,
+0xd5,
+0x96,
+0x35,
+0x17,
+0xa0,
+0x0c,
+0x1b,
+0x85,
+0x2e,
+0xb3,
+0xab,
+0x55,
+0xcb,
+0xc0,
+0xe0,
+0x75,
+0x61,
+0xf8,
+0xa6,
+0x96,
+0xa2,
+0x3f,
+0x57,
+0xd6,
+0xb7,
+0xfc,
+0x8a,
+0x6a,
+0x6c,
+0xcc,
+0x38,
+0x7f,
+0x8c,
+0x99,
+0x6e,
+0x5e,
+0x19,
+0x23,
+0x21,
+0xb0,
+0xaa,
+0x77,
+0x43,
+0x05,
+0x9b,
+0x70,
+0xd4,
+0xde,
+0xf8,
+0x7d,
+0xbb,
+0xad,
+0x28,
+0x4e,
+0xc3,
+0x5f,
+0x97,
+0x1d,
+0x79,
+0xd9,
+0x55,
+0xce,
+0x28,
+0x13,
+0x91,
+0xc0,
+0x94,
+0x5b,
+0x04,
+0x70,
+0xcc,
+0xd8,
+0x38,
+0xa9,
+0x72,
+0x0e,
+0xd1,
+0x93,
+0xe4,
+0x81,
+0x47,
+0x29,
+0xec,
+0x0f,
+0x4c,
+0x5d,
+0x50,
+0x6e,
+0xd7,
+0xa2,
+0xe3,
+0x4e,
+0x58,
+0x29,
+0x05,
+0xa2,
+0x37,
+0x27,
+0xfe,
+0x84,
+0x51,
+0xac,
+0xc2,
+0xc8,
+0x03,
+0xe6,
+0x5f,
+0xea,
+0x15,
+0x25,
+0x7a,
+0x39,
+0x60,
+0xd1,
+0xc8,
+0x94,
+0xb3,
+0xf9,
+0x8b,
+0x20,
+0x45,
+0xec,
+0x75,
+0x54,
+0xd1,
+0xdd,
+0x6a,
+0x39,
+0x31,
+0xc9,
+0x62,
+0x67,
+0x84,
+0xb3,
+0x47,
+0xbb,
+0x10,
+0xb1,
+0x0f,
+0x5e,
+0x11,
+0x00,
+0x1e,
+0xc5,
+0x79,
+0xf9,
+0xd9,
+0xb7,
+0x18,
+0x0a,
+0x55,
+0x1d,
+0x99,
+0xd5,
+0x46,
+0x2a,
+0x22,
+0xd9,
+0x2e,
+0xc3,
+0x2d,
+0xd1,
+0x9e,
+0xb9,
+0x3a,
+0x32,
+0x1a,
+0x13,
+0xdb,
+0xe5,
+0x45,
+0x1a,
+0x83,
+0x38,
+0xec,
+0x45,
+0x7b,
+0x87,
+0x01,
+0x3c,
+0x32,
+0xd3,
+0x60,
+0x09,
+0xdd,
+0xce,
+0x87,
+0xa3,
+0x03,
+0x02,
+0x91,
+0xfa,
+0xc5,
+0x4d,
+0x4d,
+0x08,
+0x6d,
+0xca,
+0x9e,
+0xfa,
+0x7e,
+0x8c,
+0xdf,
+0x3e,
+0xb3,
+0x10,
+0xfb,
+0xa3,
+0x2f,
+0xef,
+0x7e,
+0x1e,
+0x9f,
+0xd3,
+0x78,
+0xfb,
+0x5f,
+0x50,
+0x4c,
+0xaa,
+0x16,
+0x0c,
+0x35,
+0x46,
+0x10,
+0x3c,
+0x67,
+0xaa,
+0x95,
+0x8a,
+0x38,
+0xf3,
+0xb5,
+0x07,
+0xdb,
+0x6c,
+0x52,
+0xde,
+0xc6,
+0x77,
+0x0e,
+0xd0,
+0xf7,
+0xea,
+0x1f,
+0x96,
+0xae
+};
+
+uint8_t pIn_0 [] = {
+0x09,
+0xe8,
+0x1c,
+0x6f,
+0x55,
+0xe1,
+0xa3,
+0x4d,
+0xad,
+0x94,
+0x4a,
+0x9d,
+0x90,
+0xfc,
+0x76,
+0xc9,
+0x84,
+0xce,
+0xa8,
+0x18,
+0xe2,
+0x30,
+0x85,
+0x41,
+0x6f,
+0x11,
+0x09,
+0xff,
+0x07,
+0xb0,
+0x5b,
+0xe4,
+0x40,
+0xf3,
+0x8f,
+0xcd,
+0xa3,
+0x2f,
+0x85,
+0x67,
+0x31,
+0x96,
+0xc2,
+0x86,
+0x1d,
+0xe5,
+0xc1,
+0xee,
+0x0b,
+0x4d,
+0x50,
+0xa6,
+0x00,
+0x69,
+0xff,
+0x8c,
+0x4c,
+0xb5,
+0x46,
+0xd5,
+0x14,
+0xd0,
+0xdb,
+0x47,
+0xf9,
+0x2d,
+0x77,
+0x9a,
+0x92,
+0xa4,
+0xe8,
+0xab,
+0xbd,
+0xd3,
+0x88,
+0xb6,
+0x00,
+0xbc,
+0xb0,
+0xe7,
+0x60,
+0xd9,
+0x8b,
+0x1b,
+0x9e,
+0x06,
+0x95,
+0x99,
+0x8d,
+0x6e,
+0x21,
+0x20,
+0xa3,
+0x5e,
+0x99,
+0x27,
+0xb1,
+0x7d,
+0xce,
+0xdc,
+0xcf,
+0xd6,
+0xe8,
+0x16,
+0xaa,
+0xec,
+0x67,
+0x8c,
+0x23,
+0xd0,
+0x15,
+0xa0,
+0xe8,
+0x86,
+0x2b,
+0x90,
+0xe1,
+0xb2,
+0x5b,
+0x13,
+0x19,
+0xbe,
+0x8c,
+0x9e,
+0x12,
+0x13,
+0xeb,
+0x69
+};
+
+int32_t pLambda_0 [] = {
+0x00247d20,
+0x001b6100,
+0x00124840,
+0x00247d20,
+0x00247d20,
+0x00124840,
+0x001b6100,
+0x00247d20,
+0x00092c20,
+0x00247d20,
+0x001b6100,
+0x00124840,
+0x00124840,
+0x00124840,
+0x00124840,
+0x001b6100,
+0x00124840,
+0x00124840,
+0x00092c20,
+0x00124840,
+0x001b6100,
+0x00092c20,
+0x00124840,
+0x001b6100,
+0x001b6100,
+0x00247d20,
+0x00247d20,
+0x00092c20,
+0x001b6100,
+0x00092c20,
+0x00247d20,
+0x001b6100
+};
+
+int32_t pKappa_0 [] = {
+0x00000acb,
+0x00000818,
+0x00000566,
+0x00000acb,
+0x00000acb,
+0x00000566,
+0x00000818,
+0x00000acb,
+0x000002b3,
+0x00000acb,
+0x00000818,
+0x00000566,
+0x00000566,
+0x00000566,
+0x00000566,
+0x00000818,
+0x00000566,
+0x00000566,
+0x000002b3,
+0x00000566,
+0x00000818,
+0x000002b3,
+0x00000566,
+0x00000818,
+0x00000818,
+0x00000acb,
+0x00000acb,
+0x000002b3,
+0x00000818,
+0x000002b3,
+0x00000acb,
+0x00000818
+};
+
+
+/* -----------EXPECTED OUTPUTS----------- */
+uint8_t exp_outp_0 [] = {
+0xbb,
+0x94,
+0x64,
+0xc4,
+0x8c,
+0x6e,
+0x90,
+0xff,
+0x34,
+0xbe,
+0x8a,
+0x76,
+0x5f,
+0x58,
+0x58,
+0x88,
+0x6a,
+0x6b,
+0x34,
+0x5f,
+0x77,
+0x30,
+0x66,
+0xa7,
+0xa9,
+0xbe,
+0xe5,
+0x33,
+0xc3,
+0x2d,
+0xbc,
+0x80,
+0x91,
+0x97,
+0x53,
+0xb6,
+0x25,
+0x55,
+0xa3,
+0xc9,
+0x29,
+0x7f,
+0x79,
+0x66,
+0x58,
+0x3e,
+0x3c,
+0x79,
+0x46,
+0x68,
+0x2d,
+0x51,
+0x5d,
+0x2a,
+0x65,
+0x92,
+0x5b,
+0x6c,
+0xbd,
+0x33,
+0x97,
+0x26,
+0x79,
+0x7e,
+0xb6,
+0x98,
+0x45,
+0xa0,
+0x37,
+0x54,
+0x9a,
+0xd5,
+0x20,
+0x87,
+0x7d,
+0x69,
+0x5f,
+0x43,
+0x4b,
+0x7e,
+0x49,
+0x77,
+0x26,
+0x5f,
+0x54,
+0x2b,
+0x77,
+0x91,
+0x71,
+0x94,
+0xec,
+0x27,
+0x85,
+0x27,
+0x78,
+0x6a,
+0xe0,
+0xb9,
+0x76,
+0xb4,
+0x9e,
+0x6b,
+0xc0,
+0xf0,
+0x34,
+0xd6,
+0x7d,
+0x7a,
+0x5a,
+0x55,
+0x69,
+0x86,
+0x52,
+0x7e,
+0x34,
+0x6b,
+0x80,
+0x39,
+0x80,
+0xc3,
+0x90,
+0xc0,
+0xf0,
+0x3d,
+0xb4,
+0x38,
+0xc6,
+0xb3,
+0xa9,
+0x7c,
+0x4e,
+0x5d,
+0x68,
+0x74,
+0xa2,
+0xc5,
+0x26,
+0xac,
+0x7d,
+0x5e,
+0x44,
+0x48,
+0x55,
+0x77,
+0x52,
+0x6e,
+0x30,
+0x60,
+0x5d,
+0x1f,
+0x57,
+0x84,
+0x8f,
+0x99,
+0xa4,
+0x29,
+0x97,
+0x2e,
+0x80,
+0x6f,
+0x69,
+0x63,
+0x42,
+0x35,
+0x00,
+0x61,
+0x68,
+0x9c,
+0x21,
+0x82,
+0x36,
+0x38,
+0x10,
+0x24,
+0x48,
+0x39,
+0x16,
+0x55,
+0x14,
+0x2c,
+0x28,
+0x22,
+0x43,
+0x58,
+0x41,
+0x4b,
+0x8c,
+0x21,
+0x53,
+0x22,
+0x58,
+0x49,
+0x72,
+0x62,
+0x27,
+0x13,
+0x0b,
+0x52,
+0x89,
+0x7b,
+0x19,
+0x93,
+0x6b,
+0x4c,
+0x1b,
+0x22,
+0x41,
+0x36,
+0x1a,
+0x61,
+0x1e,
+0x29,
+0x4b,
+0x12,
+0x54,
+0x5d,
+0x49,
+0x68,
+0x93,
+0x1d,
+0x72,
+0x26,
+0x5a,
+0x5b,
+0xe8,
+0x9d,
+0x64,
+0x8b,
+0x71,
+0x65,
+0xb3,
+0xb4,
+0x2e,
+0xd2,
+0x8e,
+0x69,
+0x5b,
+0x48,
+0x6c,
+0x6e,
+0x33,
+0x70,
+0x2c,
+0x57,
+0x67,
+0x39,
+0x75,
+0xb2,
+0x80,
+0xba,
+0xdf,
+0x38,
+0x9c,
+0x33,
+0xc1,
+0x94,
+0x92,
+0x67,
+0x49,
+0x90,
+0x5e,
+0x75,
+0x78,
+0xd7,
+0x2b,
+0xaa,
+0x65,
+0x72,
+0x43,
+0x41,
+0x56,
+0x5e,
+0x4c,
+0x55,
+0x2d,
+0x5a,
+0x6a,
+0x25,
+0x54,
+0x97,
+0x89,
+0xcb,
+0xab,
+0x24,
+0x9f,
+0x27,
+0x84,
+0x6b,
+0x87,
+0x73,
+0x3b,
+0x25,
+0x00,
+0x62,
+0x63,
+0x82,
+0x16,
+0x7b,
+0x38,
+0x47,
+0x22,
+0x1c,
+0x49,
+0x57,
+0x24,
+0x58,
+0x26,
+0x33,
+0x25,
+0x1c,
+0x54,
+0x81,
+0x31,
+0x5f,
+0x7f,
+0x28,
+0x6f,
+0x23,
+0x52,
+0x5b,
+0x8c,
+0x65,
+0x39,
+0x34,
+0x05,
+0x5d,
+0x7f,
+0x7f,
+0x1c,
+0x7c,
+0x4d,
+0x73,
+0x11,
+0x21,
+0x3e,
+0x5d,
+0x26,
+0x55,
+0x1a,
+0x58,
+0x1d,
+0x22,
+0x43,
+0x65,
+0x42,
+0x56,
+0xac,
+0x26,
+0x5c,
+0x2c,
+0x4c,
+0x5d,
+0xf7,
+0x87,
+0x58,
+0x65,
+0x59,
+0x62,
+0x95,
+0xa8,
+0x27,
+0xad,
+0x80,
+0x5b,
+0x51,
+0x54,
+0x70,
+0x6a,
+0x34,
+0x77,
+0x27,
+0x75,
+0x75,
+0x30,
+0x6c,
+0x95,
+0x88,
+0xa9,
+0xca,
+0x2c,
+0x7c,
+0x37,
+0x8f,
+0x84,
+0xe0,
+0xa4,
+0x62,
+0x7d,
+0xaa,
+0x77,
+0xa7,
+0xfd,
+0x2f,
+0xc5,
+0x8a,
+0x79,
+0x58,
+0x68,
+0x6b,
+0x80,
+0x55,
+0x7c,
+0x43,
+0x7b,
+0x84,
+0x2f,
+0x69,
+0xa7,
+0xa8,
+0xf3,
+0xbb,
+0x35,
+0xbd,
+0x30,
+0xd8,
+0x82,
+0x9a,
+0xa9,
+0x5b,
+0x60,
+0x7f,
+0x87,
+0x82,
+0xc9,
+0x2c,
+0xcf,
+0x7f,
+0x71,
+0x51,
+0x63,
+0x61,
+0x7d,
+0x35,
+0x6a,
+0x31,
+0x5f,
+0x63,
+0x38,
+0x50,
+0xaf,
+0x6d,
+0xd0,
+0x8d,
+0x31,
+0x88,
+0x31,
+0xc9,
+0x8b,
+0xaa,
+0x8e,
+0x6a,
+0x6c,
+0x82,
+0x78,
+0x83,
+0xc4,
+0x31,
+0xdc,
+0x77,
+0x77,
+0x4c,
+0x57,
+0x6c,
+0x8b,
+0x47,
+0x6b,
+0x3b,
+0x69,
+0x7d,
+0x36,
+0x52,
+0xab,
+0x7b,
+0xda,
+0xbc,
+0x33,
+0xa2,
+0x2e,
+0xbe,
+0x75,
+0xf6,
+0xab,
+0x6a,
+0xb6,
+0xad,
+0x7b,
+0xa4,
+0xc6,
+0x33,
+0xe6,
+0x8e,
+0x7b,
+0x77,
+0x62,
+0x79,
+0x84,
+0x4a,
+0x79,
+0x34,
+0x7d,
+0x88,
+0x36,
+0x6a,
+0xbf,
+0x88,
+0xf9,
+0xdf,
+0x3d,
+0x98,
+0x3a,
+0xfd,
+0xaa
+};
+
+
+
+L1_DATA uint32_t threshs_l1 [32] = {0};
+L1_DATA int32_t kappa_l1 [32] = {0};
+L1_DATA int32_t lambda_l1 [32] = {0};
+L1_DATA uint8_t inp_l1 [128] = {0};
+L1_DATA uint8_t outp_l1 [512] = {0};
+L1_DATA int8_t wt_l1 [1152] = {0};
+L1_DATA uint8_t im2col_l1 [IM2COL_DIM] = {0};
+
+#endif
diff --git a/rt_nn_tests/xpnn_conv/pulp_nn_kernels.h b/rt_nn_tests/xpnn_conv/pulp_nn_kernels.h
new file mode 100644
index 0000000..f4ec977
--- /dev/null
+++ b/rt_nn_tests/xpnn_conv/pulp_nn_kernels.h
@@ -0,0 +1,26528 @@
+/*
+ * pulp_nn_kernels.h
+ * Nazareno Bruschi <nazareno.bruschi@unibo.it>
+ *
+ * Copyright (C) 2019-2020 University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef __XPULPNN_KERNELS__
+#define __XPULPNN_KERNELS__
+
+void xpulp_nn_conv_u8_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+#ifndef PROFILE
+                        uint8_t flag_batch_norm);
+#else
+  uint8_t flag_batch_norm,
+    uint32_t *im2col_cycles,
+    uint32_t *hotloop_prep_cycles,
+    uint32_t *hotloop_cycles,
+    uint32_t *requant_cycles,
+    uint32_t *hotloop_leftover_cycles,
+    uint32_t *matmul4x2_leftover_cycles);
+#endif
+                        
+
+void xpulp_nn_conv_i2_i2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_i8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_i8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_i8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_i4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_i4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_i4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_i2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_i2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_i2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_i8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_i8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_i8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_i4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_i4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_i4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_i2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_i2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_i2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_i8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_i8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_i8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_i4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_i4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_i4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_i2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_i2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_i2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u8_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i8_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u8_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i8_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u8_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i8_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u4_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i4_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u4_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i4_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u4_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i4_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u2_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i2_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u2_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i2_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u2_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i2_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u8_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i8_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u8_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i8_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u8_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i8_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u4_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i4_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u4_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i4_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u4_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i4_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u2_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i2_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u2_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i2_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u2_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i2_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u8_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i8_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u8_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i8_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u8_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i8_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u4_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i4_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u4_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i4_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u4_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i4_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u2_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i2_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u2_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i2_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+#ifndef PROFILE
+                        uint8_t flag_batch_norm);
+#else
+  uint8_t flag_batch_norm,
+    uint32_t *hotloop_prep_cycles,
+    uint32_t *hotloop_cycles,
+    uint32_t *requant_cycles,
+    uint32_t *hotloop_leftover_cycles);
+#endif
+
+uint8_t *xpulp_nn_matmul_u2_u2_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i2_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_i8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_i8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_i8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_i4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_i4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_i4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_i2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_i2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_i2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_i8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_i8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_i8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_i4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_i4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_i4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_i2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_i2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_i2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_i8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_i8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_i8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_i4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_i4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_i4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_i2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_i2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_i2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_i32_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_i8_i32_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_u8_i32_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_i8_i32_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_u8_i32_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_i8_i32_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_u4_i32_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_i4_i32_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_u4_i32_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_i4_i32_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_u4_i32_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_i4_i32_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_u2_i32_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_i2_i32_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_u2_i32_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_i2_i32_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_u2_i32_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_i2_i32_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_u8_u8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_i8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_u8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_i8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_u8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_i8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_u4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_i4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_u4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_i4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_u4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_i4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_u2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_i2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_u2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_i2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_u2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_i2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_u8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_i8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_u8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_i8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_u8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_i8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_u4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_i4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_u4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_i4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_u4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_i4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_u2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_i2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_u2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_i2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_u2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_i2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_u8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_i8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_u8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_i8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_u8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_i8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_u4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_i4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_u4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_i4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_u4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_i4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_u2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_i2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_u2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_i2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_u2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_i2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_maxpool_u8(
+                        uint8_t * pIn,
+                        uint8_t * pOut,
+                        uint16_t  dim_im_in_x,
+                        uint16_t  dim_im_in_y,
+                        uint16_t  ch_im_in,
+                        uint16_t  dim_im_out_x,
+                        uint16_t  dim_im_out_y,
+                        uint16_t  dim_kernel_x,
+                        uint16_t  dim_kernel_y,
+                        uint16_t  padding_t,
+                        uint16_t  padding_b,
+                        uint16_t  padding_l,
+                        uint16_t  padding_r,
+                        uint16_t  stride_x,
+                        uint16_t  stride_y);
+
+void xpulp_nn_maxpool_i8(
+                        int8_t * pIn,
+                        int8_t * pOut,
+                        uint16_t  dim_im_in_x,
+                        uint16_t  dim_im_in_y,
+                        uint16_t  ch_im_in,
+                        uint16_t  dim_im_out_x,
+                        uint16_t  dim_im_out_y,
+                        uint16_t  dim_kernel_x,
+                        uint16_t  dim_kernel_y,
+                        uint16_t  padding_t,
+                        uint16_t  padding_b,
+                        uint16_t  padding_l,
+                        uint16_t  padding_r,
+                        uint16_t  stride_x,
+                        uint16_t  stride_y);
+
+void xpulp_nn_maxpool_u4(
+                        uint8_t * pIn,
+                        uint8_t * pOut,
+                        uint16_t  dim_im_in_x,
+                        uint16_t  dim_im_in_y,
+                        uint16_t  ch_im_in,
+                        uint16_t  dim_im_out_x,
+                        uint16_t  dim_im_out_y,
+                        uint16_t  dim_kernel_x,
+                        uint16_t  dim_kernel_y,
+                        uint16_t  padding_t,
+                        uint16_t  padding_b,
+                        uint16_t  padding_l,
+                        uint16_t  padding_r,
+                        uint16_t  stride_x,
+                        uint16_t  stride_y);
+
+void xpulp_nn_maxpool_i4(
+                        int8_t * pIn,
+                        int8_t * pOut,
+                        uint16_t  dim_im_in_x,
+                        uint16_t  dim_im_in_y,
+                        uint16_t  ch_im_in,
+                        uint16_t  dim_im_out_x,
+                        uint16_t  dim_im_out_y,
+                        uint16_t  dim_kernel_x,
+                        uint16_t  dim_kernel_y,
+                        uint16_t  padding_t,
+                        uint16_t  padding_b,
+                        uint16_t  padding_l,
+                        uint16_t  padding_r,
+                        uint16_t  stride_x,
+                        uint16_t  stride_y);
+
+void xpulp_nn_maxpool_u2(
+                        uint8_t * pIn,
+                        uint8_t * pOut,
+                        uint16_t  dim_im_in_x,
+                        uint16_t  dim_im_in_y,
+                        uint16_t  ch_im_in,
+                        uint16_t  dim_im_out_x,
+                        uint16_t  dim_im_out_y,
+                        uint16_t  dim_kernel_x,
+                        uint16_t  dim_kernel_y,
+                        uint16_t  padding_t,
+                        uint16_t  padding_b,
+                        uint16_t  padding_l,
+                        uint16_t  padding_r,
+                        uint16_t  stride_x,
+                        uint16_t  stride_y);
+
+void xpulp_nn_maxpool_i2(
+                        int8_t * pIn,
+                        int8_t * pOut,
+                        uint16_t  dim_im_in_x,
+                        uint16_t  dim_im_in_y,
+                        uint16_t  ch_im_in,
+                        uint16_t  dim_im_out_x,
+                        uint16_t  dim_im_out_y,
+                        uint16_t  dim_kernel_x,
+                        uint16_t  dim_kernel_y,
+                        uint16_t  padding_t,
+                        uint16_t  padding_b,
+                        uint16_t  padding_l,
+                        uint16_t  padding_r,
+                        uint16_t  stride_x,
+                        uint16_t  stride_y);
+
+void xpulp_nn_avgpool_u8_u8(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u8_i8(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i8_u8(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i8_i8(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u8_u4(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u8_i4(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i8_u4(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i8_i4(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u8_u2(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u8_i2(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i8_u2(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i8_i2(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u4_u8(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u4_i8(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i4_u8(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i4_i8(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u4_u4(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u4_i4(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i4_u4(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i4_i4(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u4_u2(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u4_i2(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i4_u2(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i4_i2(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u2_u8(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u2_i8(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i2_u8(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i2_i8(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u2_u4(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u2_i4(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i2_u4(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i2_i4(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u2_u2(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u2_i2(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i2_u2(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i2_i2(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_add_u8_u8(
+                        uint8_t * pIn1,
+                        uint8_t * pIn2,
+                        uint8_t * pOut,
+                        uint16_t out_mult1,
+                        uint16_t out_mult2,
+                        uint16_t out_shift,
+                        uint16_t dim_im_in_x,
+                        uint16_t dim_im_in_y,
+                        uint16_t ch_im_in);
+
+void xpulp_nn_add_u8_u8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u4(
+                        uint8_t * pIn1,
+                        uint8_t * pIn2,
+                        uint8_t * pOut,
+                        uint16_t out_mult1,
+                        uint16_t out_mult2,
+                        uint16_t out_shift,
+                        uint16_t dim_im_in_x,
+                        uint16_t dim_im_in_y,
+                        uint16_t ch_im_in);
+
+void xpulp_nn_add_u8_u4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u2(
+                        uint8_t * pIn1,
+                        uint8_t * pIn2,
+                        uint8_t * pOut,
+                        uint16_t out_mult1,
+                        uint16_t out_mult2,
+                        uint16_t out_shift,
+                        uint16_t dim_im_in_x,
+                        uint16_t dim_im_in_y,
+                        uint16_t ch_im_in);
+
+void xpulp_nn_add_u8_u2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u4(
+                        uint8_t * pIn1,
+                        uint8_t * pIn2,
+                        uint8_t * pOut,
+                        uint16_t out_mult1,
+                        uint16_t out_mult2,
+                        uint16_t out_shift,
+                        uint16_t dim_im_in_x,
+                        uint16_t dim_im_in_y,
+                        uint16_t ch_im_in);
+
+void xpulp_nn_add_u4_u4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u2(
+                        uint8_t * pIn1,
+                        uint8_t * pIn2,
+                        uint8_t * pOut,
+                        uint16_t out_mult1,
+                        uint16_t out_mult2,
+                        uint16_t out_shift,
+                        uint16_t dim_im_in_x,
+                        uint16_t dim_im_in_y,
+                        uint16_t ch_im_in);
+
+void xpulp_nn_add_u4_u2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u2(
+                        uint8_t * pIn1,
+                        uint8_t * pIn2,
+                        uint8_t * pOut,
+                        uint16_t out_mult1,
+                        uint16_t out_mult2,
+                        uint16_t out_shift,
+                        uint16_t dim_im_in_x,
+                        uint16_t dim_im_in_y,
+                        uint16_t ch_im_in);
+
+void xpulp_nn_add_u2_u2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+
+
+#endif
diff --git a/rt_nn_tests/xpnn_conv/pulp_nn_mix_kernels.h b/rt_nn_tests/xpnn_conv/pulp_nn_mix_kernels.h
new file mode 100644
index 0000000..8b2a3c6
--- /dev/null
+++ b/rt_nn_tests/xpnn_conv/pulp_nn_mix_kernels.h
@@ -0,0 +1,7093 @@
+/*
+ * pulp_nn_kernels.h
+ * Nazareno Bruschi <nazareno.bruschi@unibo.it>
+ *
+ * Copyright (C) 2019-2020 University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef __XPULPNN-MIXED_KERNELS__
+#define __XPULPNN-MIXED_KERNELS__
+
+void xpulp_nn_mix_conv_u8_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u8_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u8_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u8_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u4_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u4_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u4_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u2_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u2_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u2_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u8_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u8_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u8_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u4_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u4_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u4_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u2_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u2_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u2_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u8_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u8_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u8_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u4_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u4_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u4_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u2_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u2_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u2_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_avgpool_u8_u8(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u8_i8(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i8_u8(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i8_i8(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u8_u4(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u8_i4(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i8_u4(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i8_i4(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u8_u2(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u8_i2(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i8_u2(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i8_i2(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u4_u8(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u4_i8(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i4_u8(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i4_i8(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u4_u4(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u4_i4(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i4_u4(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i4_i4(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u4_u2(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u4_i2(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i4_u2(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i4_i2(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u2_u8(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u2_i8(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i2_u8(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i2_i8(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u2_u4(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u2_i4(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i2_u4(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i2_i4(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u2_u2(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u2_i2(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i2_u2(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i2_i2(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_add_u8_u8(
+                        uint8_t * pIn1,
+                        uint8_t * pIn2,
+                        uint8_t * pOut,
+                        uint16_t out_mult1,
+                        uint16_t out_mult2,
+                        uint16_t out_shift,
+                        uint16_t dim_im_in_x,
+                        uint16_t dim_im_in_y,
+                        uint16_t ch_im_in);
+
+void xpulp_nn_add_u8_u8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u4(
+                        uint8_t * pIn1,
+                        uint8_t * pIn2,
+                        uint8_t * pOut,
+                        uint16_t out_mult1,
+                        uint16_t out_mult2,
+                        uint16_t out_shift,
+                        uint16_t dim_im_in_x,
+                        uint16_t dim_im_in_y,
+                        uint16_t ch_im_in);
+
+void xpulp_nn_add_u8_u4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u2(
+                        uint8_t * pIn1,
+                        uint8_t * pIn2,
+                        uint8_t * pOut,
+                        uint16_t out_mult1,
+                        uint16_t out_mult2,
+                        uint16_t out_shift,
+                        uint16_t dim_im_in_x,
+                        uint16_t dim_im_in_y,
+                        uint16_t ch_im_in);
+
+void xpulp_nn_add_u8_u2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u4(
+                        uint8_t * pIn1,
+                        uint8_t * pIn2,
+                        uint8_t * pOut,
+                        uint16_t out_mult1,
+                        uint16_t out_mult2,
+                        uint16_t out_shift,
+                        uint16_t dim_im_in_x,
+                        uint16_t dim_im_in_y,
+                        uint16_t ch_im_in);
+
+void xpulp_nn_add_u4_u4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u2(
+                        uint8_t * pIn1,
+                        uint8_t * pIn2,
+                        uint8_t * pOut,
+                        uint16_t out_mult1,
+                        uint16_t out_mult2,
+                        uint16_t out_shift,
+                        uint16_t dim_im_in_x,
+                        uint16_t dim_im_in_y,
+                        uint16_t ch_im_in);
+
+void xpulp_nn_add_u4_u2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u2(
+                        uint8_t * pIn1,
+                        uint8_t * pIn2,
+                        uint8_t * pOut,
+                        uint16_t out_mult1,
+                        uint16_t out_mult2,
+                        uint16_t out_shift,
+                        uint16_t dim_im_in_x,
+                        uint16_t dim_im_in_y,
+                        uint16_t ch_im_in);
+
+void xpulp_nn_add_u2_u2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+
+
+#endif
\ No newline at end of file
diff --git a/rt_nn_tests/xpnn_conv/pulp_nn_utils.h b/rt_nn_tests/xpnn_conv/pulp_nn_utils.h
new file mode 100644
index 0000000..44d2251
--- /dev/null
+++ b/rt_nn_tests/xpnn_conv/pulp_nn_utils.h
@@ -0,0 +1,2079 @@
+/*
+ * pulp_nn_utils.h
+ * Nazareno   Bruschi  <nazareno.bruschi@unibo.it>
+ * Alessandro Nadalini <alessandro.nadalini3@unibo.it>
+ * Georg Rutishauser   <georgr@iis.ee.ethz.ch>
+ *
+ * Copyright (C) 2019-2020 ETH Zurich & University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef __PULPNN_UTILS__
+#define __PULPNN_UTILS__
+
+#include <stdint.h>
+
+typedef   signed short v2s __attribute__((vector_size (4)));
+
+
+
+#define min(a,b)                                             ((a)<(b)?(a):(b))
+#define log2(x)                                              __builtin_pulp_fl1(x)
+#define PACK_INT2_SIZE(x)                                    ((x) >> 2)
+
+#define MemoryFence()                                        asm volatile("":::"memory")
+
+#define CHANS_DECOMPR(x)                                     (5*x >> 2) // equivalent to division by 0.8
+
+/* Functions for Compressed MAC */
+#define CompressedMAC(sum, ptr, config) asm volatile(                \
+    "pv.smlsdotsp.t %[shum], %[phtr], %[chonfig];"                    \
+    : [shum] "+r" (sum), [phtr] "+r" (ptr): [chonfig] "I" (config))
+
+#define CompressedMACUnsigned(sum, ptr, config) asm volatile(                   \
+    "pv.smlsdotsup.t %[shum], %[phtr], %[chonfig];" \
+    : [shum] "+r" (sum), [phtr] "+r" (ptr): [chonfig] "I" (config))
+
+#define InitNNRF(ptr, config) asm volatile(        \
+    "pv.smlsdotsp.t x0, %[phtr], %[chonfig];"       \
+    : [phtr] "+r" (ptr) : [chonfig] "I" (config))
+
+#define ThresholdCompress(res, val, thrs) asm volatile(                                                \
+    "pv.thrc %[rhes], %[vhal], %[thhrs];" : [rhes] "+r" (res) : [vhal] "r" (val), [thhrs] "r" (thrs))
+
+#define GetConfig(a_update, b_update, a_reg, b_reg) a_update << 4 | b_update << 3 | a_reg << 1 | b_reg
+
+/* Functions for threshold&compress */
+#define check_store(res, pOut)            \
+  if ((res & 0xe0000000) == 0x00000000) { \
+    *pOut = res & 0xff;                   \
+    pOut++;                               \
+    incr_val=ch_out_r; }
+
+#define check_store_4x1(res, pOut)        \
+  if ((res & 0xe0000000) == 0x00000000) { \
+    *pOut = res & 0xff;                   \
+    pOut++; }
+
+#define reset_currThr()                           \
+  if ((uint32_t *) currThr == (uint32_t *) (pThr + ch_out)) { \
+    currThr = (v2s *) pThr;                                    \
+  }
+
+#define MacLoads20(a_update, b_update, a_reg, b_reg, ptr, sum)  __builtin_pulp_mlsdotsp20_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+
+#define MacLoad20(a_update, b_update, a_reg, b_reg, ptr, sum)  __builtin_pulp_mlsdotsup20_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+
+/* Functions for compressed min/max */
+#define CompressedMax(res, in1, in2) asm volatile( \
+                                                   "pv.max.t %[res_], %[in1_], %[in2_];" : [res_] "=r" (res) : [in1_] "r" (in1), [in2_] "r" (in2))
+
+#define CompressedMin(res, in1, in2) asm volatile(                      \
+                                                   "pv.min.t %[res_], %[in1_], %[in2_];" : [res_] "=r" (res) : [in1_] "r" (in1), [in2_] "r" (in2))
+
+#define Max16(res, in1, in2) asm volatile(                      \
+                                                   "pv.max.c %[res_], %[in1_], %[in2_];" : [res_] "=r" (res) : [in1_] "r" (in1), [in2_] "r" (in2))
+
+#define Min16(res, in1, in2) asm volatile(                      \
+                                                                        "pv.min.c %[res_], %[in1_], %[in2_];" : [res_] "=r" (res) : [in1_] "r" (in1), [in2_] "r" (in2))
+
+#define thr_cmp(state, val, threshs)  __builtin_pulp_thresh_compr(state, val, threshs)
+
+typedef unsigned char  v4u __attribute__((vector_size (4)));
+typedef   signed char  v4s __attribute__((vector_size (4)));
+
+#define bitext(x,size,off)                                      __builtin_pulp_bextract(x,size,off)
+#define bitextu(x,size,off)                                     __builtin_pulp_bextractu(x,size,off)
+#ifdef __clang__
+#define bitins(dst,not_mask_imm,src,mask_imm,off)               __builtin_binsert(dst,not_mask_imm,src,mask_imm,off)
+#else
+#define bitins(dst,not_mask_imm,src,mask_imm,off)               __builtin_pulp_binsert(dst,not_mask_imm,src,mask_imm,off)
+#endif
+#define pack(x,y,z,t)                                           __builtin_pulp_pack4(x,y,z,t)
+#define max4(a,b)                                               __builtin_pulp_maxu4(a,b)
+#define maxs4(a, b)                                             __builtin_pulp_max4(a, b)
+#define max8(a, b)                                              __builtin_pulp_maxu8(a, b)
+#define maxs8(a, b)                                             __builtin_pulp_max8(a, b)
+#define max16(a, b)                                             __builtin_pulp_maxu16(a, b)
+#define maxs16(a, b)                                            __builtin_pulp_max16(a, b)
+#define maxs20(a, b)                                            __builtin_pulp_max20(a, b)
+#define max32(a,b)                                              __builtin_pulp_maxusi(a,b)
+#define maxs32(a,b)                                             __builtin_pulp_maxsi(a,b)
+#define min32(a,b)                                              __builtin_pulp_minusi(a,b)
+#define mins32(a,b)                                             __builtin_pulp_minsi(a,b)
+#define min4(a, b)                                              __builtin_pulp_minu4(a, b)
+#define mins4(a, b)                                             __builtin_pulp_min4(a, b)
+#define min8(a, b)                                              __builtin_pulp_minu8(a, b)
+#define mins8(a, b)                                             __builtin_pulp_min8(a, b)
+#define min16(a, b)                                             __builtin_pulp_minu16(a, b)
+#define mins16(a, b)                                            __builtin_pulp_min16(a, b)
+#define mins20(a, b)                                            __builtin_pulp_min20(a, b)
+#define avg4(a,b)                                               __builtin_pulp_avgu4(a,b)
+#define avg8(a,b)                                               __builtin_pulp_avgu8(a,b)
+#define avg16(a,b)                                              __builtin_pulp_avgu16(a,b)
+#define log2(x)                                                 __builtin_pulp_fl1(x)
+#define min(a,b)                                                ((a)<(b)?(a):(b))
+#define SumDotp4(a, b, c)                                       __builtin_pulp_sdotusp4(a, b, c)
+#define SumDotp8(a, b, c)                                       __builtin_pulp_sdotusp8(a, b, c)
+#define SumDotp16(a, b, c)                                      __builtin_pulp_sdotusp16(a, b, c)
+#define SumDotps4(a, b, c)                                      __builtin_pulp_sdotsp4(a, b, c)
+#define SumDotps8(a, b, c)                                      __builtin_pulp_sdotsp8(a, b, c)
+#define SumDotps16(a, b, c)                                     __builtin_pulp_sdotsp16(a, b, c)
+#define clip4(x)                                                __builtin_pulp_clipu_r(x, 15)
+#define clip2(x)                                                __builtin_pulp_clipu_r(x, 3)
+#define clip8(x)                                                __builtin_pulp_clipu_r(x, 255)
+
+#define clips4(x)                                               __builtin_pulp_clip_r(x, 7)
+#define clips2(x)                                               __builtin_pulp_clip_r(x, 1)
+#define clips8(x)                                               __builtin_pulp_clip_r(x, 127)
+#define MacLoadInit(a_update, b_update, a_reg, b_reg, ptr)      __builtin_pulp_mlinitspr_v3(a_update, b_update, a_reg, b_reg, ptr)
+#define MacLoadUpdate(ptr)                                      __builtin_pulp_mlupdatespr_v3(ptr)
+#define MacLoadAssign(ptr)                                      __builtin_pulp_mlassignspr_v3(ptr)
+#define MacLoad4(a_update, b_update, a_reg, b_reg, ptr, sum)    __builtin_pulp_mlsdotsup4_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+#define MacLoad8(a_update, b_update, a_reg, b_reg, ptr, sum)    __builtin_pulp_mlsdotsup8_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+#define MacLoad16(a_update, b_update, a_reg, b_reg, ptr, sum)   __builtin_pulp_mlsdotsup16_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+#define MacLoads4(a_update, b_update, a_reg, b_reg, ptr, sum)   __builtin_pulp_mlsdotsp4_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+#define MacLoads8(a_update, b_update, a_reg, b_reg, ptr, sum)   __builtin_pulp_mlsdotsp8_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+#define MacLoads16(a_update, b_update, a_reg, b_reg, ptr, sum)  __builtin_pulp_mlsdotsp16_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+
+#define PACK_INT8_SIZE(x)                                    (x)
+#define PACK_INT4_SIZE(x)                                    ((x) >> 1)
+#define PACK_INT2_SIZE(x)                                    ((x) >> 2)
+
+#define MemoryFence()                                        asm volatile("":::"memory")
+
+#define LEGACY_MODE(x)                                       asm volatile ("csrwi 0x010," x)
+#define IVEC_FMT(x)                                          asm volatile ("csrwi 0x00D," x)
+#define MIXED_SKIP(x)                                        asm volatile ("csrwi 0x00F," x)
+#define A_ADDRESS(x)                                         asm volatile ("csrw 0x100, %0" :: "r" (x))
+#define W_ADDRESS(x)                                         asm volatile ("csrw 0x101, %0" :: "r" (x))
+#define A_STRIDE(x)                                          asm volatile ("csrw 0x102, %0":: "r" (x))
+#define W_STRIDE(x)                                          asm volatile ("csrw 0x103, %0":: "r" (x))
+#define A_ROLLBACK(x)                                        asm volatile ("csrw 0x104, %0":: "r" (x))
+#define W_ROLLBACK(x)                                        asm volatile ("csrw 0x105, %0":: "r" (x))
+#define A_SKIP(x)                                            asm volatile ("csrwi 0x106," x)
+#define W_SKIP(x)                                            asm volatile ("csrwi 0x107," x)
+
+static uint8_t __attribute__((noinline)) pulp_nn_quant_u2 (
+  int32_t phi,
+  int16_t m,
+  int8_t d
+  ) {
+  int32_t x = (m * phi) >> d;
+  uint8_t res = clip2(x);
+  return res;
+}
+static uint8_t __attribute__((noinline)) pulp_nn_bn_quant_u2 (
+  int32_t phi,
+  int32_t k,
+  int32_t lambda,
+  int8_t d
+  ) {
+  int32_t integer_image_phi = (k * phi) + lambda;
+  int32_t x = (integer_image_phi) >> d;
+  uint8_t res = clip2(x);
+  return res;
+}
+static uint8_t __attribute__((noinline)) pulp_nn_add_quant_u2 (
+  uint8_t pix1,
+  uint8_t pix2,
+  int16_t m1,
+  int16_t m2,
+  int8_t d
+  ) {
+  uint32_t integer_image = pix1*m1 + pix2*m2;
+  uint32_t x = (integer_image) >> d;
+  uint8_t res = clip2(x);
+  return res;
+}
+
+static int8_t __attribute__((noinline)) pulp_nn_quant_i2 (
+  int32_t phi,
+  int16_t m,
+  int8_t d
+  ) {
+  int32_t x = (m * phi) >> d;
+  int8_t res = clips2(x);
+  return res;
+}
+static int8_t __attribute__((noinline)) pulp_nn_bn_quant_i2 (
+  int32_t phi,
+  int32_t k,
+  int32_t lambda,
+  int8_t d
+  ) {
+  int32_t integer_image_phi = (k * phi) + lambda;
+  int32_t x = (integer_image_phi) >> d;
+  int8_t res = clips2(x);
+  return res;
+}
+static int8_t __attribute__((noinline)) pulp_nn_add_quant_i2 (
+  int8_t pix1,
+  int8_t pix2,
+  int16_t m1,
+  int16_t m2,
+  int8_t d
+  ) {
+  int32_t integer_image = pix1*m1 + pix2*m2;
+  int32_t x = (integer_image) >> d;
+  int8_t res = clips2(x);
+  return res;
+}
+
+static uint8_t __attribute__((noinline)) pulp_nn_quant_u4 (
+  int32_t phi,
+  int16_t m,
+  int8_t d
+  ) {
+  int32_t x = (m * phi) >> d;
+  uint8_t res = clip4(x);
+  return res;
+}
+static uint8_t __attribute__((noinline)) pulp_nn_bn_quant_u4 (
+  int32_t phi,
+  int32_t k,
+  int32_t lambda,
+  int8_t d
+  ) {
+  int32_t integer_image_phi = (k * phi) + lambda;
+  int32_t x = (integer_image_phi) >> d;
+  uint8_t res = clip4(x);
+  return res;
+}
+static uint8_t __attribute__((noinline)) pulp_nn_add_quant_u4 (
+  uint8_t pix1,
+  uint8_t pix2,
+  int16_t m1,
+  int16_t m2,
+  int8_t d
+  ) {
+  uint32_t integer_image = pix1*m1 + pix2*m2;
+  uint32_t x = (integer_image) >> d;
+  uint8_t res = clip4(x);
+  return res;
+}
+
+static int8_t __attribute__((noinline)) pulp_nn_quant_i4 (
+  int32_t phi,
+  int16_t m,
+  int8_t d
+  ) {
+  int32_t x = (m * phi) >> d;
+  int8_t res = clips4(x);
+  return res;
+}
+static int8_t __attribute__((noinline)) pulp_nn_bn_quant_i4 (
+  int32_t phi,
+  int32_t k,
+  int32_t lambda,
+  int8_t d
+  ) {
+  int32_t integer_image_phi = (k * phi) + lambda;
+  int32_t x = (integer_image_phi) >> d;
+  int8_t res = clips4(x);
+  return res;
+}
+static int8_t __attribute__((noinline)) pulp_nn_add_quant_i4 (
+  int8_t pix1,
+  int8_t pix2,
+  int16_t m1,
+  int16_t m2,
+  int8_t d
+  ) {
+  int32_t integer_image = pix1*m1 + pix2*m2;
+  int32_t x = (integer_image) >> d;
+  int8_t res = clips4(x);
+  return res;
+}
+
+static uint8_t __attribute__((noinline)) pulp_nn_quant_u8 (
+  int32_t phi,
+  int16_t m,
+  int8_t d
+  ) {
+  int32_t x = (m * phi) >> d;
+  uint8_t res = clip8(x);
+  return res;
+}
+static uint8_t __attribute__((noinline)) pulp_nn_bn_quant_u8 (
+  int32_t phi,
+  int32_t k,
+  int32_t lambda,
+  int8_t d
+  ) {
+  int32_t integer_image_phi = (k * phi) + lambda;
+  int32_t x = (integer_image_phi) >> d;
+  uint8_t res = clip8(x);
+  return res;
+}
+static uint8_t __attribute__((noinline)) pulp_nn_add_quant_u8 (
+  uint8_t pix1,
+  uint8_t pix2,
+  int16_t m1,
+  int16_t m2,
+  int8_t d
+  ) {
+  uint32_t integer_image = pix1*m1 + pix2*m2;
+  uint32_t x = (integer_image) >> d;
+  uint8_t res = clip8(x);
+  return res;
+}
+
+static int8_t __attribute__((noinline)) pulp_nn_quant_i8 (
+  int32_t phi,
+  int16_t m,
+  int8_t d
+  ) {
+  int32_t x = (m * phi) >> d;
+  int8_t res = clips8(x);
+  return res;
+}
+static int8_t __attribute__((noinline)) pulp_nn_bn_quant_i8 (
+  int32_t phi,
+  int32_t k,
+  int32_t lambda,
+  int8_t d
+  ) {
+  int32_t integer_image_phi = (k * phi) + lambda;
+  int32_t x = (integer_image_phi) >> d;
+  int8_t res = clips8(x);
+  return res;
+}
+static int8_t __attribute__((noinline)) pulp_nn_add_quant_i8 (
+  int8_t pix1,
+  int8_t pix2,
+  int16_t m1,
+  int16_t m2,
+  int8_t d
+  ) {
+  int32_t integer_image = pix1*m1 + pix2*m2;
+  int32_t x = (integer_image) >> d;
+  int8_t res = clips8(x);
+  return res;
+}
+
+
+static uint8_t __attribute__((noinline)) pulp_nn_u4_quant(int input, int16_t * pThr)
+{
+  if(input <= pThr[7] )
+  {
+    if( input <= pThr[3])
+    {
+      if( input <= pThr[1])
+      {
+        if( input <= pThr[0])
+          return 0;
+        else
+          return 1;
+      }
+      else
+      {
+        if( input <= pThr[2])
+          return 2;
+        else
+          return 3;
+      }
+    }
+    else
+    {
+      if( input <= pThr[5])
+      {
+        if( input <= pThr[4])
+          return 4;
+        else
+          return 5;
+      }
+      else
+      {
+        if( input <= pThr[6])
+          return 6;
+        else
+          return 7;
+      }
+    }
+  }
+  else
+  {
+    if( input <= pThr[11])
+    {
+      if( input <= pThr[9])
+      {
+        if( input <= pThr[8])
+          return 8;
+        else
+          return 9;
+      }
+      else
+      {
+        if( input <= pThr[10])
+          return 10;
+        else
+          return 11;
+      }
+    }
+    else
+    {
+      if( input <= pThr[13])
+      {
+        if( input <= pThr[12])
+          return 12;
+        else
+          return 13;
+      }
+      else
+      {
+        if( input <= pThr[14])
+          return 14;
+        else
+          return 15;
+      }
+    }
+  }
+}
+
+static uint8_t __attribute__((noinline)) pulp_nn_u2_quant(int input, int16_t * pThr)
+{
+  if( input <= pThr[1])
+  {
+    if( input <= pThr[0])
+        {
+      return 0;
+    }
+        else
+        {
+      return 1;
+    }
+  }
+    else
+    {
+    if( input <= pThr[2])
+        {
+      return 2;
+    }
+        else
+        {
+      return 3;
+    }
+  }
+}
+
+/*
+ * Common
+ */
+
+
+static v4s __attribute__((noinline)) pulp_nn_i4_to_i8_r( int8_t *pSrc)
+{
+  v4s Src = *((v4s*) pSrc);
+  int8_t bext1, bext2, bext3, bext4;
+
+  bext1 = (int8_t) bitext((int) Src, 4, 0);
+  bext2 = (int8_t) bitext((int) Src, 4, 4);
+  bext3 = (int8_t) bitext((int) Src, 4, 8);
+  bext4 = (int8_t) bitext((int) Src, 4, 12);
+  v4s res = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+
+  return res;
+}
+
+static v4u __attribute__((noinline)) pulp_nn_u4_to_u8_r(uint8_t *pSrc)
+{
+  v4u Src = *((v4u*) pSrc);
+  uint8_t bext1, bext2, bext3, bext4;
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 4, 0);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 4, 4);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 4, 8);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 4, 12);
+  v4u res = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+
+  return res;
+}
+
+static v4s __attribute__((noinline)) pulp_nn_i2_to_i8_r( int8_t *pSrc)
+{
+  v4s Src = *((v4s*) pSrc);
+  int8_t bext1, bext2, bext3, bext4;
+
+  bext1 = (int8_t) bitext((int) Src, 2, 0);
+  bext2 = (int8_t) bitext((int) Src, 2, 2);
+  bext3 = (int8_t) bitext((int) Src, 2, 4);
+  bext4 = (int8_t) bitext((int) Src, 2, 6);
+  v4s res = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+
+  return res;
+}
+
+static v4u __attribute__((noinline)) pulp_nn_u2_to_u8_r(uint8_t *pSrc)
+{
+  v4u Src = *((v4u*) pSrc);
+  uint8_t bext1, bext2, bext3, bext4;
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 0);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 2);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 4);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 6);
+  v4u res = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+
+  return res;
+}
+
+static v4s __attribute__((noinline)) pulp_nn_i2_to_i4_r(int8_t *pSrc)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  v4s Src = *((v4s*) pSrc);
+  int8_t bext1, bext2, bext3, bext4;
+  int8_t out1, out2, out3, out4;
+
+  bext1 = (int8_t) bitextu((int) Src, 2, 0);
+  bext2 = (int8_t) bitextu((int) Src, 2, 2);
+  bext3 = (int8_t) bitextu((int) Src, 2, 4);
+  bext4 = (int8_t) bitextu((int) Src, 2, 6);
+
+  out1 = bitins(bext1, n_mask, bext2, mask, off);
+  out2 = bitins(bext3, n_mask, bext4, mask, off);
+
+  bext1 = (int8_t) bitextu((int) Src, 2, 8);
+  bext2 = (int8_t) bitextu((int) Src, 2, 10);
+  bext3 = (int8_t) bitextu((int) Src, 2, 12);
+  bext4 = (int8_t) bitextu((int) Src, 2, 14);
+
+  out3 = bitins(bext1, n_mask, bext2, mask, off);
+  out4 = bitins(bext3, n_mask, bext4, mask, off);
+
+  v4s res = pack((int8_t) out1, (int8_t) out2, (int8_t) out3, (int8_t) out4);
+
+  return res;
+}
+
+static v4u __attribute__((noinline)) pulp_nn_u2_to_u4_r(uint8_t *pSrc)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  v4u Src = *((v4u*) pSrc);
+  uint8_t bext1, bext2, bext3, bext4;
+  uint8_t out1, out2, out3, out4;
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 0);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 2);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 4);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 6);
+
+  out1 = bitins(bext1, n_mask, bext2, mask, off);
+  out2 = bitins(bext3, n_mask, bext4, mask, off);
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 8);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 10);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 12);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 14);
+
+  out3 = bitins(bext1, n_mask, bext2, mask, off);
+  out4 = bitins(bext3, n_mask, bext4, mask, off);
+
+  v4u res = pack((uint8_t) out1, (uint8_t) out2, (uint8_t) out3, (uint8_t) out4);
+
+  return res;
+}
+
+static int8_t *__attribute__((always_inline)) pulp_nn_i4_to_i8( int8_t *pSrc, int8_t *pDst)
+{
+  v4s Src = *((v4s*) pSrc);
+  int8_t bext1, bext2, bext3, bext4;
+
+  pSrc+=4;
+
+  bext1 = (int8_t) bitext((int) Src, 4, 0);
+  bext2 = (int8_t) bitext((int) Src, 4, 4);
+  bext3 = (int8_t) bitext((int) Src, 4, 8);
+  bext4 = (int8_t) bitext((int) Src, 4, 12);
+  *((v4s*)pDst) = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+   MemoryFence();
+  pDst+=4;
+  bext1 = (int8_t) bitext((int) Src, 4, 16);
+  bext2 = (int8_t) bitext((int) Src, 4, 20);
+  bext3 = (int8_t) bitext((int) Src, 4, 24);
+  bext4 = (int8_t) bitext((int) Src, 4, 28);
+  *((v4s*)pDst) = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+
+  return pSrc;
+}
+
+static uint8_t *__attribute__((always_inline)) pulp_nn_u4_to_u8(uint8_t *pSrc, uint8_t *pDst)
+{
+  v4u Src = *((v4u*) pSrc);
+  uint8_t bext1, bext2, bext3, bext4;
+
+  pSrc+=4;
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 4, 0);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 4, 4);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 4, 8);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 4, 12);
+  *((v4u*)pDst) = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 4, 16);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 4, 20);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 4, 24);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 4, 28);
+  *((v4u*)pDst) = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+
+  return pSrc;
+}
+
+static int8_t *__attribute__((always_inline)) pulp_nn_i2_to_i8( int8_t * pSrc, int8_t * pDst)
+{
+  v4s Src = *((v4s*) pSrc);
+  int8_t bext1, bext2, bext3, bext4;
+
+  pSrc+=4;
+
+  bext1 = (int8_t) bitext((int) Src, 2, 0);
+  bext2 = (int8_t) bitext((int) Src, 2, 2);
+  bext3 = (int8_t) bitext((int) Src, 2, 4);
+  bext4 = (int8_t) bitext((int) Src, 2, 6);
+  *((v4s*)pDst) = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (int8_t) bitext((int) Src, 2, 8);
+  bext2 = (int8_t) bitext((int) Src, 2, 10);
+  bext3 = (int8_t) bitext((int) Src, 2, 12);
+  bext4 = (int8_t) bitext((int) Src, 2, 14);
+  *((v4s*)pDst) = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (int8_t) bitext((int) Src, 2, 16);
+  bext2 = (int8_t) bitext((int) Src, 2, 18);
+  bext3 = (int8_t) bitext((int) Src, 2, 20);
+  bext4 = (int8_t) bitext((int) Src, 2, 22);
+  *((v4s*)pDst) = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (int8_t) bitext((int) Src, 2, 24);
+  bext2 = (int8_t) bitext((int) Src, 2, 26);
+  bext3 = (int8_t) bitext((int) Src, 2, 28);
+  bext4 = (int8_t) bitext((int) Src, 2, 30);
+  *((v4s*)pDst) = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+
+  return pSrc;
+}
+
+static uint8_t *__attribute__((always_inline)) pulp_nn_u2_to_u8(uint8_t * pSrc, uint8_t * pDst)
+{
+  v4u Src = *((v4u*) pSrc);
+  uint8_t bext1, bext2, bext3, bext4;
+
+  pSrc+=4;
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 0);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 2);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 4);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 6);
+  *((v4u*)pDst) = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 8);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 10);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 12);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 14);
+  *((v4u*)pDst) = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 16);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 18);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 20);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 22);
+  *((v4u*)pDst) = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 24);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 26);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 28);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 30);
+  *((v4u*)pDst) = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+
+  return pSrc;
+}
+
+static int8_t *__attribute__((always_inline)) pulp_nn_i2_to_i4( int8_t * pSrc, int8_t * pDst)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  v4s Src = *((v4s*) pSrc);
+  int8_t bext1, bext2, bext3, bext4;
+  int8_t out1, out2, out3, out4;
+
+  pSrc+=4;
+
+  bext1 = (int8_t) bitext((int) Src, 2, 0);
+  bext2 = (int8_t) bitext((int) Src, 2, 2);
+  bext3 = (int8_t) bitext((int) Src, 2, 4);
+  bext4 = (int8_t) bitext((int) Src, 2, 6);
+
+  out1 = bitins(bext1, n_mask, bext2, mask, off);
+  out2 = bitins(bext3, n_mask, bext4, mask, off);
+  MemoryFence();
+
+  bext1 = (int8_t) bitext((int) Src, 2, 8);
+  bext2 = (int8_t) bitext((int) Src, 2, 10);
+  bext3 = (int8_t) bitext((int) Src, 2, 12);
+  bext4 = (int8_t) bitext((int) Src, 2, 14);
+
+  out3 = bitins(bext1, n_mask, bext2, mask, off);
+  out4 = bitins(bext3, n_mask, bext4, mask, off);
+
+  *((v4s*)pDst) = pack((int8_t) out1, (int8_t) out2, (int8_t) out3, (int8_t) out4);
+  MemoryFence();
+
+  pDst+=4;
+  bext1 = (int8_t) bitext((int) Src, 2, 16);
+  bext2 = (int8_t) bitext((int) Src, 2, 18);
+  bext3 = (int8_t) bitext((int) Src, 2, 20);
+  bext4 = (int8_t) bitext((int) Src, 2, 22);
+
+  out1 = bitins(bext1, n_mask, bext2, mask, off);
+  out2 = bitins(bext3, n_mask, bext4, mask, off);
+  MemoryFence();
+
+  bext1 = (int8_t) bitext((int) Src, 2, 24);
+  bext2 = (int8_t) bitext((int) Src, 2, 26);
+  bext3 = (int8_t) bitext((int) Src, 2, 28);
+  bext4 = (int8_t) bitext((int) Src, 2, 30);
+
+  out3 = bitins(bext1, n_mask, bext2, mask, off);
+  out4 = bitins(bext3, n_mask, bext4, mask, off);
+
+  *((v4s*)pDst) = pack((int8_t) out1, (int8_t) out2, (int8_t) out3, (int8_t) out4);
+
+  return pSrc;
+}
+
+static uint8_t *__attribute__((always_inline)) pulp_nn_u2_to_u4( uint8_t * pSrc, uint8_t * pDst)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  v4u Src = *((v4u*) pSrc);
+  uint8_t bext1, bext2, bext3, bext4;
+  uint8_t out1, out2, out3, out4;
+
+  pSrc+=4;
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 0);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 2);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 4);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 6);
+
+  out1 = bitins(bext1, n_mask, bext2, mask, off);
+  out2 = bitins(bext3, n_mask, bext4, mask, off);
+  MemoryFence();
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 8);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 10);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 12);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 14);
+
+  out3 = bitins(bext1, n_mask, bext2, mask, off);
+  out4 = bitins(bext3, n_mask, bext4, mask, off);
+
+  *((v4u*)pDst) = pack((uint8_t) out1, (uint8_t) out2, (uint8_t) out3, (uint8_t) out4);
+  MemoryFence();
+
+  pDst+=4;
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 16);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 18);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 20);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 22);
+
+  out1 = bitins(bext1, n_mask, bext2, mask, off);
+  out2 = bitins(bext3, n_mask, bext4, mask, off);
+  MemoryFence();
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 24);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 26);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 28);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 30);
+
+  out3 = bitins(bext1, n_mask, bext2, mask, off);
+  out4 = bitins(bext3, n_mask, bext4, mask, off);
+  
+  *((v4u*)pDst) = pack((uint8_t) out1, (uint8_t) out2, (uint8_t) out3, (uint8_t) out4);
+
+  return pSrc;
+}
+
+/*
+ * XpulpV2
+ */
+
+static void __attribute__((noinline)) pulp_zero_mem(uint8_t * pBuffer, unsigned int size)
+{
+  int lfover = size &0x3;
+  for (int i=0; i<(size>>2); i++)
+  {
+    *((v4u *)pBuffer) = (v4u){0,0,0,0};
+    MemoryFence();
+    pBuffer+=4;
+  }
+  while(lfover)
+  {
+    *pBuffer++=0;
+    lfover--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_im2col_u2_to_u8(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_u2_to_u8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 16;
+  }
+  while (lfover)
+  {
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 0);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 2);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 4);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 6);
+    pOutput++;
+    pInput++;
+    lfover -= 4;
+  }
+}
+static void __attribute__((noinline)) pulp_nn_im2col_i2_to_i8(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_i2_to_i8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 16;
+  }
+  while (lfover)
+  {
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 0);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 2);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 4);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 6);
+    pOutput++;
+    pInput++;
+    lfover -= 4;
+  }
+}
+static void __attribute__((noinline)) pulp_nn_im2col_u4_to_u8(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 3u;
+  int lfover = blockSize & 0x07;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_u4_to_u8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 8;
+  }
+  while (lfover)
+  {
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 4, 0);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 4, 4);
+    pOutput++;
+    pInput++;
+    lfover -= 2;
+  }
+}
+static void __attribute__((noinline)) pulp_nn_im2col_i4_to_i8(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 3u;
+  int lfover = blockSize & 0x07;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_i4_to_i8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 8;
+  }
+  while (lfover)
+  {
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 4, 0);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 4, 4);
+    pOutput++;
+    pInput++;
+    lfover -= 2;
+  }
+}
+static void __attribute__((noinline)) pulp_nn_im2col_u8_to_u8(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 2u;
+  int lfover = blockSize & 0x03;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4u*)pOutput) = *((v4u*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    *((uint8_t*)pOutput) = *((uint8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover--;
+  }
+}
+static void __attribute__((noinline)) pulp_nn_im2col_i8_to_i8(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 2u;
+  int lfover = blockSize & 0x03;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4s*)pOutput) = *((v4s*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    *((int8_t*)pOutput) = *((int8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover--;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_im2col_u2_to_u2(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4u*)pOutput) = *((v4u*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    uint8_t extr;
+    *((uint8_t*)pOutput) = *((uint8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover -= 4;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_i2_to_i2(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4s*)pOutput) = *((v4s*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    int8_t extr;
+    *((int8_t*)pOutput) = *((int8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover -= 4;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_u2_to_u4(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_u2_to_u4(pInput, pOutput);
+    MemoryFence();
+    pOutput += 8;
+  }
+  while (lfover)
+  {
+    uint8_t extr;
+    // TODO: this is potentially dangerous/wrong if the number of channels
+    // is not a multiple of 8!
+    *((v4u *)pOutput) = pulp_nn_u2_to_u4_r(pInput);
+    pInput += 2;
+    pOutput += 4;
+    lfover -= 8;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_i2_to_i4(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_i2_to_i4(pInput, pOutput);
+    MemoryFence();
+    pOutput += 8;
+  }
+  while (lfover)
+  {
+    int8_t extr;
+    // TODO: this is potentially dangerous/wrong if the number of channels
+    // is not a multiple of 8!
+    *((v4s *)pOutput) = pulp_nn_i2_to_i4_r(pInput);
+    pInput += 2;
+    pOutput += 4;
+    lfover -= 8;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_u2_to_u8(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_u2_to_u8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 16;
+  }
+  while (lfover)
+  {
+    uint8_t extr;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 0);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 2);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 4);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 6);
+    pOutput++;
+    pInput++;
+    lfover -= 4;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_i2_to_i8(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_i2_to_i8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 16;
+  }
+  while (lfover)
+  {
+    int8_t extr;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 0);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 2);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 4);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 6);
+    pOutput++;
+    pInput++;
+    lfover -= 4;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_u4_to_u4(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 3u;
+  int lfover = blockSize & 0x07;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4u*)pOutput) = *((v4u*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    uint8_t extr;
+    *((uint8_t*)pOutput) = *((uint8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover -= 2;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_i4_to_i4(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 3u;
+  int lfover = blockSize & 0x07;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4s*)pOutput) = *((v4s*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    int8_t extr;
+    *((int8_t*)pOutput) = *((int8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover -= 2;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_u4_to_u8(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 3u;
+  int lfover = blockSize & 0x07;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_u4_to_u8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 8;
+  }
+  while (lfover)
+  {
+    uint8_t extr;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 4, 0);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 4, 4);
+    pOutput++;
+    pInput++;
+    lfover -= 2;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_i4_to_i8(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 3u;
+  int lfover = blockSize & 0x07;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_i4_to_i8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 8;
+  }
+  while (lfover)
+  {
+    int8_t extr;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 4, 0);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 4, 4);
+    pOutput++;
+    pInput++;
+    lfover -= 2;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_u8_to_u8(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 2u;
+  int lfover = blockSize & 0x03;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4u*)pOutput) = *((v4u*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    uint8_t extr;
+    *((uint8_t*)pOutput) = *((uint8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover -= 1;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_i8_to_i8(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 2u;
+  int lfover = blockSize & 0x03;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4s*)pOutput) = *((v4s*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    int8_t extr;
+    *((int8_t*)pOutput) = *((int8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover -= 1;
+  }
+}
+
+
+
+static void __attribute__((noinline)) pulp_nn_compare_and_replace_if_larger_u8(uint8_t * base,
+                                                uint8_t * target,
+                                                uint16_t length)
+{
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+  v4u inp;
+  v4u com;
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    inp = *((v4u*)pIn);
+    com = *((v4u*)pCom);
+
+    *((v4u*)pIn) = max4(inp, com);
+
+    pCom+=4;
+    pIn+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    if(*pIn<*pCom)
+      *pIn=*pCom;
+    
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_compare_and_replace_if_larger_i8(
+    int8_t * base, int8_t * target, uint16_t length) {
+  int8_t *pIn = base;
+  int8_t *pCom = target;
+  v4s inp;
+  v4s com;
+  int cnt = length >> 2;
+
+  while (cnt > 0u) {
+    inp = *((v4s *)pIn);
+    com = *((v4s *)pCom);
+
+    *((v4s *)pIn) = maxs4(inp, com);
+
+    pCom += 4;
+    pIn += 4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left > 0u) {
+    if (*pIn < *pCom)
+      *pIn = *pCom;
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_avg_and_replace_u8(uint8_t * base,
+                                  uint8_t * target,
+                                  uint16_t length)
+{
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+
+  while (length>0u)
+  {
+    *pIn = ((*pIn + *pCom) >> 1);
+
+    pIn++;
+    pCom++;
+    length--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_compare_and_replace_if_larger_u4(uint8_t * base,
+                                                uint8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+  v4u inp[2];
+  v4u com[2];
+  uint8_t *out;
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    pulp_nn_u4_to_u8(pIn, (uint8_t *)inp);
+    pulp_nn_u4_to_u8(pCom, (uint8_t *)com);
+
+    *((v4u *)out) = max4(inp[0], com[0]);
+
+    *((uint8_t*)pIn) = bitins(*out, n_mask, *(out + 1), mask, off);
+    pIn++;
+    *((uint8_t*)pIn) = bitins(*(out + 2), n_mask, *(out + 3), mask, off);
+    pIn++;
+
+    *((v4u *)out) = max4(inp[1], com[1]);
+
+    *((uint8_t*)pIn) = bitins(*out, n_mask, *(out + 1), mask, off);
+    pIn++;
+    *((uint8_t*)pIn) = bitins(*(out + 2), n_mask, *(out + 3), mask, off);
+    pIn++;
+
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 4, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 4, 4);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 4, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 4, 4);
+
+    if(inA0<inB0)
+      inA0=inB0;
+
+    if(inA1<inB1)
+      inA1=inB1;
+
+    *((uint8_t*)pIn) = bitins(inA0, n_mask, inA1, mask, off);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_compare_and_replace_if_larger_i4(int8_t * base,
+                                                int8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  int8_t *pIn = base;
+  int8_t *pCom = target;
+  v4s inp[2];
+  v4s com[2];
+  int8_t *out;
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    pulp_nn_i4_to_i8(pIn, (int8_t *)inp);
+    pulp_nn_i4_to_i8(pCom, (int8_t *)com);
+
+    *((v4s *)out) = maxs4(inp[0], com[0]);
+
+    *((int8_t*)pIn) = bitins(*out, n_mask, *(out + 1), mask, off);
+    pIn++;
+    *((int8_t*)pIn) = bitins(*(out + 2), n_mask, *(out + 3), mask, off);
+    pIn++;
+
+    *((v4s *)out) = maxs4(inp[1], com[1]);
+
+    *((int8_t*)pIn) = bitins(*out, n_mask, *(out + 1), mask, off);
+    pIn++;
+    *((int8_t*)pIn) = bitins(*(out + 2), n_mask, *(out + 3), mask, off);
+    pIn++;
+
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    int8_t inA0 = (int8_t) bitext((int) *pIn, 4, 0);
+    int8_t inA1 = (int8_t) bitext((int) *pIn, 4, 4);
+    int8_t inB0 = (int8_t) bitext((int) *pCom, 4, 0);
+    int8_t inB1 = (int8_t) bitext((int) *pCom, 4, 4);
+
+    if(inA0<inB0)
+      inA0=inB0;
+
+    if(inA1<inB1)
+      inA1=inB1;
+
+    *((int8_t*)pIn) = bitins(inA0, n_mask, inA1, mask, off);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_avg_and_replace_u4(uint8_t * base,
+                                  uint8_t * target,
+                                  uint16_t length)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+
+  while (length>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 4, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 4, 4);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 4, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 4, 4);
+
+    inA0 = ((inA0 + inB0) >> 1);
+    inA1 = ((inA1 + inB1) >> 1);
+
+    *((uint8_t*)pIn) = bitins(inA0, n_mask, inA1, mask, off);
+
+    pIn++;
+    pCom++;
+    length--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_compare_and_replace_if_larger_u2(uint8_t * base,
+                                                uint8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask2 = 0x0c;
+  int8_t n_mask2 = ~ mask2;
+  int8_t mask4 = 0x30;
+  int8_t n_mask4 = ~ mask4;
+  int8_t mask6 = 0xc0;
+  int8_t n_mask6 = ~ mask6;
+  int8_t off2 = 2;
+  int8_t off4 = 4;
+  int8_t off6 = 6;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+  v4u inp[4];
+  v4u com[4];
+  uint8_t *out;
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    pulp_nn_u2_to_u8(pIn, inp);
+    pulp_nn_u2_to_u8(pCom, com);
+
+    *((v4u*)out) = max4(inp[0], com[0]);
+
+    uint8_t inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    *((v4u*)out) = max4(inp[1], com[1]);
+
+    inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    *((v4u*)out) = max4(inp[2], com[2]);
+
+    inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    *((v4u*)out) = max4(inp[3], com[3]);
+
+    inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 2, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 2, 2);
+    uint8_t inA2 = (uint8_t) bitextu((unsigned int) *pIn, 2, 4);
+    uint8_t inA3 = (uint8_t) bitextu((unsigned int) *pIn, 2, 6);
+    v4u inA4 = pack((uint8_t) inA0, (uint8_t) inA1, (uint8_t) inA2, (uint8_t) inA3);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 2, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 2, 2);
+    uint8_t inB2 = (uint8_t) bitextu((unsigned int) *pCom, 2, 4);
+    uint8_t inB3 = (uint8_t) bitextu((unsigned int) *pCom, 2, 6);
+    v4u inB4 = pack((uint8_t) inB0, (uint8_t) inB1, (uint8_t) inB2, (uint8_t) inB3);
+
+    *((v4u*)out) = max4(inA4, inB4);
+
+    uint8_t inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+
+static void __attribute__((noinline)) pulp_nn_compare_and_replace_if_larger_i2(int8_t * base,
+                                                int8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask2 = 0x0c;
+  int8_t n_mask2 = ~ mask2;
+  int8_t mask4 = 0x30;
+  int8_t n_mask4 = ~ mask4;
+  int8_t mask6 = 0xc0;
+  int8_t n_mask6 = ~ mask6;
+  int8_t off2 = 2;
+  int8_t off4 = 4;
+  int8_t off6 = 6;
+
+  int8_t *pIn = base;
+  int8_t *pCom = target;
+  v4s inp[4];
+  v4s com[4];
+  int8_t *out;
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    pulp_nn_i2_to_i8(pIn, inp);
+    pulp_nn_i2_to_i8(pCom, com);
+
+    *((v4s*)out) = maxs4(inp[0], com[0]);
+
+    int8_t inA = (int8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((int8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    *((v4s*)out) = maxs4(inp[1], com[1]);
+
+    inA = (int8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((int8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    *((v4s*)out) = maxs4(inp[2], com[2]);
+
+    inA = (int8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((int8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    *((v4s*)out) = maxs4(inp[3], com[3]);
+
+    inA = (int8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((int8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    int8_t inA0 = (int8_t) bitext((unsigned int) *pIn, 2, 0);
+    int8_t inA1 = (int8_t) bitext((unsigned int) *pIn, 2, 2);
+    int8_t inA2 = (int8_t) bitext((unsigned int) *pIn, 2, 4);
+    int8_t inA3 = (int8_t) bitext((unsigned int) *pIn, 2, 6);
+    v4s inA4 = pack((int8_t) inA0, (int8_t) inA1, (int8_t) inA2, (int8_t) inA3);
+    int8_t inB0 = (int8_t) bitext((unsigned int) *pCom, 2, 0);
+    int8_t inB1 = (int8_t) bitext((unsigned int) *pCom, 2, 2);
+    int8_t inB2 = (int8_t) bitext((unsigned int) *pCom, 2, 4);
+    int8_t inB3 = (int8_t) bitext((unsigned int) *pCom, 2, 6);
+    v4s inB4 = pack((int8_t) inB0, (int8_t) inB1, (int8_t) inB2, (int8_t) inB3);
+
+    *((v4s*)out) = maxs4(inA4, inB4);
+
+    int8_t inA = (int8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((int8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_avg_and_replace_u2(uint8_t * base,
+                                  uint8_t * target,
+                                  uint16_t length)
+{
+  int8_t mask2 = 0x0c;
+  int8_t n_mask2 = ~ mask2;
+  int8_t mask4 = 0x30;
+  int8_t n_mask4 = ~ mask4;
+  int8_t mask6 = 0xc0;
+  int8_t n_mask6 = ~ mask6;
+  int8_t off2 = 2;
+  int8_t off4 = 4;
+  int8_t off6 = 6;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+
+  while (length>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 2, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 2, 2);
+    uint8_t inA2 = (uint8_t) bitextu((unsigned int) *pIn, 2, 4);
+    uint8_t inA3 = (uint8_t) bitextu((unsigned int) *pIn, 2, 6);
+
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 2, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 2, 2);
+    uint8_t inB2 = (uint8_t) bitextu((unsigned int) *pCom, 2, 4);
+    uint8_t inB3 = (uint8_t) bitextu((unsigned int) *pCom, 2, 6);
+
+    inA0 = ((inA0 + inB0) >> 1);
+    inA1 = ((inA1 + inB1) >> 1);
+    inA2 = ((inA2 + inB2) >> 1);
+    inA3 = ((inA3 + inB3) >> 1);
+
+    uint8_t inA = (uint8_t) bitins(inA0, n_mask2, inA1, mask2, off2);
+    inA = bitins(inA, n_mask4, inA2, mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, inA3, mask6, off6);
+
+    pIn++;
+    pCom++;
+    length--;
+  }
+}
+
+/*
+ * XpulpNN
+ */
+
+static void __attribute__((noinline)) xpulp_nn_zero_mem_u8(uint8_t * pBuffer, unsigned int size)
+{
+  int lfover = size &0x3;
+  for (int i=0; i<(size>>2); i++)
+  {
+    *((v4u *)pBuffer) = (v4u){0,0,0,0};
+    MemoryFence();
+    pBuffer+=4;
+  }
+  while(lfover)
+  {
+    *pBuffer++=0;
+    lfover--;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_zero_mem_u4(uint8_t * pBuffer, unsigned int size)
+{
+  int lfover = size &0x7;
+  for (int i=0; i<(size>>3); i++)
+  {
+    *((v4u *)pBuffer) = (v4u){0,0,0,0};
+    MemoryFence();
+    pBuffer+=4;
+  }
+  while(lfover)
+  {
+    *pBuffer++=0;
+    lfover-=2;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_zero_mem_u2(uint8_t * pBuffer, unsigned int size)
+{
+  int lfover = size &0xf;
+  for (int i=0; i<(size>>4); i++)
+  {
+    *((v4u *)pBuffer) = (v4u){0,0,0,0};
+    MemoryFence();
+    pBuffer+=4;
+  }
+  while(lfover)
+  {
+    *pBuffer++=0;
+    lfover-=4;
+  }
+}
+
+
+static void __attribute__((noinline)) xpulp_tnn_zero_mem_ternary(uint8_t * pBuffer, unsigned int size, unsigned int uns)
+{
+  uint8_t pad_val = 0xd9;
+  uint32_t pad_vec = 0xd9d9d9d9;
+  if (uns) {
+    // if we are using an unsigned kernel, we need to pad with -1 because the hardware will add a +1 to ALL values!
+    pad_val = 0xff;
+    pad_vec = 0xffffffff;
+  }
+  int lfover = size &0xf;
+  for (int i=0; i<(size>>4); i++)
+    {
+      *((v4u *)pBuffer) = (v4u)pad_vec;
+      MemoryFence();
+      pBuffer+=4;
+    }
+  while(lfover)
+    {
+      *pBuffer++=pad_val;
+      lfover-=4;
+    }
+}
+
+static void __attribute__((noinline)) xpulp_nn_compare_and_replace_if_larger_u4(uint8_t * base,
+                                                uint8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    *((uint32_t *)pIn) = max8(*((uint32_t *)pIn), *((uint32_t *)pCom));
+
+    pIn+=4;
+    pCom+=4;
+
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 4, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 4, 4);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 4, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 4, 4);
+
+    if(inA0<inB0)
+      inA0=inB0;
+
+    if(inA1<inB1)
+      inA1=inB1;
+
+    *((uint8_t*)pIn) = bitins(inA0, n_mask, inA1, mask, off);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_compare_and_replace_if_larger_i4(
+    int8_t * base, int8_t * target, uint16_t length) {
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~mask;
+  int8_t off = 0x04;
+
+  int8_t *pIn = base;
+  int8_t *pCom = target;
+
+  int cnt = length >> 2;
+
+  while (cnt > 0u) {
+    *((int32_t *)pIn) = maxs8(*((int32_t *)pIn), *((int32_t *)pCom));
+
+    pIn += 4;
+    pCom += 4;
+
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left > 0u) {
+    int8_t inA0 = (int8_t)bitext((int)*pIn, 4, 0);
+    int8_t inA1 = (int8_t)bitext((int)*pIn, 4, 4);
+    int8_t inB0 = (int8_t)bitext((int)*pCom, 4, 0);
+    int8_t inB1 = (int8_t)bitext((int)*pCom, 4, 4);
+
+    if (inA0 < inB0)
+      inA0 = inB0;
+
+    if (inA1 < inB1)
+      inA1 = inB1;
+
+    *((int8_t *)pIn) = bitins(inA0, n_mask, inA1, mask, off);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_avg_and_replace_u4(uint8_t * base,
+                                  uint8_t * target,
+                                  uint16_t length)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+
+  int cnt = length >> 2;
+
+  while (cnt > 0u)
+  {
+    *((uint32_t *)pIn) = avg8(*((uint32_t *)pIn), *((uint32_t *)pCom));
+
+    pIn+=4;
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+
+  while (left>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 4, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 4, 4);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 4, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 4, 4);
+
+    inA0 = ((inA0 + inB0) >> 1);
+    inA1 = ((inA1 + inB1) >> 1);
+
+    *((uint8_t*)pIn) = bitins(inA0, n_mask, inA1, mask, off);
+
+    pIn++;
+    pCom++;
+    length--;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_compare_and_replace_if_larger_u2(uint8_t * base,
+                                                uint8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask2 = 0x0c;
+  int8_t n_mask2 = ~ mask2;
+  int8_t mask4 = 0x30;
+  int8_t n_mask4 = ~ mask4;
+  int8_t mask6 = 0xc0;
+  int8_t n_mask6 = ~ mask6;
+  int8_t off2 = 2;
+  int8_t off4 = 4;
+  int8_t off6 = 6;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+  uint8_t *out;
+
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    *((uint32_t *)pIn) = max16(*((uint32_t *)pIn), *((uint32_t *)pCom));
+
+    pIn+=4;
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 2, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 2, 2);
+    uint8_t inA2 = (uint8_t) bitextu((unsigned int) *pIn, 2, 4);
+    uint8_t inA3 = (uint8_t) bitextu((unsigned int) *pIn, 2, 6);
+    v4u inA4 = pack((uint8_t) inA0, (uint8_t) inA1, (uint8_t) inA2, (uint8_t) inA3);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 2, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 2, 2);
+    uint8_t inB2 = (uint8_t) bitextu((unsigned int) *pCom, 2, 4);
+    uint8_t inB3 = (uint8_t) bitextu((unsigned int) *pCom, 2, 6);
+    v4u inB4 = pack((uint8_t) inB0, (uint8_t) inB1, (uint8_t) inB2, (uint8_t) inB3);
+
+    *((v4u*)out) = max4(inA4, inB4);
+
+    uint8_t inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_compare_and_replace_if_larger_i2(int8_t * base,
+                                                int8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask2 = 0x0c;
+  int8_t n_mask2 = ~ mask2;
+  int8_t mask4 = 0x30;
+  int8_t n_mask4 = ~ mask4;
+  int8_t mask6 = 0xc0;
+  int8_t n_mask6 = ~ mask6;
+  int8_t off2 = 2;
+  int8_t off4 = 4;
+  int8_t off6 = 6;
+
+  int8_t *pIn = base;
+  int8_t *pCom = target;
+  int8_t *out;
+
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    *((int32_t *)pIn) = maxs16(*((int32_t *)pIn), *((int32_t *)pCom));
+
+    pIn+=4;
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    int8_t inA0 = (int8_t) bitext((int) *pIn, 2, 0);
+    int8_t inA1 = (int8_t) bitext((int) *pIn, 2, 2);
+    int8_t inA2 = (int8_t) bitext((int) *pIn, 2, 4);
+    int8_t inA3 = (int8_t) bitext((int) *pIn, 2, 6);
+    v4s inA4 = pack((int8_t) inA0, (int8_t) inA1, (int8_t) inA2, (int8_t) inA3);
+    int8_t inB0 = (int8_t) bitext((int) *pCom, 2, 0);
+    int8_t inB1 = (int8_t) bitext((int) *pCom, 2, 2);
+    int8_t inB2 = (int8_t) bitext((int) *pCom, 2, 4);
+    int8_t inB3 = (int8_t) bitext((int) *pCom, 2, 6);
+    v4s inB4 = pack((int8_t) inB0, (int8_t) inB1, (int8_t) inB2, (int8_t) inB3);
+
+    *((v4s*)out) = maxs4(inA4, inB4);
+
+    int8_t inA = (int8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((int8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_tnn_compare_and_replace_if_larger_ternary(int8_t * base,
+                                                int8_t * target,
+                                                uint16_t length)
+{
+  uint8_t mask2 = 0x0c;
+  uint8_t n_mask2 = ~ mask2;
+  uint8_t mask4 = 0x30;
+  uint8_t n_mask4 = ~ mask4;
+  uint8_t mask6 = 0xc0;
+  uint8_t n_mask6 = ~ mask6;
+  uint8_t off2 = 2;
+  uint8_t off4 = 4;
+  uint8_t off6 = 6;
+
+  uint8_t *pIn = (uint8_t *) base;
+  uint8_t *pCom = (uint8_t *) target;
+  uint8_t *out;
+
+  int cnt = length >> 2;
+  uint32_t result;
+
+  while(cnt > 0u)
+  {
+    uint32_t in1 = *((uint32_t *)pIn);
+    uint32_t in2 = *((int32_t *)pCom);
+    result = maxs20(in1, in2);
+    *((uint32_t *)pIn) = result;
+
+    pIn+=4;
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  if (left>0u)
+  {
+    // do the vector max on the whole word - we won't use the leftover bytes
+    uint32_t in1 = *((uint32_t *)pIn);
+    uint32_t in2 = *((int32_t *)pCom);
+    result = maxs20(in1, in2);
+
+    // ...and copy back the relevant bytes of the result to pIn
+    for (int i=0; i<left; i++)
+      *((uint8_t*)(pIn + i)) = (uint8_t) (result >> (8*i));
+
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_avg_and_replace_u2(uint8_t * base,
+                                  uint8_t * target,
+                                  uint16_t length)
+{
+  int8_t mask2 = 0x0c;
+  int8_t n_mask2 = ~ mask2;
+  int8_t mask4 = 0x30;
+  int8_t n_mask4 = ~ mask4;
+  int8_t mask6 = 0xc0;
+  int8_t n_mask6 = ~ mask6;
+  int8_t off2 = 2;
+  int8_t off4 = 4;
+  int8_t off6 = 6;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+  uint8_t *out;
+
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    *((uint32_t *)pIn) = avg16(*((uint32_t *)pIn), *((uint32_t *)pCom));
+
+    pIn+=4;
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 2, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 2, 2);
+    uint8_t inA2 = (uint8_t) bitextu((unsigned int) *pIn, 2, 4);
+    uint8_t inA3 = (uint8_t) bitextu((unsigned int) *pIn, 2, 6);
+    v4u inA4 = pack((uint8_t) inA0, (uint8_t) inA1, (uint8_t) inA2, (uint8_t) inA3);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 2, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 2, 2);
+    uint8_t inB2 = (uint8_t) bitextu((unsigned int) *pCom, 2, 4);
+    uint8_t inB3 = (uint8_t) bitextu((unsigned int) *pCom, 2, 6);
+    v4u inB4 = pack((uint8_t) inB0, (uint8_t) inB1, (uint8_t) inB2, (uint8_t) inB3);
+
+    *((v4u*)out) = avg4(inA4, inB4);
+
+    uint8_t inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+#endif
diff --git a/rt_nn_tests/xpnn_conv/pulp_nn_utils_xpnn.h b/rt_nn_tests/xpnn_conv/pulp_nn_utils_xpnn.h
new file mode 100644
index 0000000..0c783ae
--- /dev/null
+++ b/rt_nn_tests/xpnn_conv/pulp_nn_utils_xpnn.h
@@ -0,0 +1,1937 @@
+/*
+ * pulp_nn_utils.h
+ * Nazareno   Bruschi  <nazareno.bruschi@unibo.it>
+ * Alessandro Nadalini <alessandro.nadalini3@unibo.it>
+ * Georg Rutishauser   <georgr@iis.ee.ethz.ch>
+ *
+ * Copyright (C) 2019-2020 ETH Zurich & University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef __PULPNN_UTILS__
+#define __PULPNN_UTILS__
+
+#include "pmsis.h"
+#ifdef GAP_SDK
+#include "pulp.h"
+#endif
+
+#define bitext(x,size,off)                                      __builtin_pulp_bextract(x,size,off)
+#define bitextu(x,size,off)                                     __builtin_pulp_bextractu(x,size,off)
+#define bitins(dst,not_mask_imm,src,mask_imm,off)               __builtin_pulp_binsert(dst,not_mask_imm,src,mask_imm,off)
+#define pack(x,y,z,t)                                           __builtin_pulp_pack4(x,y,z,t)
+#define max4(a,b)                                               __builtin_pulp_maxu4(a,b)
+#define maxs4(a, b)                                             __builtin_pulp_max4(a, b)
+#define max8(a, b)                                              __builtin_pulp_maxu8(a, b)
+#define maxs8(a, b)                                             __builtin_pulp_max8(a, b)
+#define max16(a, b)                                             __builtin_pulp_maxu16(a, b)
+#define maxs16(a, b)                                            __builtin_pulp_max16(a, b)
+#define max32(a,b)                                              __builtin_pulp_maxusi(a,b)
+#define maxs32(a,b)                                             __builtin_pulp_maxsi(a,b)
+#define min32(a,b)                                              __builtin_pulp_minusi(a,b)
+#define mins32(a,b)                                             __builtin_pulp_minsi(a,b)
+#define min4(a, b)                                              __builtin_pulp_minu4(a, b)
+#define mins4(a, b)                                             __builtin_pulp_min4(a, b)
+#define min8(a, b)                                              __builtin_pulp_minu8(a, b)
+#define mins8(a, b)                                             __builtin_pulp_min8(a, b)
+#define min16(a, b)                                             __builtin_pulp_minu16(a, b)
+#define mins16(a, b)                                            __builtin_pulp_min16(a, b)
+#define avg4(a,b)                                               __builtin_pulp_avgu4(a,b)
+#define avg8(a,b)                                               __builtin_pulp_avgu8(a,b)
+#define avg16(a,b)                                              __builtin_pulp_avgu16(a,b)
+#define log2(x)                                                 __builtin_pulp_fl1(x)
+#define min(a,b)                                                ((a)<(b)?(a):(b))
+#define SumDotp4(a, b, c)                                       __builtin_pulp_sdotusp4(a, b, c)
+#define SumDotp8(a, b, c)                                       __builtin_pulp_sdotusp8(a, b, c)
+#define SumDotp16(a, b, c)                                      __builtin_pulp_sdotusp16(a, b, c)
+#define SumDotps4(a, b, c)                                      __builtin_pulp_sdotsp4(a, b, c)
+#define SumDotps8(a, b, c)                                      __builtin_pulp_sdotsp8(a, b, c)
+#define SumDotps16(a, b, c)                                     __builtin_pulp_sdotsp16(a, b, c)
+#define clip4(x)                                                __builtin_pulp_clipu_r(x, 15)
+#define clip2(x)                                                __builtin_pulp_clipu_r(x, 3)
+#define clip8(x)                                                __builtin_pulp_clipu_r(x, 255)
+
+#define clips4(x)                                               __builtin_pulp_clip_r(x, 7)
+#define clips2(x)                                               __builtin_pulp_clip_r(x, 1)
+#define clips8(x)                                               __builtin_pulp_clip_r(x, 127)
+#define MacLoadInit(a_update, b_update, a_reg, b_reg, ptr)      __builtin_pulp_mlinitspr_v3(a_update, b_update, a_reg, b_reg, ptr)
+#define MacLoadUpdate(ptr)                                      __builtin_pulp_mlupdatespr_v3(ptr)
+#define MacLoadAssign(ptr)                                      __builtin_pulp_mlassignspr_v3(ptr)
+#define MacLoad4(a_update, b_update, a_reg, b_reg, ptr, sum)    __builtin_pulp_mlsdotsup4_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+#define MacLoad8(a_update, b_update, a_reg, b_reg, ptr, sum)    __builtin_pulp_mlsdotsup8_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+#define MacLoad16(a_update, b_update, a_reg, b_reg, ptr, sum)   __builtin_pulp_mlsdotsup16_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+#define MacLoads4(a_update, b_update, a_reg, b_reg, ptr, sum)   __builtin_pulp_mlsdotsp4_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+#define MacLoads8(a_update, b_update, a_reg, b_reg, ptr, sum)   __builtin_pulp_mlsdotsp8_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+#define MacLoads16(a_update, b_update, a_reg, b_reg, ptr, sum)  __builtin_pulp_mlsdotsp16_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+
+#define PACK_INT8_SIZE(x)                                    (x)
+#define PACK_INT4_SIZE(x)                                    ((x) >> 1)
+#define PACK_INT2_SIZE(x)                                    ((x) >> 2)
+
+#define MemoryFence()                                        asm volatile("":::"memory")
+
+#define LEGACY_MODE(x)                                       asm volatile ("csrwi 0x010," x)
+#define IVEC_FMT(x)                                          asm volatile ("csrwi 0x00D," x)
+#define MIXED_SKIP(x)                                        asm volatile ("csrwi 0x00F," x)
+#define A_ADDRESS(x)                                         asm volatile ("csrw 0x100, %0" :: "r" (x))
+#define W_ADDRESS(x)                                         asm volatile ("csrw 0x101, %0" :: "r" (x))
+#define A_STRIDE(x)                                          asm volatile ("csrw 0x102, %0":: "r" (x))
+#define W_STRIDE(x)                                          asm volatile ("csrw 0x103, %0":: "r" (x))
+#define A_ROLLBACK(x)                                        asm volatile ("csrw 0x104, %0":: "r" (x))
+#define W_ROLLBACK(x)                                        asm volatile ("csrw 0x105, %0":: "r" (x))
+#define A_SKIP(x)                                            asm volatile ("csrwi 0x106," x)
+#define W_SKIP(x)                                            asm volatile ("csrwi 0x107," x)
+
+static uint8_t __attribute__((noinline)) pulp_nn_quant_u2 (
+  int32_t phi,
+  int16_t m,
+  int8_t d
+  ) {
+  int32_t x = (m * phi) >> d;
+  uint8_t res = clip2(x);
+  return res;
+}
+static uint8_t __attribute__((noinline)) pulp_nn_bn_quant_u2 (
+  int32_t phi,
+  int32_t k,
+  int32_t lambda,
+  int8_t d
+  ) {
+  int32_t integer_image_phi = (k * phi) + lambda;
+  int32_t x = (integer_image_phi) >> d;
+  uint8_t res = clip2(x);
+  return res;
+}
+static uint8_t __attribute__((noinline)) pulp_nn_add_quant_u2 (
+  uint8_t pix1,
+  uint8_t pix2,
+  int16_t m1,
+  int16_t m2,
+  int8_t d
+  ) {
+  uint32_t integer_image = pix1*m1 + pix2*m2;
+  uint32_t x = (integer_image) >> d;
+  uint8_t res = clip2(x);
+  return res;
+}
+
+static int8_t __attribute__((noinline)) pulp_nn_quant_i2 (
+  int32_t phi,
+  int16_t m,
+  int8_t d
+  ) {
+  int32_t x = (m * phi) >> d;
+  int8_t res = clips2(x);
+  return res;
+}
+static int8_t __attribute__((noinline)) pulp_nn_bn_quant_i2 (
+  int32_t phi,
+  int32_t k,
+  int32_t lambda,
+  int8_t d
+  ) {
+  int32_t integer_image_phi = (k * phi) + lambda;
+  int32_t x = (integer_image_phi) >> d;
+  int8_t res = clips2(x);
+  return res;
+}
+static int8_t __attribute__((noinline)) pulp_nn_add_quant_i2 (
+  int8_t pix1,
+  int8_t pix2,
+  int16_t m1,
+  int16_t m2,
+  int8_t d
+  ) {
+  int32_t integer_image = pix1*m1 + pix2*m2;
+  int32_t x = (integer_image) >> d;
+  int8_t res = clips2(x);
+  return res;
+}
+
+static uint8_t __attribute__((noinline)) pulp_nn_quant_u4 (
+  int32_t phi,
+  int16_t m,
+  int8_t d
+  ) {
+  int32_t x = (m * phi) >> d;
+  uint8_t res = clip4(x);
+  return res;
+}
+static uint8_t __attribute__((noinline)) pulp_nn_bn_quant_u4 (
+  int32_t phi,
+  int32_t k,
+  int32_t lambda,
+  int8_t d
+  ) {
+  int32_t integer_image_phi = (k * phi) + lambda;
+  int32_t x = (integer_image_phi) >> d;
+  uint8_t res = clip4(x);
+  return res;
+}
+static uint8_t __attribute__((noinline)) pulp_nn_add_quant_u4 (
+  uint8_t pix1,
+  uint8_t pix2,
+  int16_t m1,
+  int16_t m2,
+  int8_t d
+  ) {
+  uint32_t integer_image = pix1*m1 + pix2*m2;
+  uint32_t x = (integer_image) >> d;
+  uint8_t res = clip4(x);
+  return res;
+}
+
+static int8_t __attribute__((noinline)) pulp_nn_quant_i4 (
+  int32_t phi,
+  int16_t m,
+  int8_t d
+  ) {
+  int32_t x = (m * phi) >> d;
+  int8_t res = clips4(x);
+  return res;
+}
+static int8_t __attribute__((noinline)) pulp_nn_bn_quant_i4 (
+  int32_t phi,
+  int32_t k,
+  int32_t lambda,
+  int8_t d
+  ) {
+  int32_t integer_image_phi = (k * phi) + lambda;
+  int32_t x = (integer_image_phi) >> d;
+  int8_t res = clips4(x);
+  return res;
+}
+static int8_t __attribute__((noinline)) pulp_nn_add_quant_i4 (
+  int8_t pix1,
+  int8_t pix2,
+  int16_t m1,
+  int16_t m2,
+  int8_t d
+  ) {
+  int32_t integer_image = pix1*m1 + pix2*m2;
+  int32_t x = (integer_image) >> d;
+  int8_t res = clips4(x);
+  return res;
+}
+
+static uint8_t __attribute__((noinline)) pulp_nn_quant_u8 (
+  int32_t phi,
+  int16_t m,
+  int8_t d
+  ) {
+  int32_t x = (m * phi) >> d;
+  uint8_t res = clip8(x);
+  return res;
+}
+static uint8_t __attribute__((noinline)) pulp_nn_bn_quant_u8 (
+  int32_t phi,
+  int32_t k,
+  int32_t lambda,
+  int8_t d
+  ) {
+  int32_t integer_image_phi = (k * phi) + lambda;
+  int32_t x = (integer_image_phi) >> d;
+  uint8_t res = clip8(x);
+  return res;
+}
+static uint8_t __attribute__((noinline)) pulp_nn_add_quant_u8 (
+  uint8_t pix1,
+  uint8_t pix2,
+  int16_t m1,
+  int16_t m2,
+  int8_t d
+  ) {
+  uint32_t integer_image = pix1*m1 + pix2*m2;
+  uint32_t x = (integer_image) >> d;
+  uint8_t res = clip8(x);
+  return res;
+}
+
+static int8_t __attribute__((noinline)) pulp_nn_quant_i8 (
+  int32_t phi,
+  int16_t m,
+  int8_t d
+  ) {
+  int32_t x = (m * phi) >> d;
+  int8_t res = clips8(x);
+  return res;
+}
+static int8_t __attribute__((noinline)) pulp_nn_bn_quant_i8 (
+  int32_t phi,
+  int32_t k,
+  int32_t lambda,
+  int8_t d
+  ) {
+  int32_t integer_image_phi = (k * phi) + lambda;
+  int32_t x = (integer_image_phi) >> d;
+  int8_t res = clips8(x);
+  return res;
+}
+static int8_t __attribute__((noinline)) pulp_nn_add_quant_i8 (
+  int8_t pix1,
+  int8_t pix2,
+  int16_t m1,
+  int16_t m2,
+  int8_t d
+  ) {
+  int32_t integer_image = pix1*m1 + pix2*m2;
+  int32_t x = (integer_image) >> d;
+  int8_t res = clips8(x);
+  return res;
+}
+
+
+static uint8_t __attribute__((noinline)) pulp_nn_u4_quant(int input, int16_t * pThr)
+{
+  if(input <= pThr[7] )
+  {
+    if( input <= pThr[3])
+    {
+      if( input <= pThr[1])
+      {
+        if( input <= pThr[0])
+          return 0;
+        else
+          return 1;
+      }
+      else
+      {
+        if( input <= pThr[2])
+          return 2;
+        else
+          return 3;
+      }
+    }
+    else
+    {
+      if( input <= pThr[5])
+      {
+        if( input <= pThr[4])
+          return 4;
+        else
+          return 5;
+      }
+      else
+      {
+        if( input <= pThr[6])
+          return 6;
+        else
+          return 7;
+      }
+    }
+  }
+  else
+  {
+    if( input <= pThr[11])
+    {
+      if( input <= pThr[9])
+      {
+        if( input <= pThr[8])
+          return 8;
+        else
+          return 9;
+      }
+      else
+      {
+        if( input <= pThr[10])
+          return 10;
+        else
+          return 11;
+      }
+    }
+    else
+    {
+      if( input <= pThr[13])
+      {
+        if( input <= pThr[12])
+          return 12;
+        else
+          return 13;
+      }
+      else
+      {
+        if( input <= pThr[14])
+          return 14;
+        else
+          return 15;
+      }
+    }
+  }
+}
+
+static uint8_t __attribute__((noinline)) pulp_nn_u2_quant(int input, int16_t * pThr)
+{
+  if( input <= pThr[1])
+  {
+    if( input <= pThr[0])
+        {
+      return 0;
+    }
+        else
+        {
+      return 1;
+    }
+  }
+    else
+    {
+    if( input <= pThr[2])
+        {
+      return 2;
+    }
+        else
+        {
+      return 3;
+    }
+  }
+}
+
+/*
+ * Common
+ */
+
+
+static v4s __attribute__((noinline)) pulp_nn_i4_to_i8_r( int8_t *pSrc)
+{
+  v4s Src = *((v4s*) pSrc);
+  int8_t bext1, bext2, bext3, bext4;
+
+  bext1 = (int8_t) bitext((int) Src, 4, 0);
+  bext2 = (int8_t) bitext((int) Src, 4, 4);
+  bext3 = (int8_t) bitext((int) Src, 4, 8);
+  bext4 = (int8_t) bitext((int) Src, 4, 12);
+  v4s res = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+
+  return res;
+}
+
+static v4u __attribute__((noinline)) pulp_nn_u4_to_u8_r(uint8_t *pSrc)
+{
+  v4u Src = *((v4u*) pSrc);
+  uint8_t bext1, bext2, bext3, bext4;
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 4, 0);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 4, 4);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 4, 8);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 4, 12);
+  v4u res = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+
+  return res;
+}
+
+static v4s __attribute__((noinline)) pulp_nn_i2_to_i8_r( int8_t *pSrc)
+{
+  v4s Src = *((v4s*) pSrc);
+  int8_t bext1, bext2, bext3, bext4;
+
+  bext1 = (int8_t) bitext((int) Src, 2, 0);
+  bext2 = (int8_t) bitext((int) Src, 2, 2);
+  bext3 = (int8_t) bitext((int) Src, 2, 4);
+  bext4 = (int8_t) bitext((int) Src, 2, 6);
+  v4s res = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+
+  return res;
+}
+
+static v4u __attribute__((noinline)) pulp_nn_u2_to_u8_r(uint8_t *pSrc)
+{
+  v4u Src = *((v4u*) pSrc);
+  uint8_t bext1, bext2, bext3, bext4;
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 0);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 2);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 4);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 6);
+  v4u res = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+
+  return res;
+}
+
+static v4s __attribute__((noinline)) pulp_nn_i2_to_i4_r(int8_t *pSrc)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  v4s Src = *((v4s*) pSrc);
+  int8_t bext1, bext2, bext3, bext4;
+  int8_t out1, out2, out3, out4;
+
+  bext1 = (int8_t) bitextu((int) Src, 2, 0);
+  bext2 = (int8_t) bitextu((int) Src, 2, 2);
+  bext3 = (int8_t) bitextu((int) Src, 2, 4);
+  bext4 = (int8_t) bitextu((int) Src, 2, 6);
+
+  out1 = bitins(bext1, n_mask, bext2, mask, off);
+  out2 = bitins(bext3, n_mask, bext4, mask, off);
+
+  bext1 = (int8_t) bitextu((int) Src, 2, 8);
+  bext2 = (int8_t) bitextu((int) Src, 2, 10);
+  bext3 = (int8_t) bitextu((int) Src, 2, 12);
+  bext4 = (int8_t) bitextu((int) Src, 2, 14);
+
+  out3 = bitins(bext1, n_mask, bext2, mask, off);
+  out4 = bitins(bext3, n_mask, bext4, mask, off);
+
+  v4s res = pack((int8_t) out1, (int8_t) out2, (int8_t) out3, (int8_t) out4);
+
+  return res;
+}
+
+static v4u __attribute__((noinline)) pulp_nn_u2_to_u4_r(uint8_t *pSrc)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  v4u Src = *((v4u*) pSrc);
+  uint8_t bext1, bext2, bext3, bext4;
+  uint8_t out1, out2, out3, out4;
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 0);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 2);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 4);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 6);
+
+  out1 = bitins(bext1, n_mask, bext2, mask, off);
+  out2 = bitins(bext3, n_mask, bext4, mask, off);
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 8);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 10);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 12);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 14);
+
+  out3 = bitins(bext1, n_mask, bext2, mask, off);
+  out4 = bitins(bext3, n_mask, bext4, mask, off);
+
+  v4u res = pack((uint8_t) out1, (uint8_t) out2, (uint8_t) out3, (uint8_t) out4);
+
+  return res;
+}
+
+static int8_t *__attribute__((always_inline)) pulp_nn_i4_to_i8( int8_t *pSrc, int8_t *pDst)
+{
+  v4s Src = *((v4s*) pSrc);
+  int8_t bext1, bext2, bext3, bext4;
+
+  pSrc+=4;
+
+  bext1 = (int8_t) bitext((int) Src, 4, 0);
+  bext2 = (int8_t) bitext((int) Src, 4, 4);
+  bext3 = (int8_t) bitext((int) Src, 4, 8);
+  bext4 = (int8_t) bitext((int) Src, 4, 12);
+  *((v4s*)pDst) = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+   MemoryFence();
+  pDst+=4;
+  bext1 = (int8_t) bitext((int) Src, 4, 16);
+  bext2 = (int8_t) bitext((int) Src, 4, 20);
+  bext3 = (int8_t) bitext((int) Src, 4, 24);
+  bext4 = (int8_t) bitext((int) Src, 4, 28);
+  *((v4s*)pDst) = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+
+  return pSrc;
+}
+
+static uint8_t *__attribute__((always_inline)) pulp_nn_u4_to_u8(uint8_t *pSrc, uint8_t *pDst)
+{
+  v4u Src = *((v4u*) pSrc);
+  uint8_t bext1, bext2, bext3, bext4;
+
+  pSrc+=4;
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 4, 0);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 4, 4);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 4, 8);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 4, 12);
+  *((v4u*)pDst) = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 4, 16);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 4, 20);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 4, 24);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 4, 28);
+  *((v4u*)pDst) = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+
+  return pSrc;
+}
+
+static int8_t *__attribute__((always_inline)) pulp_nn_i2_to_i8( int8_t * pSrc, int8_t * pDst)
+{
+  v4s Src = *((v4s*) pSrc);
+  int8_t bext1, bext2, bext3, bext4;
+
+  pSrc+=4;
+
+  bext1 = (int8_t) bitext((int) Src, 2, 0);
+  bext2 = (int8_t) bitext((int) Src, 2, 2);
+  bext3 = (int8_t) bitext((int) Src, 2, 4);
+  bext4 = (int8_t) bitext((int) Src, 2, 6);
+  *((v4s*)pDst) = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (int8_t) bitext((int) Src, 2, 8);
+  bext2 = (int8_t) bitext((int) Src, 2, 10);
+  bext3 = (int8_t) bitext((int) Src, 2, 12);
+  bext4 = (int8_t) bitext((int) Src, 2, 14);
+  *((v4s*)pDst) = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (int8_t) bitext((int) Src, 2, 16);
+  bext2 = (int8_t) bitext((int) Src, 2, 18);
+  bext3 = (int8_t) bitext((int) Src, 2, 20);
+  bext4 = (int8_t) bitext((int) Src, 2, 22);
+  *((v4s*)pDst) = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (int8_t) bitext((int) Src, 2, 24);
+  bext2 = (int8_t) bitext((int) Src, 2, 26);
+  bext3 = (int8_t) bitext((int) Src, 2, 28);
+  bext4 = (int8_t) bitext((int) Src, 2, 30);
+  *((v4s*)pDst) = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+
+  return pSrc;
+}
+
+static uint8_t *__attribute__((always_inline)) pulp_nn_u2_to_u8(uint8_t * pSrc, uint8_t * pDst)
+{
+  v4u Src = *((v4u*) pSrc);
+  uint8_t bext1, bext2, bext3, bext4;
+
+  pSrc+=4;
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 0);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 2);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 4);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 6);
+  *((v4u*)pDst) = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 8);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 10);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 12);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 14);
+  *((v4u*)pDst) = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 16);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 18);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 20);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 22);
+  *((v4u*)pDst) = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 24);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 26);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 28);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 30);
+  *((v4u*)pDst) = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+
+  return pSrc;
+}
+
+static int8_t *__attribute__((always_inline)) pulp_nn_i2_to_i4( int8_t * pSrc, int8_t * pDst)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  v4s Src = *((v4s*) pSrc);
+  int8_t bext1, bext2, bext3, bext4;
+  int8_t out1, out2, out3, out4;
+
+  pSrc+=4;
+
+  bext1 = (int8_t) bitext((int) Src, 2, 0);
+  bext2 = (int8_t) bitext((int) Src, 2, 2);
+  bext3 = (int8_t) bitext((int) Src, 2, 4);
+  bext4 = (int8_t) bitext((int) Src, 2, 6);
+
+  out1 = bitins(bext1, n_mask, bext2, mask, off);
+  out2 = bitins(bext3, n_mask, bext4, mask, off);
+  MemoryFence();
+
+  bext1 = (int8_t) bitext((int) Src, 2, 8);
+  bext2 = (int8_t) bitext((int) Src, 2, 10);
+  bext3 = (int8_t) bitext((int) Src, 2, 12);
+  bext4 = (int8_t) bitext((int) Src, 2, 14);
+
+  out3 = bitins(bext1, n_mask, bext2, mask, off);
+  out4 = bitins(bext3, n_mask, bext4, mask, off);
+
+  *((v4s*)pDst) = pack((int8_t) out1, (int8_t) out2, (int8_t) out3, (int8_t) out4);
+  MemoryFence();
+
+  pDst+=4;
+  bext1 = (int8_t) bitext((int) Src, 2, 16);
+  bext2 = (int8_t) bitext((int) Src, 2, 18);
+  bext3 = (int8_t) bitext((int) Src, 2, 20);
+  bext4 = (int8_t) bitext((int) Src, 2, 22);
+
+  out1 = bitins(bext1, n_mask, bext2, mask, off);
+  out2 = bitins(bext3, n_mask, bext4, mask, off);
+  MemoryFence();
+
+  bext1 = (int8_t) bitext((int) Src, 2, 24);
+  bext2 = (int8_t) bitext((int) Src, 2, 26);
+  bext3 = (int8_t) bitext((int) Src, 2, 28);
+  bext4 = (int8_t) bitext((int) Src, 2, 30);
+
+  out3 = bitins(bext1, n_mask, bext2, mask, off);
+  out4 = bitins(bext3, n_mask, bext4, mask, off);
+
+  *((v4s*)pDst) = pack((int8_t) out1, (int8_t) out2, (int8_t) out3, (int8_t) out4);
+
+  return pSrc;
+}
+
+static uint8_t *__attribute__((always_inline)) pulp_nn_u2_to_u4( uint8_t * pSrc, uint8_t * pDst)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  v4u Src = *((v4u*) pSrc);
+  uint8_t bext1, bext2, bext3, bext4;
+  uint8_t out1, out2, out3, out4;
+
+  pSrc+=4;
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 0);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 2);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 4);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 6);
+
+  out1 = bitins(bext1, n_mask, bext2, mask, off);
+  out2 = bitins(bext3, n_mask, bext4, mask, off);
+  MemoryFence();
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 8);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 10);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 12);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 14);
+
+  out3 = bitins(bext1, n_mask, bext2, mask, off);
+  out4 = bitins(bext3, n_mask, bext4, mask, off);
+
+  *((v4u*)pDst) = pack((uint8_t) out1, (uint8_t) out2, (uint8_t) out3, (uint8_t) out4);
+  MemoryFence();
+
+  pDst+=4;
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 16);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 18);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 20);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 22);
+
+  out1 = bitins(bext1, n_mask, bext2, mask, off);
+  out2 = bitins(bext3, n_mask, bext4, mask, off);
+  MemoryFence();
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 24);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 26);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 28);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 30);
+
+  out3 = bitins(bext1, n_mask, bext2, mask, off);
+  out4 = bitins(bext3, n_mask, bext4, mask, off);
+  
+  *((v4u*)pDst) = pack((uint8_t) out1, (uint8_t) out2, (uint8_t) out3, (uint8_t) out4);
+
+  return pSrc;
+}
+
+/*
+ * XpulpV2
+ */
+
+static void __attribute__((noinline)) pulp_zero_mem(uint8_t * pBuffer, unsigned int size)
+{
+  int lfover = size &0x3;
+  for (int i=0; i<(size>>2); i++)
+  {
+    *((v4u *)pBuffer) = (v4u){0,0,0,0};
+    MemoryFence();
+    pBuffer+=4;
+  }
+  while(lfover)
+  {
+    *pBuffer++=0;
+    lfover--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_im2col_u2_to_u8(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_u2_to_u8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 16;
+  }
+  while (lfover)
+  {
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 0);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 2);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 4);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 6);
+    pOutput++;
+    pInput++;
+    lfover -= 4;
+  }
+}
+static void __attribute__((noinline)) pulp_nn_im2col_i2_to_i8(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_i2_to_i8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 16;
+  }
+  while (lfover)
+  {
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 0);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 2);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 4);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 6);
+    pOutput++;
+    pInput++;
+    lfover -= 4;
+  }
+}
+static void __attribute__((noinline)) pulp_nn_im2col_u4_to_u8(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 3u;
+  int lfover = blockSize & 0x07;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_u4_to_u8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 8;
+  }
+  while (lfover)
+  {
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 4, 0);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 4, 4);
+    pOutput++;
+    pInput++;
+    lfover -= 2;
+  }
+}
+static void __attribute__((noinline)) pulp_nn_im2col_i4_to_i8(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 3u;
+  int lfover = blockSize & 0x07;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_i4_to_i8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 8;
+  }
+  while (lfover)
+  {
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 4, 0);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 4, 4);
+    pOutput++;
+    pInput++;
+    lfover -= 2;
+  }
+}
+static void __attribute__((noinline)) pulp_nn_im2col_u8_to_u8(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 2u;
+  int lfover = blockSize & 0x03;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4u*)pOutput) = *((v4u*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    *((uint8_t*)pOutput) = *((uint8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover--;
+  }
+}
+static void __attribute__((noinline)) pulp_nn_im2col_i8_to_i8(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 2u;
+  int lfover = blockSize & 0x03;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4s*)pOutput) = *((v4s*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    *((int8_t*)pOutput) = *((int8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover--;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_im2col_u2_to_u2(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4u*)pOutput) = *((v4u*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    uint8_t extr;
+    *((uint8_t*)pOutput) = *((uint8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover -= 4;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_i2_to_i2(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4s*)pOutput) = *((v4s*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    int8_t extr;
+    *((int8_t*)pOutput) = *((int8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover -= 4;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_u2_to_u4(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_u2_to_u4(pInput, pOutput);
+    MemoryFence();
+    pOutput += 8;
+  }
+  while (lfover)
+  {
+    uint8_t extr;
+    // TODO: this is potentially dangerous/wrong if the number of channels
+    // is not a multiple of 8!
+    *((v4u *)pOutput) = pulp_nn_u2_to_u4_r(pInput);
+    pInput += 2;
+    pOutput += 4;
+    lfover -= 8;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_i2_to_i4(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_i2_to_i4(pInput, pOutput);
+    MemoryFence();
+    pOutput += 8;
+  }
+  while (lfover)
+  {
+    int8_t extr;
+    // TODO: this is potentially dangerous/wrong if the number of channels
+    // is not a multiple of 8!
+    *((v4s *)pOutput) = pulp_nn_i2_to_i4_r(pInput);
+    pInput += 2;
+    pOutput += 4;
+    lfover -= 8;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_u2_to_u8(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_u2_to_u8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 16;
+  }
+  while (lfover)
+  {
+    uint8_t extr;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 0);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 2);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 4);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 6);
+    pOutput++;
+    pInput++;
+    lfover -= 4;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_i2_to_i8(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_i2_to_i8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 16;
+  }
+  while (lfover)
+  {
+    int8_t extr;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 0);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 2);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 4);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 6);
+    pOutput++;
+    pInput++;
+    lfover -= 4;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_u4_to_u4(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 3u;
+  int lfover = blockSize & 0x07;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4u*)pOutput) = *((v4u*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    uint8_t extr;
+    *((uint8_t*)pOutput) = *((uint8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover -= 2;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_i4_to_i4(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 3u;
+  int lfover = blockSize & 0x07;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4s*)pOutput) = *((v4s*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    int8_t extr;
+    *((int8_t*)pOutput) = *((int8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover -= 2;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_u4_to_u8(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 3u;
+  int lfover = blockSize & 0x07;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_u4_to_u8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 8;
+  }
+  while (lfover)
+  {
+    uint8_t extr;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 4, 0);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 4, 4);
+    pOutput++;
+    pInput++;
+    lfover -= 2;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_i4_to_i8(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 3u;
+  int lfover = blockSize & 0x07;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_i4_to_i8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 8;
+  }
+  while (lfover)
+  {
+    int8_t extr;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 4, 0);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 4, 4);
+    pOutput++;
+    pInput++;
+    lfover -= 2;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_u8_to_u8(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 2u;
+  int lfover = blockSize & 0x03;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4u*)pOutput) = *((v4u*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    uint8_t extr;
+    *((uint8_t*)pOutput) = *((uint8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover -= 1;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_i8_to_i8(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 2u;
+  int lfover = blockSize & 0x03;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4s*)pOutput) = *((v4s*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    int8_t extr;
+    *((int8_t*)pOutput) = *((int8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover -= 1;
+  }
+}
+
+
+
+static void __attribute__((noinline)) pulp_nn_compare_and_replace_if_larger_u8(uint8_t * base,
+                                                uint8_t * target,
+                                                uint16_t length)
+{
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+  v4u inp;
+  v4u com;
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    inp = *((v4u*)pIn);
+    com = *((v4u*)pCom);
+
+    *((v4u*)pIn) = max4(inp, com);
+
+    pCom+=4;
+    pIn+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    if(*pIn<*pCom)
+      *pIn=*pCom;
+    
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_compare_and_replace_if_larger_i8(
+    int8_t * base, int8_t * target, uint16_t length) {
+  int8_t *pIn = base;
+  int8_t *pCom = target;
+  v4s inp;
+  v4s com;
+  int cnt = length >> 2;
+
+  while (cnt > 0u) {
+    inp = *((v4s *)pIn);
+    com = *((v4s *)pCom);
+
+    *((v4s *)pIn) = maxs4(inp, com);
+
+    pCom += 4;
+    pIn += 4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left > 0u) {
+    if (*pIn < *pCom)
+      *pIn = *pCom;
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_avg_and_replace_u8(uint8_t * base,
+                                  uint8_t * target,
+                                  uint16_t length)
+{
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+
+  while (length>0u)
+  {
+    *pIn = ((*pIn + *pCom) >> 1);
+
+    pIn++;
+    pCom++;
+    length--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_compare_and_replace_if_larger_u4(uint8_t * base,
+                                                uint8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+  v4u inp[2];
+  v4u com[2];
+  uint8_t *out;
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    pulp_nn_u4_to_u8(pIn, (uint8_t *)inp);
+    pulp_nn_u4_to_u8(pCom, (uint8_t *)com);
+
+    *((v4u *)out) = max4(inp[0], com[0]);
+
+    *((uint8_t*)pIn) = bitins(*out, n_mask, *(out + 1), mask, off);
+    pIn++;
+    *((uint8_t*)pIn) = bitins(*(out + 2), n_mask, *(out + 3), mask, off);
+    pIn++;
+
+    *((v4u *)out) = max4(inp[1], com[1]);
+
+    *((uint8_t*)pIn) = bitins(*out, n_mask, *(out + 1), mask, off);
+    pIn++;
+    *((uint8_t*)pIn) = bitins(*(out + 2), n_mask, *(out + 3), mask, off);
+    pIn++;
+
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 4, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 4, 4);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 4, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 4, 4);
+
+    if(inA0<inB0)
+      inA0=inB0;
+
+    if(inA1<inB1)
+      inA1=inB1;
+
+    *((uint8_t*)pIn) = bitins(inA0, n_mask, inA1, mask, off);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_compare_and_replace_if_larger_i4(int8_t * base,
+                                                int8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  int8_t *pIn = base;
+  int8_t *pCom = target;
+  v4s inp[2];
+  v4s com[2];
+  int8_t *out;
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    pulp_nn_i4_to_i8(pIn, (int8_t *)inp);
+    pulp_nn_i4_to_i8(pCom, (int8_t *)com);
+
+    *((v4s *)out) = maxs4(inp[0], com[0]);
+
+    *((int8_t*)pIn) = bitins(*out, n_mask, *(out + 1), mask, off);
+    pIn++;
+    *((int8_t*)pIn) = bitins(*(out + 2), n_mask, *(out + 3), mask, off);
+    pIn++;
+
+    *((v4s *)out) = maxs4(inp[1], com[1]);
+
+    *((int8_t*)pIn) = bitins(*out, n_mask, *(out + 1), mask, off);
+    pIn++;
+    *((int8_t*)pIn) = bitins(*(out + 2), n_mask, *(out + 3), mask, off);
+    pIn++;
+
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    int8_t inA0 = (int8_t) bitext((int) *pIn, 4, 0);
+    int8_t inA1 = (int8_t) bitext((int) *pIn, 4, 4);
+    int8_t inB0 = (int8_t) bitext((int) *pCom, 4, 0);
+    int8_t inB1 = (int8_t) bitext((int) *pCom, 4, 4);
+
+    if(inA0<inB0)
+      inA0=inB0;
+
+    if(inA1<inB1)
+      inA1=inB1;
+
+    *((int8_t*)pIn) = bitins(inA0, n_mask, inA1, mask, off);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_avg_and_replace_u4(uint8_t * base,
+                                  uint8_t * target,
+                                  uint16_t length)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+
+  while (length>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 4, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 4, 4);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 4, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 4, 4);
+
+    inA0 = ((inA0 + inB0) >> 1);
+    inA1 = ((inA1 + inB1) >> 1);
+
+    *((uint8_t*)pIn) = bitins(inA0, n_mask, inA1, mask, off);
+
+    pIn++;
+    pCom++;
+    length--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_compare_and_replace_if_larger_u2(uint8_t * base,
+                                                uint8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask2 = 0x0c;
+  int8_t n_mask2 = ~ mask2;
+  int8_t mask4 = 0x30;
+  int8_t n_mask4 = ~ mask4;
+  int8_t mask6 = 0xc0;
+  int8_t n_mask6 = ~ mask6;
+  int8_t off2 = 2;
+  int8_t off4 = 4;
+  int8_t off6 = 6;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+  v4u inp[4];
+  v4u com[4];
+  uint8_t *out;
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    pulp_nn_u2_to_u8(pIn, inp);
+    pulp_nn_u2_to_u8(pCom, com);
+
+    *((v4u*)out) = max4(inp[0], com[0]);
+
+    uint8_t inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    *((v4u*)out) = max4(inp[1], com[1]);
+
+    inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    *((v4u*)out) = max4(inp[2], com[2]);
+
+    inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    *((v4u*)out) = max4(inp[3], com[3]);
+
+    inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 2, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 2, 2);
+    uint8_t inA2 = (uint8_t) bitextu((unsigned int) *pIn, 2, 4);
+    uint8_t inA3 = (uint8_t) bitextu((unsigned int) *pIn, 2, 6);
+    v4u inA4 = pack((uint8_t) inA0, (uint8_t) inA1, (uint8_t) inA2, (uint8_t) inA3);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 2, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 2, 2);
+    uint8_t inB2 = (uint8_t) bitextu((unsigned int) *pCom, 2, 4);
+    uint8_t inB3 = (uint8_t) bitextu((unsigned int) *pCom, 2, 6);
+    v4u inB4 = pack((uint8_t) inB0, (uint8_t) inB1, (uint8_t) inB2, (uint8_t) inB3);
+
+    *((v4u*)out) = max4(inA4, inB4);
+
+    uint8_t inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+
+static void __attribute__((noinline)) pulp_nn_compare_and_replace_if_larger_i2(int8_t * base,
+                                                int8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask2 = 0x0c;
+  int8_t n_mask2 = ~ mask2;
+  int8_t mask4 = 0x30;
+  int8_t n_mask4 = ~ mask4;
+  int8_t mask6 = 0xc0;
+  int8_t n_mask6 = ~ mask6;
+  int8_t off2 = 2;
+  int8_t off4 = 4;
+  int8_t off6 = 6;
+
+  int8_t *pIn = base;
+  int8_t *pCom = target;
+  v4s inp[4];
+  v4s com[4];
+  int8_t *out;
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    pulp_nn_i2_to_i8(pIn, inp);
+    pulp_nn_i2_to_i8(pCom, com);
+
+    *((v4s*)out) = maxs4(inp[0], com[0]);
+
+    int8_t inA = (int8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((int8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    *((v4s*)out) = maxs4(inp[1], com[1]);
+
+    inA = (int8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((int8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    *((v4s*)out) = maxs4(inp[2], com[2]);
+
+    inA = (int8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((int8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    *((v4s*)out) = maxs4(inp[3], com[3]);
+
+    inA = (int8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((int8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    int8_t inA0 = (int8_t) bitext((unsigned int) *pIn, 2, 0);
+    int8_t inA1 = (int8_t) bitext((unsigned int) *pIn, 2, 2);
+    int8_t inA2 = (int8_t) bitext((unsigned int) *pIn, 2, 4);
+    int8_t inA3 = (int8_t) bitext((unsigned int) *pIn, 2, 6);
+    v4s inA4 = pack((int8_t) inA0, (int8_t) inA1, (int8_t) inA2, (int8_t) inA3);
+    int8_t inB0 = (int8_t) bitext((unsigned int) *pCom, 2, 0);
+    int8_t inB1 = (int8_t) bitext((unsigned int) *pCom, 2, 2);
+    int8_t inB2 = (int8_t) bitext((unsigned int) *pCom, 2, 4);
+    int8_t inB3 = (int8_t) bitext((unsigned int) *pCom, 2, 6);
+    v4s inB4 = pack((int8_t) inB0, (int8_t) inB1, (int8_t) inB2, (int8_t) inB3);
+
+    *((v4s*)out) = maxs4(inA4, inB4);
+
+    int8_t inA = (int8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((int8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_avg_and_replace_u2(uint8_t * base,
+                                  uint8_t * target,
+                                  uint16_t length)
+{
+  int8_t mask2 = 0x0c;
+  int8_t n_mask2 = ~ mask2;
+  int8_t mask4 = 0x30;
+  int8_t n_mask4 = ~ mask4;
+  int8_t mask6 = 0xc0;
+  int8_t n_mask6 = ~ mask6;
+  int8_t off2 = 2;
+  int8_t off4 = 4;
+  int8_t off6 = 6;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+
+  while (length>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 2, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 2, 2);
+    uint8_t inA2 = (uint8_t) bitextu((unsigned int) *pIn, 2, 4);
+    uint8_t inA3 = (uint8_t) bitextu((unsigned int) *pIn, 2, 6);
+
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 2, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 2, 2);
+    uint8_t inB2 = (uint8_t) bitextu((unsigned int) *pCom, 2, 4);
+    uint8_t inB3 = (uint8_t) bitextu((unsigned int) *pCom, 2, 6);
+
+    inA0 = ((inA0 + inB0) >> 1);
+    inA1 = ((inA1 + inB1) >> 1);
+    inA2 = ((inA2 + inB2) >> 1);
+    inA3 = ((inA3 + inB3) >> 1);
+
+    uint8_t inA = (uint8_t) bitins(inA0, n_mask2, inA1, mask2, off2);
+    inA = bitins(inA, n_mask4, inA2, mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, inA3, mask6, off6);
+
+    pIn++;
+    pCom++;
+    length--;
+  }
+}
+
+/*
+ * XpulpNN
+ */
+
+static void __attribute__((noinline)) xpulp_nn_zero_mem_u8(uint8_t * pBuffer, unsigned int size)
+{
+  int lfover = size &0x3;
+  for (int i=0; i<(size>>2); i++)
+  {
+    *((v4u *)pBuffer) = (v4u){0,0,0,0};
+    MemoryFence();
+    pBuffer+=4;
+  }
+  while(lfover)
+  {
+    *pBuffer++=0;
+    lfover--;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_zero_mem_u4(uint8_t * pBuffer, unsigned int size)
+{
+  int lfover = size &0x7;
+  for (int i=0; i<(size>>3); i++)
+  {
+    *((v4u *)pBuffer) = (v4u){0,0,0,0};
+    MemoryFence();
+    pBuffer+=4;
+  }
+  while(lfover)
+  {
+    *pBuffer++=0;
+    lfover-=2;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_zero_mem_u2(uint8_t * pBuffer, unsigned int size)
+{
+  int lfover = size &0xf;
+  for (int i=0; i<(size>>4); i++)
+  {
+    *((v4u *)pBuffer) = (v4u){0,0,0,0};
+    MemoryFence();
+    pBuffer+=4;
+  }
+  while(lfover)
+  {
+    *pBuffer++=0;
+    lfover-=4;
+  }
+}
+
+
+static void __attribute__((noinline)) xpulp_nn_compare_and_replace_if_larger_u4(uint8_t * base,
+                                                uint8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    *((uint32_t *)pIn) = max8(*((uint32_t *)pIn), *((uint32_t *)pCom));
+
+    pIn+=4;
+    pCom+=4;
+
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 4, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 4, 4);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 4, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 4, 4);
+
+    if(inA0<inB0)
+      inA0=inB0;
+
+    if(inA1<inB1)
+      inA1=inB1;
+
+    *((uint8_t*)pIn) = bitins(inA0, n_mask, inA1, mask, off);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_compare_and_replace_if_larger_i4(
+    int8_t * base, int8_t * target, uint16_t length) {
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~mask;
+  int8_t off = 0x04;
+
+  int8_t *pIn = base;
+  int8_t *pCom = target;
+
+  int cnt = length >> 2;
+
+  while (cnt > 0u) {
+    *((int32_t *)pIn) = maxs8(*((int32_t *)pIn), *((int32_t *)pCom));
+
+    pIn += 4;
+    pCom += 4;
+
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left > 0u) {
+    int8_t inA0 = (int8_t)bitext((int)*pIn, 4, 0);
+    int8_t inA1 = (int8_t)bitext((int)*pIn, 4, 4);
+    int8_t inB0 = (int8_t)bitext((int)*pCom, 4, 0);
+    int8_t inB1 = (int8_t)bitext((int)*pCom, 4, 4);
+
+    if (inA0 < inB0)
+      inA0 = inB0;
+
+    if (inA1 < inB1)
+      inA1 = inB1;
+
+    *((int8_t *)pIn) = bitins(inA0, n_mask, inA1, mask, off);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_avg_and_replace_u4(uint8_t * base,
+                                  uint8_t * target,
+                                  uint16_t length)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+
+  int cnt = length >> 2;
+
+  while (cnt > 0u)
+  {
+    *((uint32_t *)pIn) = avg8(*((uint32_t *)pIn), *((uint32_t *)pCom));
+
+    pIn+=4;
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+
+  while (left>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 4, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 4, 4);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 4, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 4, 4);
+
+    inA0 = ((inA0 + inB0) >> 1);
+    inA1 = ((inA1 + inB1) >> 1);
+
+    *((uint8_t*)pIn) = bitins(inA0, n_mask, inA1, mask, off);
+
+    pIn++;
+    pCom++;
+    length--;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_compare_and_replace_if_larger_u2(uint8_t * base,
+                                                uint8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask2 = 0x0c;
+  int8_t n_mask2 = ~ mask2;
+  int8_t mask4 = 0x30;
+  int8_t n_mask4 = ~ mask4;
+  int8_t mask6 = 0xc0;
+  int8_t n_mask6 = ~ mask6;
+  int8_t off2 = 2;
+  int8_t off4 = 4;
+  int8_t off6 = 6;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+  uint8_t *out;
+
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    *((uint32_t *)pIn) = max16(*((uint32_t *)pIn), *((uint32_t *)pCom));
+
+    pIn+=4;
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 2, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 2, 2);
+    uint8_t inA2 = (uint8_t) bitextu((unsigned int) *pIn, 2, 4);
+    uint8_t inA3 = (uint8_t) bitextu((unsigned int) *pIn, 2, 6);
+    v4u inA4 = pack((uint8_t) inA0, (uint8_t) inA1, (uint8_t) inA2, (uint8_t) inA3);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 2, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 2, 2);
+    uint8_t inB2 = (uint8_t) bitextu((unsigned int) *pCom, 2, 4);
+    uint8_t inB3 = (uint8_t) bitextu((unsigned int) *pCom, 2, 6);
+    v4u inB4 = pack((uint8_t) inB0, (uint8_t) inB1, (uint8_t) inB2, (uint8_t) inB3);
+
+    *((v4u*)out) = max4(inA4, inB4);
+
+    uint8_t inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_compare_and_replace_if_larger_i2(int8_t * base,
+                                                int8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask2 = 0x0c;
+  int8_t n_mask2 = ~ mask2;
+  int8_t mask4 = 0x30;
+  int8_t n_mask4 = ~ mask4;
+  int8_t mask6 = 0xc0;
+  int8_t n_mask6 = ~ mask6;
+  int8_t off2 = 2;
+  int8_t off4 = 4;
+  int8_t off6 = 6;
+
+  int8_t *pIn = base;
+  int8_t *pCom = target;
+  int8_t *out;
+
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    *((int32_t *)pIn) = maxs16(*((int32_t *)pIn), *((int32_t *)pCom));
+
+    pIn+=4;
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    int8_t inA0 = (int8_t) bitext((int) *pIn, 2, 0);
+    int8_t inA1 = (int8_t) bitext((int) *pIn, 2, 2);
+    int8_t inA2 = (int8_t) bitext((int) *pIn, 2, 4);
+    int8_t inA3 = (int8_t) bitext((int) *pIn, 2, 6);
+    v4s inA4 = pack((int8_t) inA0, (int8_t) inA1, (int8_t) inA2, (int8_t) inA3);
+    int8_t inB0 = (int8_t) bitext((int) *pCom, 2, 0);
+    int8_t inB1 = (int8_t) bitext((int) *pCom, 2, 2);
+    int8_t inB2 = (int8_t) bitext((int) *pCom, 2, 4);
+    int8_t inB3 = (int8_t) bitext((int) *pCom, 2, 6);
+    v4s inB4 = pack((int8_t) inB0, (int8_t) inB1, (int8_t) inB2, (int8_t) inB3);
+
+    *((v4s*)out) = maxs4(inA4, inB4);
+
+    int8_t inA = (int8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((int8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_avg_and_replace_u2(uint8_t * base,
+                                  uint8_t * target,
+                                  uint16_t length)
+{
+  int8_t mask2 = 0x0c;
+  int8_t n_mask2 = ~ mask2;
+  int8_t mask4 = 0x30;
+  int8_t n_mask4 = ~ mask4;
+  int8_t mask6 = 0xc0;
+  int8_t n_mask6 = ~ mask6;
+  int8_t off2 = 2;
+  int8_t off4 = 4;
+  int8_t off6 = 6;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+  uint8_t *out;
+
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    *((uint32_t *)pIn) = avg16(*((uint32_t *)pIn), *((uint32_t *)pCom));
+
+    pIn+=4;
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 2, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 2, 2);
+    uint8_t inA2 = (uint8_t) bitextu((unsigned int) *pIn, 2, 4);
+    uint8_t inA3 = (uint8_t) bitextu((unsigned int) *pIn, 2, 6);
+    v4u inA4 = pack((uint8_t) inA0, (uint8_t) inA1, (uint8_t) inA2, (uint8_t) inA3);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 2, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 2, 2);
+    uint8_t inB2 = (uint8_t) bitextu((unsigned int) *pCom, 2, 4);
+    uint8_t inB3 = (uint8_t) bitextu((unsigned int) *pCom, 2, 6);
+    v4u inB4 = pack((uint8_t) inB0, (uint8_t) inB1, (uint8_t) inB2, (uint8_t) inB3);
+
+    *((v4u*)out) = avg4(inA4, inB4);
+
+    uint8_t inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+
+#endif
diff --git a/rt_nn_tests/xpnn_conv/test.c b/rt_nn_tests/xpnn_conv/test.c
new file mode 100644
index 0000000..5ce4bde
--- /dev/null
+++ b/rt_nn_tests/xpnn_conv/test.c
@@ -0,0 +1,221 @@
+#include <stdio.h>
+#include <stdint.h>
+#include <stdlib.h>
+
+
+#include "pmsis.h"
+
+#include "data_statstest.h"
+//#include "pulp_nn_kernels.h"
+#include "pulp_nn_mix_kernels.h"
+
+
+#define start_cycle_counter()                                   asm volatile("csrw 0xCC0, 0x01;")
+#define stop_cycle_counter()                                    asm volatile("csrw 0xCC0, 0x00;")
+#define read_cycle_counter(x)                                   asm volatile("csrr %0, 0x780;" : "=r" (x))
+#define reset_cycle_counter()                                   asm volatile("csrw 0x780, 0x0;")
+
+uint8_t im2col[IM2COL_DIM] = {0};
+uint8_t outputs[OUTPUT_DIM] = {0};
+
+int32_t outputs_fp[OUTPUT_DIM_FP] = {0};
+
+#ifndef PROFILE
+int num_cycles;
+#else
+int im2col_cycles;
+int hotloop_prep_cycles;
+int hotloop_cycles;
+int threshold_cycles;
+int requant_cycles;
+int hotloop_leftover_cycles;
+int matmul4x2_leftover_cycles;
+#endif
+
+void call_krnl_0();
+void test_0();
+
+int main(int argc, char *argv[])
+{
+#if KRAKEN_PTEST == 1
+  kraken_padframe_aon_pad_gpioa_cfg_rxe_set(24, 0);
+  kraken_padframe_aon_pad_gpioa_cfg_trie_set(24, 0);
+  kraken_padframe_aon_pad_gpioa_cfg_chip2pad_set(24, 0);
+#endif
+  int n_mismatches = 0;
+  int mismatches_tot = 0;
+
+    
+
+  #ifndef PROFILE
+  num_cycles = 0;
+  #endif
+
+  if (get_core_id() == 0) {
+  printf("===> TEST 0: Running xpulp_nn_conv_u4_u8_i2...\n");
+  printf("  dims_in     = [4, 4]\n");
+  printf("  dims_kernel = [3, 3]\n");
+  printf("  ch_in/out   = [16, 32]\n");
+  //printf("  padding_y_top    = [%d]\n", padding_y_top);
+  //printf("  padding_y_bottom = [%d]\n", padding_y_bottom);
+  //printf("  padding_x_left   = [%d]\n", padding_x_left);
+  //printf("  padding_x_right  = [%d]\n", padding_x_right);
+  //printf("  stride_x         = [%d]\n", stride_x);
+  //printf("  stride_y         = [%d]\n", stride_y);
+  }
+  test_0();
+  #ifndef PROFILE
+  stop_cycle_counter();
+  read_cycle_counter(num_cycles);
+  if (get_core_id() == 0) {
+  printf("===> TEST 0: Finished running xpulp_nn_conv_u4_u8_i2\n");
+  printf("num_cycles = %d\n", num_cycles);
+  printf("MACs = 73728\n");
+  printf("MACs/cycle = %.4f\n", 73728/num_cycles);
+  }
+  #endif
+  if (get_core_id() == 0) {
+  printf("Checking for mismatches..\n");
+  n_mismatches = 0;
+
+  for(int i=0; i < 512; i++) {
+    if (outputs[i] != exp_outp_0[i]){
+        printf("***Mismatch in test 0 at iteration %d: Expected: %x, got: %x\n", i, exp_outp_0[i], outputs[i]);
+        n_mismatches++;
+    }
+  }
+  }
+  mismatches_tot += n_mismatches;
+
+  if (get_core_id() == 0) {
+  printf("Got %d mismatches in %d tests\n", mismatches_tot, 1);
+  }
+  return mismatches_tot;
+}
+
+
+void call_krnl_0(void) {
+  uint8_t * pInp;
+  uint8_t * pIm2ColBuffer;
+  int8_t * pBias = NULL;
+  uint8_t * pOut;
+  int8_t * pWeight;
+  uint32_t * pThr;
+  int32_t * pKappa, pLambda;
+  #ifdef PROFILE
+  int32_t im2col_cycles = 0;
+  int32_t hotloop_prep_cycles = 0;
+  int32_t hotloop_cycles = 0;
+  int32_t threshold_cycles = 0;
+  int32_t requant_cycles = 0;
+  int32_t hotloop_leftover_cycles = 0;
+  int32_t matmul4x2_leftover_cycles = 0;
+  #endif
+  pInp = inp_l1;
+  
+  pOut = outp_l1;
+  pIm2ColBuffer = im2col_l1;
+  pWeight = wt_l1;
+  pThr = threshs_l1;
+  pKappa = kappa_l1;
+  pLambda = lambda_l1;
+#if KRAKEN_PTEST == 1
+  if (pi_core_id() == 0) {
+    kraken_padframe_aon_pad_gpioa_cfg_chip2pad_set(24, 1);
+  } 
+#endif
+  xpulp_nn_conv_u4_u8_i2(
+    pInp,
+    pIm2ColBuffer,
+    pBias,
+    pOut,
+    pWeight,
+    pKappa,
+    pLambda,
+    1,
+    13,
+    4,
+    4,
+    16,
+    4,
+    4,
+    32,
+    3,
+    3,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+#ifndef PROFILE
+    1
+#else
+    1,
+    &im2col_cycles,
+    &hotloop_prep_cycles,
+    &hotloop_cycles,
+    &requant_cycles,
+    &hotloop_leftover_cycles,
+    &matmul4x2_leftover_cycles
+#endif
+    );
+
+#if KRAKEN_PTEST == 1
+  if (pi_core_id() == 0) {
+    kraken_padframe_aon_pad_gpioa_cfg_chip2pad_set(24, 0);
+  }
+#endif
+  #ifdef PROFILE
+  if (pi_core_id() == 0) {
+  printf("im2col_cycles = %d\n", im2col_cycles);
+  printf("hotloop_prep_cycles = %d\n", hotloop_prep_cycles);
+  printf("hotloop_cycles = %d\n", hotloop_cycles);
+  printf("requant_cycles = %d\n", requant_cycles);
+  printf("threshold_cycles = %d\n", threshold_cycles);
+  printf("hotloop_leftover_cycles = %d\n", hotloop_leftover_cycles);
+  printf("matmul4x2_leftover_cycles = %d\n", matmul4x2_leftover_cycles);
+  }
+  #endif
+}
+
+void test_0(void) {
+  // DMA transfer inputs from L2 to L1
+  if (pi_core_id() == 0) {
+     plp_dma_memcpy(pIn_0, inp_l1, 128, 1);
+     plp_dma_barrier();
+  }
+  pi_cl_team_barrier(0);
+  if (pi_core_id() == 0) {
+     plp_dma_memcpy(pLambda_0, lambda_l1, 32 * 4, 1); // 4 bytes per lambda item
+     plp_dma_barrier();
+  }
+  if (pi_core_id() == 0) {
+     plp_dma_memcpy(pKappa_0, kappa_l1, 32 * 4, 1); // 4 bytes per lambda item
+     plp_dma_barrier();
+  }
+  pi_cl_team_barrier(0);
+  // transfer weights
+  if (pi_core_id() == 0) {
+     plp_dma_memcpy(pWeight_0, wt_l1, 1152, 1);
+     plp_dma_barrier();
+  }
+  pi_cl_team_barrier(0);
+  call_krnl_0();
+  // get outputs back with DMA
+  if (pi_core_id() == 0) {
+     plp_dma_memcpy(outputs, outp_l1, 512, 0);
+     plp_dma_barrier();
+  }
+}
+
+
+
+
+
+
+
+
+
+
diff --git a/rt_nn_tests/xpnn_conv/xpulp_nn_conv_u4_u8_i2.c b/rt_nn_tests/xpnn_conv/xpulp_nn_conv_u4_u8_i2.c
new file mode 100644
index 0000000..f625bfc
--- /dev/null
+++ b/rt_nn_tests/xpnn_conv/xpulp_nn_conv_u4_u8_i2.c
@@ -0,0 +1,303 @@
+/*
+ * xpulp_nn_conv_u4_u8_i2.c
+ * Nazareno Bruschi <nazareno.bruschi@unibo.it>
+ *
+ * Copyright (C) 2019-2020 University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include "pmsis.h"
+#include "pulp_nn_utils.h"
+#include "pulp_nn_kernels.h"
+
+
+
+
+void __attribute__((noinline)) xpulp_nn_conv_u4_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mult,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batch_norm)
+{
+  uint16_t ch_in_r = PACK_INT4_SIZE(ch_in);
+  uint16_t ch_out_r = PACK_INT8_SIZE(ch_out);
+
+  int core_id = pi_core_id();
+  uint8_t * pIm2ColBase = pIm2ColBuffer + (2 * core_id * PACK_INT4_SIZE(ch_in) * dim_kernel_x * dim_kernel_y);
+  int i_out_y, i_out_x, i_ker_y, i_ker_x;
+  int Log2Core;
+
+  uint8_t extra_chunk = ((dim_out_y & (NUM_CORES-1)) != 0);
+  uint8_t extra_chunk_r;
+  uint16_t dim_out_x_r;
+  uint8_t section;
+  int core_id_r;
+
+  if(extra_chunk && dim_out_x > 1)
+  {
+    Log2Core = log2(NUM_CORES >> 1);
+    core_id_r = (core_id >> 1);
+    dim_out_x_r = (dim_out_x >> 1);
+    section = (core_id & 0x1);
+    extra_chunk_r = ((dim_out_y & ((NUM_CORES >> 1) - 1)) != 0);
+  }
+  else
+  {
+    Log2Core = log2(NUM_CORES);
+    core_id_r = core_id;
+    dim_out_x_r = dim_out_x;
+    section = 0;
+    extra_chunk_r = extra_chunk;
+    extra_chunk = 0;
+  }
+
+  uint8_t flag_dim_out_x_odd = dim_out_x & 0x01;
+
+  int chunk = (dim_out_y >> Log2Core) + extra_chunk_r;
+
+  int start_pixel = min((chunk * core_id_r), dim_out_y);
+  int stop_pixel = min(start_pixel + chunk, dim_out_y);
+
+  uint8_t *pIm2Col = pIm2ColBase;
+  uint8_t *pOutBuffer = pOut + (start_pixel * ch_out_r * dim_out_x) + (section * ch_out_r * dim_out_x_r);
+
+  for (i_out_y = start_pixel; i_out_y < stop_pixel; i_out_y++)
+  {
+    for(i_out_x=(section * dim_out_x_r); i_out_x<(dim_out_x_r + (section * (dim_out_x_r + flag_dim_out_x_odd))); i_out_x++)
+    {
+      if(i_out_y < padding_y_top)
+      {
+        for(i_ker_y=((i_out_y * stride_y) - padding_y_top); i_ker_y<((i_out_y * stride_y) - padding_y_top + dim_kernel_y); i_ker_y++)
+        {
+          for(i_ker_x=((i_out_x * stride_x) - padding_x_left); i_ker_x<((i_out_x * stride_x) - padding_x_left + dim_kernel_x); i_ker_x++)
+          {
+            if((i_ker_y < 0) || (i_ker_y >= dim_in_y) || (i_ker_x < 0) || (i_ker_x >= dim_in_x))
+            {
+              xpulp_nn_zero_mem_u4(pIm2Col, ch_in);
+            }
+            else
+            {
+              xpulp_nn_im2col_u4_to_u4((uint8_t*) (pIn + ((i_ker_y * dim_in_x + i_ker_x) * ch_in_r)), pIm2Col, ch_in);
+            }
+            pIm2Col+=PACK_INT4_SIZE(ch_in);
+          }
+        }
+      }
+      else if(i_out_y < dim_out_y - padding_y_bottom)
+      {
+        if(i_out_x < padding_x_left)
+        {
+          for(i_ker_y=((i_out_y * stride_y) - padding_y_top); i_ker_y<((i_out_y * stride_y) - padding_y_top + dim_kernel_y); i_ker_y++)
+          {
+            for(i_ker_x=((i_out_x * stride_x) - padding_x_left); i_ker_x<((i_out_x * stride_x) - padding_x_left + dim_kernel_x); i_ker_x++)
+            {
+              if((i_ker_x < 0) || (i_ker_x >= dim_in_x))
+              {
+                xpulp_nn_zero_mem_u4(pIm2Col, ch_in);
+              }
+              else
+              {
+                xpulp_nn_im2col_u4_to_u4((uint8_t*) (pIn + ((i_ker_y * dim_in_x + i_ker_x) * ch_in_r)), pIm2Col, ch_in);
+              }
+              pIm2Col+=PACK_INT4_SIZE(ch_in);
+            }
+          }
+        }
+        else if(i_out_x < (dim_out_x - padding_x_right))
+        {
+          for(i_ker_y=((i_out_y * stride_y) - padding_y_top); i_ker_y<((i_out_y * stride_y) - padding_y_top + dim_kernel_y); i_ker_y++)
+          {
+            xpulp_nn_im2col_u4_to_u4((uint8_t*) pIn + (i_ker_y * dim_in_x + i_out_x * stride_x - padding_x_left)*ch_in_r,pIm2Col,ch_in * dim_kernel_x);
+            pIm2Col+=PACK_INT4_SIZE(ch_in * dim_kernel_x);
+          }
+        }
+        else
+        {
+          for(i_ker_y=((i_out_y * stride_y) - padding_y_top); i_ker_y<((i_out_y * stride_y) - padding_y_top + dim_kernel_y); i_ker_y++)
+          {
+            for(i_ker_x = i_out_x * stride_x - padding_x_left; i_ker_x < i_out_x * stride_x - padding_x_left + dim_kernel_x; i_ker_x++)
+            {
+              if((i_ker_x < 0) || (i_ker_x >= dim_in_x))
+              {
+                xpulp_nn_zero_mem_u4(pIm2Col, ch_in);
+              }
+              else
+              {
+                xpulp_nn_im2col_u4_to_u4((uint8_t *)pIn + (i_ker_y*dim_in_x+i_ker_x)* ch_in_r, pIm2Col, ch_in);
+              }
+              pIm2Col+=PACK_INT4_SIZE(ch_in);
+            }
+          }
+        }
+      }
+      else
+      {
+        for(i_ker_y=((i_out_y * stride_y) - padding_y_top); i_ker_y<((i_out_y * stride_y) - padding_y_top + dim_kernel_y); i_ker_y++)
+        {
+          for(i_ker_x = i_out_x * stride_x - padding_x_left; i_ker_x < i_out_x * stride_x - padding_x_left + dim_kernel_x; i_ker_x++)
+          {
+            if(i_ker_y < 0 || (i_ker_y >= dim_in_y) || i_ker_x < 0 || i_ker_x >= dim_in_x)
+            {
+              xpulp_nn_zero_mem_u4(pIm2Col, ch_in);
+            }
+            else
+            {
+              xpulp_nn_im2col_u4_to_u4((uint8_t *) pIn + (i_ker_y * dim_in_x + i_ker_x) * ch_in_r, pIm2Col, ch_in);
+            }
+            pIm2Col+=PACK_INT4_SIZE(ch_in);
+          }
+        }
+      }
+      if(pIm2Col == (pIm2ColBase + ((PACK_INT4_SIZE(ch_in) * dim_kernel_x * dim_kernel_y) << 1)))
+      {
+        pOutBuffer = xpulp_nn_matmul_u4_u8_i2(
+          pIm2ColBase,
+          pBias,
+          pOutBuffer,
+          pOutBuffer + ch_out_r,
+          pWeight,
+          pKappa,
+          pLambda,
+          out_mult,
+          out_shift,
+          (ch_in * dim_kernel_x * dim_kernel_y),
+          ch_out,
+          flag_relu,
+          flag_batch_norm
+          );
+
+        pIm2Col = pIm2ColBase;
+      }
+    }
+
+    if(pIm2Col != pIm2ColBase)
+    {
+      const int8_t *pA = pWeight;
+      int i;
+      int32_t * k1 = pKappa;
+      int32_t * lambda1 = pLambda;
+
+      v4s inA[2];
+      uint8_t out[1];
+      uint16_t num_col_im2col = ch_in * dim_kernel_x * dim_kernel_y;
+      uint16_t num_col_im2col_w = PACK_INT2_SIZE(ch_in) * dim_kernel_x * dim_kernel_y;
+
+      for(i = 0; i < ch_out; i++)
+      {
+        int sum = 0;
+        if (pBias != NULL)
+        {
+          sum = *((int*) pBias);
+          pBias+= 4;
+        }
+
+        uint8_t *pB = pIm2ColBase;
+
+        int32_t *ptrA  = (int32_t *)pA;
+        uint32_t *ptrB = (uint32_t *)pB;
+
+        for(int j=0; j < (num_col_im2col >> 4); j++)
+        {
+          pA = pulp_nn_i2_to_i4(pA,inA);
+
+          ptrA = (int32_t *)inA;
+
+          sum = SumDotp8(*(uint32_t *)ptrB, *(int32_t *)ptrA, sum);
+
+          ptrA++;
+          ptrB++;
+
+          sum = SumDotp8(*(uint32_t *)ptrB, *(int32_t *)ptrA, sum);
+
+          ptrA++;
+          ptrB++;
+        }
+
+        int col_cnt_im2col = num_col_im2col & 0xf;
+
+        if(col_cnt_im2col)
+        {
+
+          uint16_t loop_cnt_im2col_a = (num_col_im2col >> 4) << 3;
+          pB+=loop_cnt_im2col_a;
+
+          do
+          {
+            int8_t inA1 = (int8_t) bitext((int) *pA, 2, 0);
+            uint8_t inB1 = (uint8_t) bitextu((uint32_t) *pB, 4, 0);
+            sum += inA1 * inB1;
+            inA1 = (int8_t) bitext((int) *pA, 2, 2);
+            inB1 = (uint8_t) bitextu((uint32_t) *pB, 4, 4);
+            sum += inA1 * inB1;
+            pB++;
+            inA1 = (int8_t) bitext((int) *pA, 2, 4);
+            inB1 = (uint8_t) bitextu((uint32_t) *pB, 4, 0);
+            sum += inA1 * inB1;
+            inA1 = (int8_t) bitext((int) *pA, 2, 6);
+            inB1 = (uint8_t) bitextu((uint32_t) *pB, 4, 4);
+            sum += inA1 * inB1;
+
+            pA++;
+            pB++;
+            col_cnt_im2col-=4;
+          } while(col_cnt_im2col);
+        }
+        if (flag_batch_norm && flag_relu)
+        {
+          *pOutBuffer = pulp_nn_bn_quant_u8(sum, *k1, *lambda1, out_shift);
+          k1++;
+          lambda1++;
+          pOutBuffer++;
+        }
+        else
+        {
+          if(flag_relu == 1)
+          {
+            *pOutBuffer = pulp_nn_quant_u8(sum, out_mult, out_shift);
+            pOutBuffer++;
+          }
+          else
+          {
+            *pOutBuffer = (uint8_t) clip8(sum >> out_shift);
+            pOutBuffer++;
+          }
+        }
+      }
+    }
+    pOutBuffer+=(extra_chunk * ((dim_out_x_r + ((1 - section) * flag_dim_out_x_odd)) * ch_out_r));
+    pIm2Col = pIm2ColBase;
+  }
+  pi_cl_team_barrier();
+}
diff --git a/rt_nn_tests/xpnn_conv/xpulp_nn_matmul_u4_u8_i2.c b/rt_nn_tests/xpnn_conv/xpulp_nn_matmul_u4_u8_i2.c
new file mode 100644
index 0000000..5713be9
--- /dev/null
+++ b/rt_nn_tests/xpnn_conv/xpulp_nn_matmul_u4_u8_i2.c
@@ -0,0 +1,490 @@
+/*
+ * xpulp_nn_matmul_u4_u8_i2.c
+ * Nazareno Bruschi <nazareno.bruschi@unibo.it>
+ *
+ * Copyright (C) 2019-2020 University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include "pmsis.h"
+#include "pulp_nn_utils.h"
+
+
+
+uint8_t * __attribute__((noinline)) xpulp_nn_matmul_u4_u8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mult,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batch_norm)
+{
+  int32_t vecA[2];
+  int32_t vecA2[2];
+  int32_t vecA3[2];
+  int32_t vecA4[2];
+
+  uint16_t ch_out_r = PACK_INT8_SIZE(ch_out);
+
+  uint16_t num_col_im2col_w = PACK_INT2_SIZE(num_col_im2col);
+  uint16_t num_col_im2col_a = PACK_INT4_SIZE(num_col_im2col);
+
+  //uint8_t *pOut2 = pOut + ch_out_r;
+  int8_t *pA = pWeight;
+
+  uint16_t chan_left = ch_out & 0x3;
+
+  for(int i=0; i < (ch_out >> 2); i++)
+  {
+    uint8_t *pB =  pIn;
+    uint8_t *pB2 = (pB + num_col_im2col_a);
+
+    uint32_t *ptrB  = (uint32_t *) pB;
+    uint32_t *ptrB2 = (uint32_t *) pB2;
+
+    int8_t *pA2 = (pA + num_col_im2col_w);
+    int8_t *pA3 = (pA2 + num_col_im2col_w);
+    int8_t *pA4 = (pA3 + num_col_im2col_w);
+
+    pA  = pulp_nn_i2_to_i4(pA , vecA);
+    pA2 = pulp_nn_i2_to_i4(pA2, vecA2);
+    pA3 = pulp_nn_i2_to_i4(pA3, vecA3);
+    pA4 = pulp_nn_i2_to_i4(pA4, vecA4);
+
+    int32_t *startA;
+    int32_t *startA2;
+    int32_t *startA3;
+    int32_t *startA4;
+
+    asm volatile("mv %0, %1":"=r"(startA):"r"(vecA));
+    asm volatile("mv %0, %1":"=r"(startA2):"r"(vecA2));
+    asm volatile("mv %0, %1":"=r"(startA3):"r"(vecA3));
+    asm volatile("mv %0, %1":"=r"(startA4):"r"(vecA4));
+
+    int32_t *ptrA  = (int32_t *) vecA ;
+    int32_t *ptrA2 = (int32_t *) vecA2;
+    int32_t *ptrA3 = (int32_t *) vecA3;
+    int32_t *ptrA4 = (int32_t *) vecA4;
+
+    ptrA  = MacLoadInit(1, 0, 0, 0, ptrA);
+    ptrA2 = MacLoadInit(1, 0, 1, 0, ptrA2);
+    ptrA3 = MacLoadInit(1, 0, 2, 0, ptrA3);
+    ptrA4 = MacLoadInit(1, 0, 3, 0, ptrA4);
+
+    ptrB  = MacLoadInit(0, 1, 0, 0, ptrB);
+
+    int sum = 0;
+    int sum2 = 0;
+    int sum3 = 0;
+    int sum4 = 0;
+    int sum5 = 0;
+    int sum6 = 0;
+    int sum7 = 0;
+    int sum8 = 0;
+
+
+    if (pBias != NULL)
+    {
+      sum = *((int*)  pBias);
+      pBias+= 4;
+      sum2 = *((int*)  pBias);
+      pBias+= 4;
+      sum3 = *((int*)  pBias);
+      pBias+= 4;
+      sum4 = *((int*)  pBias);
+      pBias+= 4;
+
+      sum5 = sum;
+      sum6 = sum2;
+      sum7 = sum3;
+      sum8 = sum4;
+    }
+
+    for(int j=0; j<(num_col_im2col >> 4); j++)
+    {
+      ptrB2 = MacLoadInit(0, 1, 0, 1, ptrB2);
+
+      sum  = MacLoad8(0, 0, 0, 0, ptrA, sum);
+      sum2 = MacLoad8(0, 0, 1, 0, ptrA2, sum2);
+      sum3 = MacLoad8(0, 0, 2, 0, ptrA3, sum3);
+      sum4 = MacLoad8(0, 1, 3, 0, ptrB, sum4);
+      ptrB = MacLoadUpdate(ptrB);
+
+
+      sum5 = MacLoad8(1, 0, 0, 1, ptrA, sum5);
+      ptrA = MacLoadUpdate(ptrA);
+
+      sum6 = MacLoad8(1, 0, 1, 1, ptrA2, sum6);
+      ptrA2 = MacLoadUpdate(ptrA2);
+
+      sum7 = MacLoad8(1, 0, 2, 1, ptrA3, sum7);
+      ptrA3 = MacLoadUpdate(ptrA3);
+
+      sum8 = MacLoad8(1, 0, 3, 1, ptrA4, sum8);
+      ptrA4 = MacLoadUpdate(ptrA4);
+
+      ptrB2  = MacLoadInit(0, 1, 0, 1, ptrB2);
+
+      sum  = MacLoad8(0, 0, 0, 0, ptrA, sum);
+      sum2 = MacLoad8(0, 0, 1, 0, ptrA2, sum2);
+      sum3 = MacLoad8(0, 0, 2, 0, ptrA3, sum3);
+      sum4 = MacLoad8(0, 1, 3, 0, ptrB, sum4);
+      ptrB = MacLoadUpdate(ptrB);
+
+      pA  = pulp_nn_i2_to_i4(pA , vecA);
+      pA2 = pulp_nn_i2_to_i4(pA2, vecA2);
+      pA3 = pulp_nn_i2_to_i4(pA3, vecA3);
+      pA4 = pulp_nn_i2_to_i4(pA4, vecA4);
+
+      ptrA   = MacLoadAssign(vecA);
+      ptrA2  = MacLoadAssign(vecA2);
+      ptrA3  = MacLoadAssign(vecA3);
+      ptrA4  = MacLoadAssign(vecA4);
+
+      sum5  = MacLoad8(1, 0, 0, 1, ptrA, sum5);
+      ptrA  = MacLoadUpdate(ptrA);
+
+      sum6  = MacLoad8(1, 0, 1, 1, ptrA2, sum6);
+      ptrA2 = MacLoadUpdate(ptrA2);
+
+      sum7  = MacLoad8(1, 0, 2, 1, ptrA3, sum7);
+      ptrA3 = MacLoadUpdate(ptrA3);
+
+      sum8  = MacLoad8(1, 0, 3, 1, ptrA4, sum8);
+      ptrA4 = MacLoadUpdate(ptrA4);
+    }
+    pA-=4;
+    pA2-=4;
+    pA3-=4;
+    pA4-=4;
+
+    int col_cnt_im2col = num_col_im2col & 0xf;
+
+    if(col_cnt_im2col)
+    {
+
+      uint16_t loop_cnt_im2col_a = (num_col_im2col >> 4) << 3;
+      pB+=loop_cnt_im2col_a;
+      pB2+=loop_cnt_im2col_a;
+
+      do
+      {
+        int8_t inA = (int8_t) bitext((int) *pA, 2, 0);
+        int8_t inA2 = (int8_t) bitext((int) *pA2, 2, 0);
+        int8_t inA3 = (int8_t) bitext((int) *pA3, 2, 0);
+        int8_t inA4 = (int8_t) bitext((int) *pA4, 2, 0);
+
+        uint8_t inB = (uint8_t)bitextu((uint32_t) *pB, 4, 0);
+        uint8_t inB2 = (uint8_t)bitextu((uint32_t) *pB2, 4, 0);
+
+        sum += inA * inB;
+        sum2 += inA2 * inB;
+        sum3 += inA3 * inB;
+        sum4 += inA4 * inB;
+
+        sum5 += inA * inB2;
+        sum6 += inA2 * inB2;
+        sum7 += inA3 * inB2;
+        sum8 += inA4 * inB2;
+
+        inA = (int8_t) bitext((int) *pA, 2, 2);
+        inA2 = (int8_t) bitext((int) *pA2, 2, 2);
+        inA3 = (int8_t) bitext((int) *pA3, 2, 2);
+        inA4 = (int8_t) bitext((int) *pA4, 2, 2);
+
+        inB = (uint8_t)bitextu((uint32_t) *pB, 4, 4);
+        inB2 = (uint8_t)bitextu((uint32_t) *pB2, 4, 4);
+
+        sum += inA * inB;
+        sum2 += inA2 * inB;
+        sum3 += inA3 * inB;
+        sum4 += inA4 * inB;
+
+        sum5 += inA * inB2;
+        sum6 += inA2 * inB2;
+        sum7 += inA3 * inB2;
+        sum8 += inA4 * inB2;
+
+        pB++;
+        pB2++;
+
+        inA = (int8_t) bitext((int) *pA, 2, 4);
+        inA2 = (int8_t) bitext((int) *pA2, 2, 4);
+        inA3 = (int8_t) bitext((int) *pA3, 2, 4);
+        inA4 = (int8_t) bitext((int) *pA4, 2, 4);
+
+        inB = (uint8_t)bitextu((uint32_t) *pB, 4, 0);
+        inB2 = (uint8_t)bitextu((uint32_t) *pB2, 4, 0);
+
+        sum += inA * inB;
+        sum2 += inA2 * inB;
+        sum3 += inA3 * inB;
+        sum4 += inA4 * inB;
+
+        sum5 += inA * inB2;
+        sum6 += inA2 * inB2;
+        sum7 += inA3 * inB2;
+        sum8 += inA4 * inB2;
+
+        inA = (int8_t) bitext((int) *pA, 2, 6);
+        inA2 = (int8_t) bitext((int) *pA2, 2, 6);
+        inA3 = (int8_t) bitext((int) *pA3, 2, 6);
+        inA4 = (int8_t) bitext((int) *pA4, 2, 6);
+
+        inB = (uint8_t)bitextu((uint32_t) *pB, 4, 4);
+        inB2 = (uint8_t)bitextu((uint32_t) *pB2, 4, 4);
+
+        sum += inA * inB;
+        sum2 += inA2 * inB;
+        sum3 += inA3 * inB;
+        sum4 += inA4 * inB;
+
+        sum5 += inA * inB2;
+        sum6 += inA2 * inB2;
+        sum7 += inA3 * inB2;
+        sum8 += inA4 * inB2;
+
+        pA++;
+        pA2++;
+        pA3++;
+        pA4++;
+
+        pB++;
+        pB2++;
+
+        col_cnt_im2col-=4;
+      } while(col_cnt_im2col > 0);
+    }
+    if (flag_batch_norm && flag_relu)
+    {
+      *pOut = pulp_nn_bn_quant_u8(sum, *pKappa, *pLambda, out_shift);
+      pOut++;
+      *pOut2 = pulp_nn_bn_quant_u8(sum5, *pKappa, *pLambda, out_shift);
+      pOut2++;
+      pKappa++;
+      pLambda++;
+
+      *pOut = pulp_nn_bn_quant_u8(sum2, *pKappa, *pLambda, out_shift);
+      pOut++;
+      *pOut2 = pulp_nn_bn_quant_u8(sum6, *pKappa, *pLambda, out_shift);
+      pOut2++;
+      pKappa++;
+      pLambda++;
+
+      *pOut = pulp_nn_bn_quant_u8(sum3, *pKappa, *pLambda, out_shift);
+      pOut++;
+      *pOut2 = pulp_nn_bn_quant_u8(sum7, *pKappa, *pLambda, out_shift);
+      pOut2++;
+      pKappa++;
+      pLambda++;
+
+      *pOut = pulp_nn_bn_quant_u8(sum4, *pKappa, *pLambda, out_shift);
+      pOut++;
+      *pOut2 = pulp_nn_bn_quant_u8(sum8, *pKappa, *pLambda, out_shift);
+      pOut2++;
+      pKappa++;
+      pLambda++;
+    }
+    else
+    {
+      if (flag_relu == 1)
+      {
+        *pOut = pulp_nn_quant_u8(sum, out_mult, out_shift);
+        pOut++;
+        *pOut = pulp_nn_quant_u8(sum2, out_mult, out_shift);
+        pOut++;
+        *pOut = pulp_nn_quant_u8(sum3, out_mult, out_shift);
+        pOut++;
+        *pOut = pulp_nn_quant_u8(sum4, out_mult, out_shift);
+        pOut++;
+
+        *pOut2 = pulp_nn_quant_u8(sum5, out_mult, out_shift);
+        pOut2++;
+        *pOut2 = pulp_nn_quant_u8(sum6, out_mult, out_shift);
+        pOut2++;
+        *pOut2 = pulp_nn_quant_u8(sum7, out_mult, out_shift);
+        pOut2++;
+        *pOut2 = pulp_nn_quant_u8(sum8, out_mult, out_shift);
+        pOut2++;
+
+      }
+      else
+      {
+        *pOut = (uint8_t) clip8(sum >> out_shift);
+        pOut++;
+        *pOut = (uint8_t) clip8(sum2 >> out_shift);
+        pOut++;
+        *pOut = (uint8_t) clip8(sum3 >> out_shift);
+        pOut++;
+        *pOut = (uint8_t) clip8(sum4 >> out_shift);
+        pOut++;
+
+        *pOut2 = (uint8_t) clip8(sum5 >> out_shift);
+        pOut2++;
+        *pOut2 = (uint8_t) clip8(sum6 >> out_shift);
+        pOut2++;
+        *pOut2 = (uint8_t) clip8(sum7 >> out_shift);
+        pOut2++;
+        *pOut2 = (uint8_t) clip8(sum8 >> out_shift);
+        pOut2++;
+
+      }
+    }
+    pA+=(3 * num_col_im2col_w);
+  }
+  while(chan_left)
+  {
+    uint8_t *pB = pIn;
+    uint8_t *pB2 = (pB + num_col_im2col_a);
+
+    uint32_t *ptrB  = (uint32_t *) pB;
+    uint32_t *ptrB2 = (uint32_t *) pB2;
+
+    pA  = pulp_nn_i2_to_i4(pA , vecA);
+
+    int32_t *startA;
+
+    asm volatile("mv %0, %1":"=r"(startA):"r"(vecA));
+
+    int32_t *ptrA  = (int32_t *) vecA;
+
+    ptrA  = MacLoadInit(1, 0, 0, 0, ptrA);
+
+    ptrB  = MacLoadInit(0, 1, 0, 0, ptrB);
+
+    int sum = 0;
+    if (pBias != NULL)
+    {
+      sum = *((int*) pBias++);
+    }
+    int sum2 = sum;
+
+    for(int j=0; j < (num_col_im2col >> 4); j++)
+    {
+      ptrB2 = MacLoadInit(0, 1, 0, 1, ptrB2);
+
+      sum  = MacLoad8(0, 1, 0, 0, ptrB, sum);
+      ptrB = MacLoadUpdate(ptrB);
+
+      sum2 = MacLoad8(1, 0, 0, 1, ptrA, sum2);
+      ptrA = MacLoadUpdate(ptrA);
+      ptrB2  = MacLoadInit(0, 1, 0, 1, ptrB2);
+
+      sum  = MacLoad8(0, 1, 0, 0, ptrB, sum);
+      ptrB = MacLoadUpdate(ptrB);
+
+      pA  = pulp_nn_i2_to_i4(pA , vecA);
+
+      ptrA   = MacLoadAssign(vecA);
+
+      sum2  = MacLoad8(1, 0, 0, 1, ptrA, sum2);
+      ptrA  = MacLoadUpdate(ptrA);
+    }
+    pA-=4;
+    int col_cnt_im2col = num_col_im2col & 0xf;
+
+    if(col_cnt_im2col)
+    {
+
+      uint16_t loop_cnt_im2col_a = (num_col_im2col >> 4) << 3;
+      pB+=loop_cnt_im2col_a;
+      pB2+=loop_cnt_im2col_a;
+
+      do
+      {
+        int8_t inA = (int8_t) bitext((int) *pA, 2, 0);
+
+        uint8_t inB = (uint8_t)bitextu((uint32_t) *pB, 4, 0);
+        uint8_t inB2 = (uint8_t)bitextu((uint32_t) *pB2, 4, 0);
+
+        sum += inA * inB;
+
+        sum2 += inA * inB2;
+
+        inA = (int8_t) bitext((int) *pA, 2, 2);
+
+        inB = (uint8_t)bitextu((uint32_t) *pB, 4, 4);
+        inB2 = (uint8_t)bitextu((uint32_t) *pB2, 4, 4);
+
+        sum += inA * inB;
+
+        sum2 += inA * inB2;
+
+        pB++;
+        pB2++;
+
+        inA = (int8_t) bitext((int) *pA, 2, 4);
+
+        inB = (uint8_t)bitextu((uint32_t) *pB, 4, 0);
+        inB2 = (uint8_t)bitextu((uint32_t) *pB2, 4, 0);
+
+        sum += inA * inB;
+
+        sum2 += inA * inB2;
+
+        inA = (int8_t) bitext((int) *pA, 2, 6);
+
+        inB = (uint8_t)bitextu((uint32_t) *pB, 4, 4);
+        inB2 = (uint8_t)bitextu((uint32_t) *pB2, 4, 4);
+
+        sum += inA * inB;
+
+        sum2 += inA * inB2;
+
+        pA++;
+
+        pB++;
+        pB2++;
+
+        col_cnt_im2col-=4;
+      } while(col_cnt_im2col > 0);
+    }
+    if (flag_batch_norm && flag_relu)
+    {
+      *pOut = pulp_nn_bn_quant_u8(sum, *pKappa, *pLambda, out_shift);
+      pOut++;
+      *pOut2 = pulp_nn_bn_quant_u8(sum2, *pKappa, *pLambda, out_shift);
+      pOut2++;
+      pKappa++;
+      pLambda++;
+    }
+    else
+    {
+      if (flag_relu == 1)
+      {
+        *pOut = pulp_nn_quant_u8(sum, out_mult, out_shift);
+        pOut++;
+        *pOut2 = pulp_nn_quant_u8(sum2, out_mult, out_shift);
+        pOut2++;
+      }
+      else
+      {
+        *pOut = (uint8_t) clip8(sum >> out_shift);
+        pOut++;
+        *pOut2 = (uint8_t) clip8(sum2 >> out_shift);
+        pOut2++;
+      }
+    }
+    chan_left--;
+  }
+  pOut+=ch_out_r;
+  return pOut;
+}
diff --git a/rt_nn_tests/xpnn_maxpool_2b/Makefile b/rt_nn_tests/xpnn_maxpool_2b/Makefile
new file mode 100644
index 0000000..ed30544
--- /dev/null
+++ b/rt_nn_tests/xpnn_maxpool_2b/Makefile
@@ -0,0 +1,18 @@
+APP = test
+PULP_APP = test
+PULP_APP_SRCS = test.c
+PULP_APP_SRCS += xpulp_nn_maxpool_i2.c
+
+CORE=8
+
+
+PULP_CFLAGS += -DNUM_CORES=$(CORE) -I. -O3
+PULP_LDFLAGS += -lc -lm -lgcc -Wl,-print-memory-usage 
+
+PULP_APP_CFLAGS += -DNUM_CORES=$(CORE) -I. -O3
+PULP_APP_LDFLAGS += -lc -lm -lgcc -Wl,-print-memory-usage 
+
+PULP_CFLAGS += -DPULP_RUNTIME -DSINGLE_CORE_DMA -DALWAYS_BLOCK_DMA_TRANSFERS
+
+
+include $(PULP_SDK_HOME)/install/rules/pulp.mk
diff --git a/rt_nn_tests/xpnn_maxpool_2b/data_statstest.h b/rt_nn_tests/xpnn_maxpool_2b/data_statstest.h
new file mode 100644
index 0000000..91291ed
--- /dev/null
+++ b/rt_nn_tests/xpnn_maxpool_2b/data_statstest.h
@@ -0,0 +1,10270 @@
+// this file is generated automatically by the python script `generate_statstest.py`
+
+#ifndef _DATA_H
+#define _DATA_H
+
+
+
+#define IM2COL_DIM (32 * NUM_CORES * 2) // for NUM_CORES cores
+#define OUTPUT_DIM 2048
+#define OUTPUT_DIM_FP (1)
+
+uint8_t pIn_0 [] = {
+0xb7,
+0xd3,
+0x41,
+0xbd,
+0xa7,
+0x84,
+0xc9,
+0x6c,
+0xb7,
+0x85,
+0xef,
+0xc3,
+0xe6,
+0xd1,
+0x57,
+0xda,
+0x0a,
+0x5a,
+0x40,
+0xcd,
+0x0b,
+0xd9,
+0xf0,
+0xbd,
+0x01,
+0x9c,
+0x1a,
+0x06,
+0x55,
+0xf8,
+0xee,
+0xcf,
+0x62,
+0x05,
+0xdc,
+0x27,
+0xac,
+0x99,
+0xc0,
+0xc4,
+0xc3,
+0x9a,
+0x69,
+0x04,
+0x51,
+0xf6,
+0xf3,
+0x08,
+0x91,
+0x8e,
+0xd7,
+0xae,
+0x9d,
+0xa6,
+0x43,
+0x4e,
+0x72,
+0xfb,
+0xdd,
+0x59,
+0x39,
+0x37,
+0x69,
+0x9b,
+0x57,
+0x94,
+0xd7,
+0x7b,
+0x42,
+0x0a,
+0x9e,
+0x3f,
+0xe7,
+0xeb,
+0x54,
+0x43,
+0xc8,
+0x0a,
+0xa5,
+0x03,
+0x9e,
+0x11,
+0x6b,
+0x3a,
+0x7e,
+0x6f,
+0x6c,
+0x18,
+0xe7,
+0x87,
+0x7d,
+0x6d,
+0x42,
+0x7b,
+0x84,
+0x90,
+0x13,
+0x38,
+0xbc,
+0xbb,
+0x73,
+0x80,
+0x7a,
+0x66,
+0x38,
+0x73,
+0x2a,
+0xed,
+0x6f,
+0x17,
+0x86,
+0x41,
+0x16,
+0x60,
+0xe3,
+0x2e,
+0x41,
+0x88,
+0x91,
+0x21,
+0x4b,
+0x9e,
+0x5c,
+0xdf,
+0x7d,
+0xbd,
+0x87,
+0xcb,
+0xae,
+0x7f,
+0xa0,
+0xf4,
+0x09,
+0xad,
+0xbd,
+0xb1,
+0x48,
+0xd9,
+0x96,
+0xfc,
+0x9a,
+0x92,
+0xaf,
+0x92,
+0x0a,
+0x49,
+0x0c,
+0x40,
+0xf0,
+0xa7,
+0xb2,
+0xbd,
+0x1c,
+0x91,
+0x5e,
+0x47,
+0xcc,
+0x4e,
+0x9a,
+0xc9,
+0x16,
+0x50,
+0x50,
+0xb9,
+0xb6,
+0x6f,
+0xea,
+0x1b,
+0x75,
+0xb7,
+0x5d,
+0x5a,
+0xec,
+0x47,
+0x9a,
+0x89,
+0x36,
+0xb2,
+0x2e,
+0x97,
+0x4c,
+0xd8,
+0x3e,
+0xb7,
+0xfa,
+0xaa,
+0x4f,
+0x2e,
+0xe0,
+0x68,
+0x1d,
+0xd7,
+0xc9,
+0xa3,
+0x65,
+0xa4,
+0xff,
+0x0a,
+0xa3,
+0x1f,
+0xc1,
+0x92,
+0xf1,
+0x17,
+0x70,
+0x55,
+0xb7,
+0x9c,
+0x7c,
+0xb2,
+0x11,
+0x1c,
+0xc5,
+0x0a,
+0xfd,
+0x6e,
+0xfe,
+0x5c,
+0x28,
+0xd9,
+0x90,
+0xf9,
+0xbc,
+0x5d,
+0xf5,
+0xf9,
+0x9f,
+0x47,
+0x52,
+0x49,
+0x79,
+0xcb,
+0x8c,
+0xec,
+0x2c,
+0xe9,
+0x84,
+0xe3,
+0x66,
+0x50,
+0x18,
+0xb7,
+0x8a,
+0x57,
+0xd5,
+0x8f,
+0xb4,
+0xc9,
+0x37,
+0xe2,
+0x15,
+0x49,
+0x4d,
+0xac,
+0x05,
+0x6c,
+0xbe,
+0x1a,
+0x50,
+0xf8,
+0xa8,
+0x94,
+0x60,
+0x3d,
+0xfa,
+0x90,
+0x6d,
+0x5b,
+0x27,
+0x22,
+0x77,
+0xa5,
+0x20,
+0x9b,
+0xd2,
+0xf1,
+0x5c,
+0x03,
+0x8c,
+0x04,
+0x53,
+0xcf,
+0x1c,
+0x6a,
+0x0d,
+0xfd,
+0xba,
+0x8a,
+0x55,
+0x95,
+0x25,
+0x65,
+0x2c,
+0x4d,
+0x4b,
+0x77,
+0x6c,
+0xc6,
+0x46,
+0xf1,
+0x57,
+0x2c,
+0x1d,
+0x41,
+0x8b,
+0xf4,
+0x08,
+0x62,
+0x26,
+0xd6,
+0x14,
+0x65,
+0xe3,
+0x8c,
+0x63,
+0x34,
+0x7c,
+0x64,
+0xb1,
+0x0e,
+0xfb,
+0x1b,
+0x6c,
+0x9c,
+0x19,
+0xff,
+0xcd,
+0xfc,
+0x7d,
+0xf3,
+0x8c,
+0x4b,
+0x61,
+0xb8,
+0xeb,
+0xd9,
+0x6b,
+0x5b,
+0x79,
+0xb4,
+0x3e,
+0x10,
+0x4c,
+0x1b,
+0x98,
+0xd2,
+0x1a,
+0x62,
+0xf7,
+0x30,
+0xea,
+0xe6,
+0xcf,
+0x1c,
+0x46,
+0xba,
+0x57,
+0x1e,
+0x4e,
+0x73,
+0x02,
+0xbb,
+0x5b,
+0x27,
+0x1b,
+0xe8,
+0xe8,
+0x7b,
+0x20,
+0x95,
+0x70,
+0xa9,
+0x87,
+0x0a,
+0x36,
+0x83,
+0xc8,
+0x1a,
+0x03,
+0x8e,
+0xd9,
+0x10,
+0x9c,
+0x04,
+0x14,
+0xe2,
+0xed,
+0x31,
+0x43,
+0xd2,
+0x76,
+0x75,
+0x5f,
+0x30,
+0xc5,
+0x4c,
+0x7b,
+0x86,
+0xdd,
+0x35,
+0x42,
+0xbe,
+0x90,
+0xad,
+0xb6,
+0x3a,
+0x1c,
+0x7f,
+0x67,
+0xfd,
+0x9f,
+0xa3,
+0xd7,
+0xf3,
+0xdf,
+0xbb,
+0x8d,
+0x16,
+0x20,
+0x51,
+0xa0,
+0xe7,
+0x9b,
+0x68,
+0x0f,
+0x75,
+0xfd,
+0xbc,
+0xa3,
+0x58,
+0x98,
+0x7b,
+0xaa,
+0xb2,
+0x8b,
+0xba,
+0x6f,
+0xe3,
+0x36,
+0xd8,
+0x89,
+0xbb,
+0xbf,
+0xf1,
+0x97,
+0x5c,
+0x9b,
+0xa4,
+0xb8,
+0x83,
+0xc0,
+0x39,
+0x2a,
+0x78,
+0xba,
+0xbf,
+0x87,
+0x92,
+0xcd,
+0xe8,
+0x96,
+0x56,
+0xa9,
+0x18,
+0x57,
+0xc0,
+0xf3,
+0x19,
+0x0b,
+0x70,
+0x84,
+0xa7,
+0x60,
+0x18,
+0x64,
+0x55,
+0x44,
+0xc8,
+0xcb,
+0xa7,
+0x19,
+0xb4,
+0x80,
+0x28,
+0x68,
+0xcb,
+0xfb,
+0x05,
+0xbc,
+0xce,
+0x90,
+0x58,
+0xdb,
+0x1b,
+0x2e,
+0xd4,
+0x7a,
+0xd2,
+0x73,
+0x64,
+0x4e,
+0x5a,
+0x88,
+0x55,
+0xc1,
+0x27,
+0x23,
+0x77,
+0xc8,
+0xbb,
+0x3f,
+0x1b,
+0x06,
+0xbf,
+0x30,
+0x41,
+0xbb,
+0xcc,
+0xa7,
+0xda,
+0x58,
+0x20,
+0xe5,
+0x85,
+0xc6,
+0x08,
+0x16,
+0x4d,
+0x2a,
+0x1e,
+0x0b,
+0x7a,
+0x99,
+0xf9,
+0xc7,
+0x31,
+0x2f,
+0xc5,
+0xea,
+0xc1,
+0xb1,
+0xb2,
+0x62,
+0x90,
+0xcd,
+0xa3,
+0x1f,
+0x10,
+0xb4,
+0xe8,
+0x18,
+0x4d,
+0xa8,
+0x6e,
+0x3c,
+0xf2,
+0xbc,
+0x8d,
+0xcd,
+0xcb,
+0xce,
+0x6d,
+0x31,
+0x13,
+0xad,
+0x38,
+0xb1,
+0x7a,
+0xc8,
+0x53,
+0x86,
+0x5c,
+0x85,
+0x4e,
+0x85,
+0x91,
+0x19,
+0xc0,
+0x2f,
+0xcd,
+0xdb,
+0x3a,
+0x87,
+0xfa,
+0x72,
+0x70,
+0x2b,
+0xdb,
+0x5f,
+0xdc,
+0x50,
+0xf5,
+0xf9,
+0x77,
+0x01,
+0xb1,
+0x67,
+0x60,
+0xd5,
+0xdc,
+0x8f,
+0x6c,
+0xaf,
+0x16,
+0xb7,
+0xdb,
+0x05,
+0xc3,
+0x54,
+0x3e,
+0x04,
+0xa6,
+0xe2,
+0x6b,
+0xb4,
+0x65,
+0x6d,
+0x29,
+0xf3,
+0x2d,
+0x91,
+0xa7,
+0xec,
+0xa5,
+0x25,
+0x27,
+0x5f,
+0xd3,
+0x75,
+0x32,
+0x70,
+0x59,
+0xbe,
+0x04,
+0x15,
+0xcd,
+0xf8,
+0xc9,
+0x98,
+0xfa,
+0x5f,
+0x01,
+0xe7,
+0x7c,
+0x4b,
+0xe0,
+0xfa,
+0x97,
+0xea,
+0xea,
+0xbb,
+0x7d,
+0xbb,
+0x5d,
+0x72,
+0x93,
+0xac,
+0xb6,
+0xac,
+0x90,
+0xdd,
+0x42,
+0xd1,
+0xb6,
+0xf6,
+0x4a,
+0x0e,
+0x70,
+0x36,
+0x31,
+0x10,
+0x3f,
+0x5a,
+0xf9,
+0xa8,
+0xfd,
+0x57,
+0x84,
+0x05,
+0x16,
+0x1a,
+0x3d,
+0x10,
+0x58,
+0x76,
+0x49,
+0xb7,
+0xf0,
+0x23,
+0xdb,
+0x71,
+0xfb,
+0x4a,
+0xa5,
+0x27,
+0xec,
+0xde,
+0x65,
+0xa5,
+0x51,
+0xa1,
+0x07,
+0xab,
+0x10,
+0x8d,
+0x6a,
+0x80,
+0x61,
+0x27,
+0x22,
+0xc0,
+0x61,
+0x4e,
+0x51,
+0x4f,
+0x21,
+0x44,
+0x8a,
+0xdb,
+0x26,
+0xee,
+0x78,
+0xed,
+0xa2,
+0xdd,
+0x59,
+0xc0,
+0x7a,
+0xa0,
+0x09,
+0x7d,
+0x83,
+0x4b,
+0x2b,
+0x09,
+0x80,
+0x9a,
+0x2e,
+0x4a,
+0x9b,
+0x28,
+0x53,
+0x79,
+0xc4,
+0xf0,
+0x6e,
+0x0e,
+0xff,
+0x2e,
+0x15,
+0xf8,
+0xa0,
+0xec,
+0xb0,
+0xd3,
+0xd3,
+0x16,
+0x5f,
+0xdd,
+0x80,
+0xfe,
+0xd3,
+0x2b,
+0xb7,
+0x8e,
+0x53,
+0x0b,
+0x20,
+0x1d,
+0xeb,
+0x52,
+0x05,
+0xb3,
+0x7f,
+0x0a,
+0xf1,
+0x2f,
+0x11,
+0x86,
+0xb2,
+0xf2,
+0xfc,
+0x42,
+0x87,
+0xd2,
+0x4f,
+0x50,
+0x5f,
+0x9e,
+0x73,
+0x15,
+0xaf,
+0xc5,
+0x94,
+0x14,
+0xfa,
+0x65,
+0x12,
+0xf0,
+0x6a,
+0x43,
+0xa8,
+0x66,
+0x11,
+0x8d,
+0xf9,
+0xde,
+0x96,
+0xac,
+0x72,
+0x9d,
+0x16,
+0x7a,
+0x3a,
+0x2a,
+0xdd,
+0x7c,
+0x3c,
+0xfe,
+0x11,
+0x35,
+0xe7,
+0xfd,
+0x8b,
+0xeb,
+0xe1,
+0x33,
+0x92,
+0x59,
+0xb9,
+0xff,
+0x52,
+0x1c,
+0x41,
+0x95,
+0x11,
+0xcb,
+0x5f,
+0xc9,
+0x90,
+0xf0,
+0x50,
+0x48,
+0xa6,
+0x99,
+0x53,
+0x3d,
+0x7a,
+0x79,
+0x78,
+0xd9,
+0x65,
+0x40,
+0x86,
+0x3a,
+0x3b,
+0x32,
+0xcb,
+0xb7,
+0x32,
+0x21,
+0x7c,
+0x4d,
+0x80,
+0xba,
+0xf6,
+0x29,
+0x13,
+0x2d,
+0xb0,
+0xae,
+0xdf,
+0x32,
+0x5c,
+0x9f,
+0x47,
+0xc1,
+0xd1,
+0xb9,
+0xf3,
+0xda,
+0x67,
+0x7d,
+0x21,
+0x02,
+0xf0,
+0x1e,
+0x7e,
+0xc0,
+0x9d,
+0xd4,
+0x68,
+0x4e,
+0x27,
+0x01,
+0xe6,
+0x1c,
+0xe5,
+0xa8,
+0x30,
+0xd8,
+0x0d,
+0x87,
+0x5d,
+0xb7,
+0x36,
+0x4e,
+0x29,
+0xae,
+0x14,
+0xe4,
+0xab,
+0x37,
+0x81,
+0xdb,
+0x2a,
+0x76,
+0xb8,
+0x75,
+0x93,
+0x39,
+0x84,
+0x2b,
+0xb0,
+0xd6,
+0xa7,
+0x1f,
+0x31,
+0x2e,
+0xda,
+0x09,
+0xf4,
+0xc8,
+0xc5,
+0x64,
+0x10,
+0xaa,
+0xf4,
+0xcf,
+0x74,
+0x3c,
+0x98,
+0xc8,
+0x48,
+0xb8,
+0xea,
+0xd2,
+0x47,
+0x1e,
+0x96,
+0xce,
+0x71,
+0x47,
+0x44,
+0x8f,
+0xde,
+0x7a,
+0x01,
+0x8d,
+0x3d,
+0x37,
+0x1f,
+0xd8,
+0xc1,
+0x21,
+0xe3,
+0xa2,
+0x3d,
+0x51,
+0x82,
+0x5e,
+0x13,
+0x4b,
+0x08,
+0x76,
+0x6c,
+0x17,
+0x67,
+0x9a,
+0xfb,
+0x21,
+0x46,
+0x97,
+0xe0,
+0xd8,
+0xf3,
+0x1a,
+0x74,
+0x5e,
+0x48,
+0xab,
+0x33,
+0xc7,
+0x05,
+0xa6,
+0x9b,
+0x42,
+0x88,
+0x31,
+0xed,
+0xf6,
+0x7c,
+0xe9,
+0x07,
+0x05,
+0xea,
+0xff,
+0x0a,
+0x43,
+0x3f,
+0xe2,
+0x98,
+0x30,
+0x61,
+0x17,
+0xe9,
+0x2f,
+0x31,
+0xae,
+0x87,
+0x0d,
+0xe1,
+0x66,
+0x2f,
+0x57,
+0xe0,
+0xe4,
+0x34,
+0x1b,
+0xc7,
+0x92,
+0x4b,
+0xee,
+0x1a,
+0x69,
+0x73,
+0x14,
+0x4d,
+0x51,
+0x07,
+0x04,
+0xe5,
+0x97,
+0x91,
+0xea,
+0xa9,
+0x7b,
+0x06,
+0xb9,
+0x54,
+0xbe,
+0x2f,
+0x8d,
+0x6e,
+0x4f,
+0xca,
+0x8d,
+0x17,
+0xb8,
+0x18,
+0x2e,
+0xa5,
+0x31,
+0x9f,
+0x9b,
+0x28,
+0x21,
+0x5f,
+0xcc,
+0x95,
+0xe2,
+0x79,
+0x72,
+0x0c,
+0xcb,
+0x6a,
+0x5f,
+0xa2,
+0x28,
+0x8e,
+0x8e,
+0x85,
+0x2d,
+0xcc,
+0x59,
+0x0e,
+0xaf,
+0x00,
+0x34,
+0xad,
+0xda,
+0x9a,
+0x0a,
+0xda,
+0x44,
+0x07,
+0x1e,
+0x4b,
+0x7c,
+0x6e,
+0xb7,
+0xa9,
+0x72,
+0x66,
+0x88,
+0xc7,
+0xe9,
+0xd6,
+0x72,
+0x32,
+0xc8,
+0xb2,
+0xe1,
+0x9b,
+0xa0,
+0x39,
+0x25,
+0xe0,
+0xd6,
+0xf6,
+0x85,
+0xda,
+0x5f,
+0x8d,
+0x13,
+0xc9,
+0x26,
+0x1c,
+0x21,
+0x0c,
+0x78,
+0x72,
+0x56,
+0xb5,
+0x7a,
+0x0c,
+0xb3,
+0xd4,
+0x04,
+0xc0,
+0x58,
+0x63,
+0x89,
+0xfd,
+0x2c,
+0x02,
+0x98,
+0x91,
+0x21,
+0x80,
+0xb7,
+0xf6,
+0x3b,
+0x5e,
+0x3d,
+0x1e,
+0xa5,
+0x9d,
+0xcc,
+0xeb,
+0x0f,
+0x31,
+0xf4,
+0x3b,
+0xbb,
+0x06,
+0x9b,
+0x4d,
+0x5c,
+0xa7,
+0xf5,
+0xfb,
+0x85,
+0xca,
+0xaf,
+0xe4,
+0x9e,
+0x4e,
+0x5e,
+0x31,
+0x53,
+0x3d,
+0x6e,
+0x94,
+0x4c,
+0x5c,
+0x3d,
+0x7b,
+0x92,
+0xd6,
+0x6b,
+0x2f,
+0x5b,
+0x26,
+0x2f,
+0xe7,
+0xc1,
+0x79,
+0xde,
+0xe5,
+0x4a,
+0x73,
+0xcb,
+0xeb,
+0x7d,
+0x56,
+0x9a,
+0x49,
+0x33,
+0xcd,
+0xb2,
+0xc8,
+0xa2,
+0x4b,
+0x47,
+0x41,
+0x6b,
+0x6b,
+0xb5,
+0x66,
+0x78,
+0x05,
+0xf4,
+0xc9,
+0xc2,
+0x9e,
+0x77,
+0x69,
+0x5f,
+0x39,
+0x50,
+0xeb,
+0x77,
+0x6e,
+0xe9,
+0xb9,
+0xd3,
+0x22,
+0xe2,
+0xa9,
+0x30,
+0x03,
+0xdb,
+0xb2,
+0xea,
+0xae,
+0x10,
+0x8b,
+0x46,
+0x6b,
+0x9c,
+0xf8,
+0xea,
+0x88,
+0x47,
+0x62,
+0x78,
+0x89,
+0x4c,
+0x5e,
+0x5e,
+0xff,
+0x7a,
+0x4e,
+0x24,
+0x33,
+0x0e,
+0x4d,
+0x43,
+0x02,
+0x68,
+0x92,
+0x1b,
+0x6d,
+0xe6,
+0xe5,
+0x94,
+0xeb,
+0x40,
+0x8c,
+0x77,
+0x4c,
+0x50,
+0x36,
+0x26,
+0x9d,
+0x9f,
+0x6e,
+0xea,
+0x4c,
+0xf5,
+0xeb,
+0xf1,
+0x6c,
+0x3c,
+0x06,
+0x9e,
+0xea,
+0x17,
+0xcd,
+0x48,
+0x8f,
+0x29,
+0xec,
+0x9a,
+0x11,
+0xb2,
+0x47,
+0xff,
+0x75,
+0x23,
+0x22,
+0x15,
+0x2f,
+0x5c,
+0xa1,
+0x7a,
+0x8f,
+0x20,
+0x4b,
+0x2f,
+0xc6,
+0x2c,
+0x3b,
+0x5d,
+0x67,
+0xbd,
+0x67,
+0xcf,
+0x66,
+0x87,
+0x2a,
+0x9a,
+0x49,
+0xcc,
+0x49,
+0x66,
+0xd1,
+0x17,
+0xb1,
+0x65,
+0x0d,
+0xe6,
+0x14,
+0x9a,
+0x57,
+0x84,
+0x5a,
+0x55,
+0x00,
+0x70,
+0x32,
+0x56,
+0x09,
+0xa7,
+0x2a,
+0xfd,
+0x94,
+0xb6,
+0x81,
+0x8c,
+0x38,
+0x46,
+0x0b,
+0x7f,
+0x52,
+0x53,
+0x2a,
+0x90,
+0xd1,
+0x1c,
+0xd8,
+0x35,
+0xb6,
+0xea,
+0xec,
+0xdc,
+0xae,
+0x16,
+0xa6,
+0x76,
+0x29,
+0x2d,
+0xc7,
+0x06,
+0xb5,
+0x10,
+0x97,
+0x28,
+0x3c,
+0xdd,
+0x0b,
+0x94,
+0x08,
+0x97,
+0x82,
+0xa1,
+0x1b,
+0x66,
+0x86,
+0x51,
+0x03,
+0x89,
+0x1e,
+0x2d,
+0x5a,
+0xb6,
+0x56,
+0xd8,
+0xc9,
+0x26,
+0xc0,
+0x51,
+0x6a,
+0xb6,
+0x7c,
+0x99,
+0x32,
+0xd3,
+0x8c,
+0xe8,
+0x76,
+0xb4,
+0x87,
+0x65,
+0x99,
+0x60,
+0xfa,
+0x6c,
+0x17,
+0xa5,
+0x47,
+0x31,
+0x25,
+0x0b,
+0x25,
+0x85,
+0xc2,
+0xb4,
+0xb2,
+0x38,
+0x72,
+0x4b,
+0xe6,
+0x59,
+0x40,
+0xf5,
+0x70,
+0x5c,
+0x50,
+0x86,
+0xa4,
+0xe1,
+0x77,
+0x3a,
+0xe3,
+0x4c,
+0x7a,
+0xc2,
+0x9d,
+0x08,
+0x0c,
+0x7c,
+0xcb,
+0x3b,
+0xf1,
+0xd0,
+0x5f,
+0x1a,
+0xde,
+0x31,
+0x2b,
+0xc0,
+0xd2,
+0x3f,
+0xff,
+0xe2,
+0xbe,
+0x3e,
+0x07,
+0xa5,
+0x47,
+0xe2,
+0x1f,
+0x2e,
+0xcf,
+0x87,
+0xa3,
+0xe2,
+0x59,
+0x71,
+0x2c,
+0x3a,
+0xb4,
+0xd0,
+0x78,
+0xff,
+0xb1,
+0xe3,
+0x76,
+0xa2,
+0xf2,
+0x05,
+0xb2,
+0xd7,
+0x1f,
+0x4b,
+0x6c,
+0x93,
+0xe7,
+0xf6,
+0x29,
+0x53,
+0xec,
+0x7f,
+0xb9,
+0x12,
+0x30,
+0x88,
+0x01,
+0xb5,
+0xe9,
+0x5b,
+0x90,
+0x37,
+0x31,
+0x3d,
+0xba,
+0x5f,
+0x17,
+0x20,
+0x65,
+0x84,
+0xbf,
+0x3b,
+0x89,
+0x0d,
+0x28,
+0xcb,
+0x6e,
+0x9f,
+0x51,
+0xb3,
+0x58,
+0xb0,
+0x6f,
+0xc3,
+0x34,
+0x03,
+0x1f,
+0x58,
+0xf0,
+0x7f,
+0xe8,
+0x13,
+0x47,
+0x7f,
+0x55,
+0x80,
+0xeb,
+0x55,
+0x2f,
+0x7e,
+0xd1,
+0x61,
+0x35,
+0x6c,
+0x95,
+0x87,
+0x27,
+0x7c,
+0x83,
+0x13,
+0x6b,
+0x3b,
+0x81,
+0xdf,
+0x89,
+0xcb,
+0xa5,
+0x22,
+0xe5,
+0x7a,
+0x23,
+0x35,
+0x41,
+0x26,
+0x90,
+0x6b,
+0x62,
+0x90,
+0x2f,
+0x21,
+0x68,
+0x5c,
+0x2f,
+0x0a,
+0xb3,
+0xaa,
+0xed,
+0x22,
+0x75,
+0x12,
+0xff,
+0x0e,
+0xbf,
+0xf7,
+0x71,
+0xca,
+0xee,
+0xa5,
+0x56,
+0x82,
+0x65,
+0x7c,
+0xe2,
+0x53,
+0x53,
+0x86,
+0x9c,
+0x36,
+0x72,
+0x13,
+0x7b,
+0x43,
+0x64,
+0xa5,
+0x5a,
+0x96,
+0x30,
+0xb7,
+0x4f,
+0xea,
+0xcd,
+0x23,
+0x6a,
+0x37,
+0x35,
+0x1d,
+0x42,
+0x0f,
+0x8f,
+0x1c,
+0x52,
+0xad,
+0xa7,
+0xa7,
+0xfc,
+0xc5,
+0xa0,
+0x21,
+0xcd,
+0x0d,
+0x13,
+0x4f,
+0x9d,
+0xad,
+0xe9,
+0x03,
+0x5c,
+0x36,
+0x65,
+0xff,
+0xb1,
+0xb8,
+0xc5,
+0xd4,
+0x4d,
+0x36,
+0x47,
+0x90,
+0x65,
+0xd1,
+0xca,
+0xf7,
+0x60,
+0x10,
+0xdc,
+0xf7,
+0x32,
+0xd2,
+0xb9,
+0xc6,
+0x28,
+0x63,
+0x4a,
+0x47,
+0xf9,
+0xc7,
+0xd6,
+0x2d,
+0x29,
+0xb1,
+0x3c,
+0xf4,
+0x91,
+0x62,
+0xec,
+0x17,
+0x1c,
+0x2c,
+0xc3,
+0xb9,
+0x7e,
+0x96,
+0x49,
+0xb0,
+0x39,
+0xc9,
+0xd4,
+0x2c,
+0x8c,
+0xc1,
+0xe1,
+0x60,
+0x33,
+0x97,
+0x28,
+0x26,
+0x20,
+0x57,
+0x65,
+0xb8,
+0xf4,
+0x7b,
+0xcf,
+0xa0,
+0x29,
+0x54,
+0x9b,
+0x82,
+0x75,
+0x35,
+0x61,
+0x15,
+0xb0,
+0xbd,
+0x9c,
+0xdd,
+0x8d,
+0x8c,
+0x41,
+0x45,
+0x86,
+0xdb,
+0xa4,
+0xdf,
+0xbb,
+0xf5,
+0x52,
+0x89,
+0x23,
+0xb7,
+0x10,
+0xba,
+0x17,
+0xf6,
+0x56,
+0x08,
+0x2c,
+0xac,
+0x64,
+0x32,
+0x48,
+0x16,
+0x2e,
+0x47,
+0xce,
+0x07,
+0x7e,
+0x42,
+0x48,
+0x0e,
+0x24,
+0x50,
+0x49,
+0xe3,
+0xc9,
+0xb5,
+0x90,
+0x66,
+0x0b,
+0xf2,
+0x01,
+0xe8,
+0x9b,
+0x5f,
+0x7d,
+0xa0,
+0x20,
+0x14,
+0x4e,
+0x47,
+0xbc,
+0x82,
+0xc3,
+0xce,
+0xb3,
+0x60,
+0xdc,
+0xef,
+0x42,
+0xed,
+0x02,
+0x68,
+0x64,
+0x01,
+0x7d,
+0xbf,
+0xef,
+0x1b,
+0xc3,
+0xdd,
+0xe4,
+0xfa,
+0xa4,
+0x02,
+0x3e,
+0x7c,
+0xf6,
+0x60,
+0x94,
+0xd2,
+0xa0,
+0xbd,
+0x58,
+0xda,
+0xff,
+0xe7,
+0x2a,
+0x51,
+0x6e,
+0x30,
+0x7a,
+0x09,
+0xac,
+0x55,
+0xce,
+0xb5,
+0xe0,
+0x7a,
+0x39,
+0x2f,
+0x53,
+0x56,
+0x25,
+0xa2,
+0x7c,
+0x19,
+0xec,
+0x07,
+0x07,
+0xf6,
+0xbb,
+0xc3,
+0xb3,
+0xe1,
+0x3a,
+0x35,
+0x9c,
+0x45,
+0xc0,
+0x24,
+0x71,
+0x39,
+0x60,
+0xf8,
+0x96,
+0xa5,
+0xe7,
+0x5c,
+0x6a,
+0x73,
+0x62,
+0xf8,
+0xda,
+0x20,
+0xef,
+0x08,
+0xe6,
+0xa9,
+0x8d,
+0xb7,
+0x59,
+0x45,
+0x46,
+0xc1,
+0x66,
+0x18,
+0x6a,
+0x25,
+0x62,
+0xdd,
+0x11,
+0xa1,
+0xf1,
+0x0e,
+0xdd,
+0x4e,
+0x2b,
+0x43,
+0x52,
+0x0b,
+0xd7,
+0xe5,
+0x67,
+0xce,
+0x0b,
+0xdf,
+0x08,
+0x84,
+0x23,
+0x73,
+0x99,
+0xf3,
+0x77,
+0x3a,
+0x90,
+0x24,
+0x53,
+0xee,
+0xf9,
+0x0c,
+0x6f,
+0xf5,
+0x0a,
+0x6f,
+0xa5,
+0xc2,
+0xda,
+0x3a,
+0x78,
+0x23,
+0x32,
+0x4d,
+0x9b,
+0x8e,
+0x3c,
+0x4c,
+0x81,
+0x69,
+0x7f,
+0x96,
+0xcc,
+0xe0,
+0xe2,
+0x05,
+0xfc,
+0xec,
+0x76,
+0x7f,
+0x85,
+0x0c,
+0xec,
+0x32,
+0x73,
+0xb9,
+0x01,
+0xa8,
+0x1c,
+0x00,
+0x69,
+0x0e,
+0x0e,
+0xe0,
+0xd4,
+0x49,
+0x05,
+0xc6,
+0xb9,
+0x78,
+0x6c,
+0x94,
+0x3c,
+0x7a,
+0xab,
+0x32,
+0x42,
+0x07,
+0x31,
+0xce,
+0xc9,
+0xcd,
+0x0d,
+0xc4,
+0xce,
+0x36,
+0x02,
+0x96,
+0xd5,
+0x51,
+0x58,
+0x64,
+0xd5,
+0x01,
+0xe0,
+0x72,
+0xcf,
+0x81,
+0xf0,
+0xc3,
+0xd4,
+0x75,
+0x96,
+0x7a,
+0x4d,
+0x6a,
+0x6a,
+0xa0,
+0x40,
+0xf2,
+0xce,
+0x1e,
+0x01,
+0xfb,
+0xa5,
+0x97,
+0xfa,
+0x3c,
+0xdd,
+0x7c,
+0xac,
+0xf3,
+0x5c,
+0xf4,
+0xe8,
+0xcc,
+0x8e,
+0x4d,
+0xb4,
+0xc7,
+0xd6,
+0x47,
+0xc5,
+0x4d,
+0x24,
+0x44,
+0x6a,
+0x43,
+0x33,
+0x38,
+0x84,
+0x69,
+0xbc,
+0x1d,
+0x62,
+0xa0,
+0x98,
+0xa5,
+0x82,
+0xfc,
+0x5a,
+0x71,
+0x53,
+0xbe,
+0x78,
+0xa0,
+0x09,
+0xc6,
+0xc2,
+0x10,
+0x30,
+0xd3,
+0xea,
+0xcb,
+0x1f,
+0x61,
+0x8c,
+0xd3,
+0xd9,
+0x28,
+0x5b,
+0xf5,
+0x1a,
+0x9d,
+0xaf,
+0x42,
+0xbf,
+0xbe,
+0xca,
+0x2e,
+0x88,
+0xe4,
+0x8f,
+0xcd,
+0x16,
+0x34,
+0xc5,
+0x77,
+0x50,
+0x37,
+0xd1,
+0xe9,
+0x2a,
+0xbf,
+0xd1,
+0x77,
+0x56,
+0xed,
+0x74,
+0xa3,
+0x1c,
+0x26,
+0x08,
+0x4d,
+0x36,
+0xf3,
+0x79,
+0xd5,
+0x13,
+0xb4,
+0xce,
+0x62,
+0xdc,
+0x47,
+0xd5,
+0xa8,
+0xee,
+0xe9,
+0x7c,
+0x42,
+0x02,
+0xf9,
+0x9e,
+0xe5,
+0x15,
+0xeb,
+0xd9,
+0xdc,
+0x2c,
+0xf5,
+0x50,
+0x3d,
+0x58,
+0xa8,
+0x98,
+0x74,
+0xb4,
+0x22,
+0x85,
+0x71,
+0x91,
+0xb9,
+0x3c,
+0xc4,
+0x4d,
+0x1c,
+0xa5,
+0xc8,
+0x86,
+0xd6,
+0xb0,
+0xe6,
+0x1f,
+0xbc,
+0x29,
+0x2b,
+0xaf,
+0x6f,
+0x2d,
+0xfc,
+0x7a,
+0x8e,
+0x6f,
+0xca,
+0x34,
+0x2e,
+0x64,
+0xda,
+0xb4,
+0x0f,
+0xc9,
+0x2d,
+0x28,
+0x2f,
+0x36,
+0xc9,
+0xd9,
+0x53,
+0xc9,
+0xd0,
+0x81,
+0x03,
+0xdd,
+0x08,
+0xd5,
+0xfc,
+0xf9,
+0x59,
+0x45,
+0x76,
+0x7e,
+0x20,
+0xbf,
+0x4f,
+0x55,
+0x00,
+0x48,
+0x45,
+0x9d,
+0x1c,
+0x2f,
+0x09,
+0x20,
+0xe9,
+0x27,
+0xcf,
+0x00,
+0x6f,
+0x48,
+0x0d,
+0xf0,
+0x00,
+0x18,
+0x58,
+0x4e,
+0x8e,
+0x51,
+0x61,
+0x8f,
+0xa9,
+0x15,
+0x43,
+0x72,
+0xab,
+0xdc,
+0x1a,
+0xe6,
+0x7d,
+0xfd,
+0x21,
+0x94,
+0xf4,
+0x5c,
+0x2e,
+0x88,
+0x81,
+0x82,
+0xf3,
+0x45,
+0xd4,
+0x48,
+0x92,
+0x5d,
+0x76,
+0xf6,
+0x97,
+0xcf,
+0x0a,
+0x8a,
+0xb1,
+0x4d,
+0x2f,
+0x31,
+0x04,
+0x26,
+0xee,
+0x15,
+0x77,
+0x88,
+0x91,
+0x06,
+0xcb,
+0x31,
+0x8f,
+0x2a,
+0xe5,
+0x4a,
+0x54,
+0x22,
+0x9e,
+0x12,
+0xf5,
+0x83,
+0x99,
+0x12,
+0xe4,
+0x10,
+0x77,
+0x59,
+0x15,
+0x59,
+0x51,
+0xa8,
+0xa3,
+0xc3,
+0x8f,
+0xe9,
+0x33,
+0xeb,
+0xb9,
+0x0c,
+0xb4,
+0xdf,
+0x7c,
+0xc2,
+0xdc,
+0xe0,
+0x65,
+0xfc,
+0xd2,
+0x70,
+0x84,
+0xaa,
+0x81,
+0xac,
+0xd1,
+0x29,
+0xf6,
+0x9a,
+0x78,
+0x8b,
+0xe1,
+0x0c,
+0x12,
+0x3b,
+0x70,
+0x96,
+0x32,
+0xba,
+0xf7,
+0x6a,
+0x04,
+0xa2,
+0x74,
+0xb0,
+0x77,
+0xd7,
+0x69,
+0x69,
+0x43,
+0x27,
+0xc4,
+0x8b,
+0x9c,
+0x1a,
+0x26,
+0x74,
+0xfd,
+0x8e,
+0x42,
+0x7c,
+0x35,
+0x84,
+0x51,
+0xdb,
+0x88,
+0x2d,
+0x83,
+0x1e,
+0x2e,
+0x8b,
+0xb5,
+0xfa,
+0x15,
+0xc7,
+0xaa,
+0x43,
+0x2e,
+0x50,
+0x91,
+0x6a,
+0xdb,
+0xf9,
+0x26,
+0xd8,
+0x76,
+0xa3,
+0xe8,
+0x54,
+0x11,
+0x34,
+0xd2,
+0x8a,
+0x2f,
+0x50,
+0x0a,
+0x3e,
+0x19,
+0x5e,
+0x76,
+0x9b,
+0x68,
+0x52,
+0x83,
+0x6e,
+0x7a,
+0x04,
+0x59,
+0x5c,
+0xbf,
+0x77,
+0x20,
+0xd2,
+0x23,
+0xb0,
+0x9a,
+0xe2,
+0xe3,
+0x27,
+0x32,
+0x03,
+0xa2,
+0xaf,
+0xbe,
+0xf6,
+0xf5,
+0x87,
+0x4a,
+0x60,
+0xd8,
+0xca,
+0x20,
+0xf7,
+0x40,
+0x3d,
+0xae,
+0x23,
+0x9e,
+0xdf,
+0x68,
+0x2a,
+0x30,
+0x0a,
+0x6c,
+0xb0,
+0x01,
+0x9c,
+0xee,
+0x4f,
+0xe4,
+0x4f,
+0x5f,
+0xb1,
+0x11,
+0xce,
+0x28,
+0xaf,
+0x27,
+0xb5,
+0x13,
+0xa1,
+0x76,
+0xb9,
+0x81,
+0xa5,
+0xbb,
+0x38,
+0x5a,
+0xa1,
+0xd7,
+0x27,
+0x00,
+0x8a,
+0x3e,
+0x72,
+0xb4,
+0x9e,
+0x31,
+0x61,
+0x27,
+0x71,
+0x63,
+0xa7,
+0xc0,
+0xdd,
+0x48,
+0x50,
+0x0f,
+0x1a,
+0xd8,
+0x7e,
+0x97,
+0x52,
+0xc5,
+0xf1,
+0xfe,
+0x28,
+0x5c,
+0x4c,
+0xfa,
+0xd6,
+0xd4,
+0x85,
+0x6d,
+0x40,
+0x02,
+0x82,
+0x46,
+0x6c,
+0x3e,
+0x5d,
+0xf0,
+0x2f,
+0x31,
+0xa6,
+0x65,
+0x91,
+0x83,
+0xa6,
+0x20,
+0xdc,
+0xc2,
+0xb3,
+0xbe,
+0x85,
+0x15,
+0xaa,
+0x84,
+0x72,
+0xc5,
+0x2e,
+0xa0,
+0x2c,
+0xc2,
+0x5d,
+0xf4,
+0xa9,
+0x64,
+0x60,
+0xc4,
+0x07,
+0xef,
+0xee,
+0x0b,
+0x91,
+0x54,
+0x0a,
+0x5e,
+0xb5,
+0xb8,
+0xa9,
+0x49,
+0xbc,
+0xb0,
+0x0f,
+0x0b,
+0x4c,
+0xf0,
+0x50,
+0x11,
+0xac,
+0x11,
+0x38,
+0xfe,
+0x91,
+0x1d,
+0x7e,
+0x8d,
+0x39,
+0x29,
+0x48,
+0x66,
+0x31,
+0x94,
+0xaf,
+0xb5,
+0xbf,
+0xe1,
+0xd2,
+0x5d,
+0x44,
+0x83,
+0xb4,
+0xe2,
+0x7c,
+0xf2,
+0x81,
+0x79,
+0x8e,
+0xf2,
+0x95,
+0x17,
+0xe7,
+0xfa,
+0x5f,
+0xa4,
+0xf0,
+0xb4,
+0x32,
+0xcd,
+0xaa,
+0xad,
+0xdb,
+0x9b,
+0x9c,
+0x1c,
+0x20,
+0x75,
+0xf0,
+0xae,
+0x37,
+0xc5,
+0x3d,
+0x4d,
+0x86,
+0x31,
+0xc6,
+0xdd,
+0x2e,
+0x67,
+0x97,
+0x43,
+0x99,
+0x67,
+0x4d,
+0x0b,
+0x95,
+0x49,
+0xf1,
+0x6c,
+0xb0,
+0xe3,
+0x8a,
+0xfa,
+0x26,
+0x7b,
+0x44,
+0x11,
+0x35,
+0x79,
+0x7e,
+0xa9,
+0x44,
+0x65,
+0x94,
+0x66,
+0xf7,
+0x93,
+0x7b,
+0x4b,
+0xd8,
+0x6c,
+0xc1,
+0x7a,
+0xca,
+0x6a,
+0x82,
+0x8c,
+0xc0,
+0xa0,
+0x5d,
+0xfc,
+0x9c,
+0x1c,
+0x51,
+0x62,
+0x5d,
+0x21,
+0x94,
+0xda,
+0x05,
+0x9a,
+0x25,
+0x69,
+0x13,
+0x37,
+0x19,
+0x06,
+0xd1,
+0x59,
+0x03,
+0xca,
+0x10,
+0x1b,
+0x0a,
+0xfb,
+0x73,
+0xfb,
+0xf5,
+0xfe,
+0x5b,
+0xd6,
+0x55,
+0x02,
+0x2f,
+0x1d,
+0x4c,
+0xd1,
+0x8f,
+0x2a,
+0xba,
+0x75,
+0xc8,
+0xb1,
+0x0e,
+0xc4,
+0x4f,
+0x7c,
+0xc2,
+0x2b,
+0x69,
+0x90,
+0x99,
+0x83,
+0x98,
+0x23,
+0x3a,
+0x4e,
+0x89,
+0x8f,
+0x50,
+0xff,
+0x8e,
+0x0a,
+0x75,
+0xf4,
+0xc9,
+0x3d,
+0x10,
+0x2d,
+0xea,
+0x79,
+0x9c,
+0xf7,
+0x9f,
+0xac,
+0x94,
+0xdd,
+0x5d,
+0xf9,
+0xb6,
+0x6d,
+0x82,
+0x67,
+0x95,
+0x6e,
+0xae,
+0xe2,
+0x2c,
+0x7e,
+0x01,
+0x2c,
+0xc3,
+0x56,
+0x6c,
+0x75,
+0x59,
+0xf2,
+0xca,
+0x2e,
+0xa4,
+0x4e,
+0xb1,
+0x83,
+0x4f,
+0x51,
+0x39,
+0x76,
+0x5e,
+0x78,
+0x2a,
+0x2b,
+0x48,
+0xc2,
+0x95,
+0x63,
+0x1a,
+0xda,
+0xb3,
+0xbf,
+0x1a,
+0x5e,
+0x78,
+0x4b,
+0x95,
+0x47,
+0x0b,
+0x80,
+0xf8,
+0xf4,
+0x9b,
+0x27,
+0x03,
+0x21,
+0x2f,
+0x49,
+0x66,
+0xdf,
+0x7d,
+0xad,
+0x9b,
+0xa9,
+0xf3,
+0x26,
+0xe2,
+0x3c,
+0x17,
+0x9f,
+0x29,
+0xc8,
+0xa3,
+0x23,
+0x4a,
+0x50,
+0xa1,
+0x74,
+0x9c,
+0xda,
+0x01,
+0xfd,
+0xd8,
+0xe7,
+0xed,
+0x72,
+0x70,
+0x98,
+0x6c,
+0x3a,
+0xd8,
+0x88,
+0xd7,
+0xc6,
+0x81,
+0x89,
+0xf1,
+0xa4,
+0x3d,
+0x4c,
+0x3c,
+0x57,
+0xe5,
+0xfa,
+0xe3,
+0x6f,
+0x04,
+0x01,
+0xe4,
+0x7f,
+0xba,
+0x2b,
+0x05,
+0x6e,
+0xc8,
+0xe9,
+0x52,
+0x49,
+0x8a,
+0x7a,
+0x5c,
+0xca,
+0x0f,
+0xdb,
+0x58,
+0x1f,
+0xe4,
+0xbd,
+0xf5,
+0x8c,
+0xe5,
+0x7c,
+0xa0,
+0xb3,
+0xd9,
+0x00,
+0x95,
+0xe4,
+0x05,
+0xe4,
+0x3a,
+0x4b,
+0x67,
+0xf4,
+0x51,
+0xca,
+0x26,
+0x9e,
+0x55,
+0xdd,
+0xbd,
+0x14,
+0x2f,
+0x50,
+0x29,
+0x0f,
+0x02,
+0xb2,
+0x42,
+0x0a,
+0xc3,
+0x45,
+0xa2,
+0x67,
+0xef,
+0x48,
+0x6b,
+0x5c,
+0x58,
+0xc3,
+0x5e,
+0xc3,
+0x10,
+0x87,
+0x1e,
+0x14,
+0x2d,
+0x07,
+0x1a,
+0x55,
+0xe1,
+0x46,
+0x6f,
+0x9e,
+0x0f,
+0xa5,
+0x7e,
+0xb1,
+0x79,
+0xe1,
+0x21,
+0x09,
+0x52,
+0xc2,
+0x50,
+0xcc,
+0xe5,
+0x05,
+0xf5,
+0x57,
+0x88,
+0x21,
+0x2c,
+0xd7,
+0xc1,
+0x2b,
+0x77,
+0x44,
+0xc2,
+0x4e,
+0x3f,
+0xf0,
+0xea,
+0x7f,
+0x6a,
+0x4f,
+0x97,
+0xe1,
+0x36,
+0xaf,
+0x7d,
+0x1f,
+0x12,
+0xdc,
+0x65,
+0x3b,
+0xcd,
+0x07,
+0xc2,
+0x6f,
+0xdf,
+0xe7,
+0x39,
+0x18,
+0x54,
+0xd3,
+0x3a,
+0x64,
+0x84,
+0x34,
+0x08,
+0xc8,
+0x3e,
+0xa8,
+0xdf,
+0x4e,
+0xc8,
+0x0d,
+0x48,
+0x61,
+0x2b,
+0x9a,
+0xdc,
+0xe8,
+0xfa,
+0xf0,
+0x37,
+0x9b,
+0xaf,
+0x06,
+0x8b,
+0x78,
+0x0f,
+0x67,
+0xc2,
+0x5b,
+0x6b,
+0xa4,
+0x01,
+0x40,
+0x51,
+0xa9,
+0x4a,
+0x9b,
+0x01,
+0xd2,
+0xde,
+0x31,
+0x1d,
+0x46,
+0xe4,
+0xbc,
+0xd0,
+0x93,
+0x57,
+0x16,
+0x91,
+0x79,
+0x67,
+0x12,
+0x3b,
+0x83,
+0x95,
+0x21,
+0x2f,
+0xec,
+0xf1,
+0x62,
+0x4f,
+0x24,
+0x5b,
+0xdc,
+0xfe,
+0x19,
+0xec,
+0x45,
+0x1c,
+0x9f,
+0x8d,
+0x1c,
+0x7c,
+0x13,
+0x50,
+0x54,
+0x29,
+0x6b,
+0xe7,
+0x03,
+0x73,
+0x29,
+0x5f,
+0xd8,
+0x5e,
+0x05,
+0x51,
+0x94,
+0xb7,
+0x08,
+0x40,
+0xd7,
+0x8f,
+0x09,
+0x9b,
+0xc6,
+0x11,
+0x5c,
+0x86,
+0xae,
+0xfa,
+0x8b,
+0xb4,
+0x3b,
+0x86,
+0x9d,
+0x3d,
+0x7c,
+0x0a,
+0x1b,
+0x46,
+0x28,
+0x58,
+0x2f,
+0xd5,
+0xda,
+0x8f,
+0x10,
+0x9f,
+0xa5,
+0x86,
+0x76,
+0xec,
+0x0f,
+0x07,
+0x16,
+0xf0,
+0xc6,
+0x82,
+0x96,
+0xfa,
+0xfa,
+0xe2,
+0xd3,
+0xf7,
+0x57,
+0xd0,
+0x59,
+0x46,
+0x63,
+0x69,
+0x3e,
+0xb7,
+0xdc,
+0x2e,
+0xaf,
+0x0a,
+0x5d,
+0xb3,
+0xe4,
+0x7a,
+0x1c,
+0x07,
+0x04,
+0x5e,
+0x24,
+0x58,
+0x08,
+0x92,
+0x21,
+0xc3,
+0xd0,
+0xb8,
+0x16,
+0x30,
+0x1d,
+0xd3,
+0x1f,
+0xba,
+0xb1,
+0xc5,
+0xf1,
+0xfd,
+0x32,
+0xfe,
+0x1e,
+0x5e,
+0x07,
+0xb4,
+0xea,
+0x1c,
+0xce,
+0xa0,
+0x7a,
+0x1a,
+0x87,
+0x94,
+0xaa,
+0xbf,
+0xc0,
+0x15,
+0x32,
+0x12,
+0xd4,
+0x49,
+0xf4,
+0xa0,
+0xd9,
+0x2d,
+0xf4,
+0x36,
+0xc2,
+0x67,
+0xb7,
+0x56,
+0x1d,
+0xa4,
+0x1a,
+0xb0,
+0xc8,
+0x84,
+0x89,
+0x38,
+0xf6,
+0x35,
+0x9f,
+0x1a,
+0x47,
+0x81,
+0xc9,
+0x6a,
+0x2a,
+0xb3,
+0x24,
+0x19,
+0x10,
+0xdc,
+0xfd,
+0x78,
+0x7c,
+0xc8,
+0xbf,
+0x85,
+0xe3,
+0xb4,
+0x4a,
+0x9a,
+0x77,
+0x33,
+0x81,
+0x7e,
+0x6d,
+0xa1,
+0x29,
+0xf8,
+0xd5,
+0x7a,
+0x36,
+0x09,
+0xb3,
+0x6b,
+0xa3,
+0x92,
+0x6a,
+0x45,
+0x89,
+0xbf,
+0x79,
+0x8d,
+0xda,
+0x5d,
+0x06,
+0xce,
+0x65,
+0xf2,
+0xd7,
+0x9d,
+0xca,
+0xd7,
+0x0a,
+0x0a,
+0x68,
+0x56,
+0x20,
+0x02,
+0xd2,
+0xa0,
+0x5d,
+0xbc,
+0xb9,
+0x79,
+0xf2,
+0xf9,
+0x5b,
+0xb3,
+0xac,
+0xe7,
+0x72,
+0x55,
+0xda,
+0x94,
+0xf1,
+0x4a,
+0x1f,
+0xea,
+0xbb,
+0x75,
+0x3a,
+0x58,
+0x9c,
+0x7b,
+0xa4,
+0xbc,
+0x67,
+0x23,
+0xeb,
+0xb8,
+0xd1,
+0x59,
+0x1c,
+0x38,
+0xb8,
+0x78,
+0x07,
+0xd8,
+0x5c,
+0xca,
+0x30,
+0xa5,
+0xc8,
+0x8d,
+0xbf,
+0xa2,
+0x78,
+0x6c,
+0x3a,
+0x4b,
+0x3b,
+0x5f,
+0x0c,
+0x39,
+0xbb,
+0x68,
+0x67,
+0xa5,
+0x78,
+0x25,
+0x4e,
+0x35,
+0x81,
+0xc9,
+0xee,
+0xfb,
+0x90,
+0x7a,
+0x14,
+0x9d,
+0xfc,
+0x9a,
+0x1d,
+0x9d,
+0xb7,
+0xae,
+0x90,
+0x69,
+0x79,
+0x89,
+0xba,
+0x9a,
+0x40,
+0x4b,
+0x60,
+0x74,
+0x58,
+0x16,
+0x2c,
+0xc6,
+0xbf,
+0x5d,
+0x16,
+0x46,
+0xe7,
+0x81,
+0x63,
+0x08,
+0xbf,
+0x43,
+0xdf,
+0xaa,
+0x70,
+0xac,
+0x82,
+0x63,
+0x08,
+0x82,
+0x44,
+0x9b,
+0xdd,
+0xce,
+0x90,
+0x0e,
+0xff,
+0x52,
+0xe7,
+0x65,
+0x9a,
+0xdb,
+0x51,
+0x98,
+0xfa,
+0xd4,
+0xc9,
+0x2f,
+0x25,
+0x28,
+0x82,
+0x65,
+0x34,
+0x70,
+0x23,
+0x32,
+0x2b,
+0x53,
+0x59,
+0xaa,
+0x9d,
+0xec,
+0x57,
+0x7d,
+0x26,
+0x59,
+0x2f,
+0xa2,
+0xd4,
+0xf8,
+0x68,
+0x27,
+0x81,
+0xb5,
+0x61,
+0x19,
+0x8e,
+0x24,
+0x97,
+0x7c,
+0x41,
+0xe4,
+0xf7,
+0x53,
+0xf2,
+0xe0,
+0x40,
+0xe3,
+0xff,
+0x29,
+0x3f,
+0x01,
+0x38,
+0x2c,
+0x4e,
+0x4d,
+0xa8,
+0xbe,
+0x5b,
+0x48,
+0x2e,
+0xc9,
+0xf9,
+0x7d,
+0x6d,
+0xed,
+0x14,
+0x6c,
+0x57,
+0x4f,
+0x03,
+0x9e,
+0xb7,
+0xf6,
+0x35,
+0x9a,
+0x92,
+0x73,
+0xf6,
+0xc1,
+0x18,
+0xa6,
+0x3f,
+0xaf,
+0x9e,
+0xca,
+0x0d,
+0x9b,
+0xa6,
+0x25,
+0x3b,
+0x5c,
+0x1b,
+0x6d,
+0x6b,
+0x9e,
+0x58,
+0x9b,
+0xbf,
+0x29,
+0x98,
+0xc5,
+0x00,
+0xc2,
+0xd8,
+0x92,
+0xa6,
+0xdf,
+0xeb,
+0x29,
+0x47,
+0x3a,
+0xc4,
+0xd8,
+0x08,
+0xc7,
+0xbb,
+0xc5,
+0x63,
+0x49,
+0xfe,
+0x7d,
+0x9b,
+0xb3,
+0x74,
+0x73,
+0xfc,
+0xaa,
+0x18,
+0xd3,
+0x01,
+0xfd,
+0xa4,
+0xad,
+0x72,
+0x85,
+0xa5,
+0x58,
+0x13,
+0x5b,
+0x7a,
+0xcd,
+0x3e,
+0x9e,
+0x56,
+0x73,
+0x00,
+0x75,
+0xdf,
+0x2a,
+0xe4,
+0xc9,
+0xff,
+0x17,
+0xfd,
+0x0c,
+0x36,
+0xd9,
+0xa5,
+0xff,
+0xe6,
+0x13,
+0xe6,
+0xcb,
+0x0f,
+0xa4,
+0x80,
+0xa8,
+0x15,
+0xd2,
+0x85,
+0x30,
+0x2c,
+0x4a,
+0xe5,
+0xb5,
+0xf6,
+0xdc,
+0x69,
+0x76,
+0x5f,
+0xbc,
+0x4c,
+0x98,
+0xab,
+0x1d,
+0x71,
+0x56,
+0x84,
+0x49,
+0x1d,
+0x8e,
+0x13,
+0x1e,
+0x85,
+0x80,
+0x5b,
+0x74,
+0xda,
+0xe4,
+0xbd,
+0x38,
+0xf9,
+0x88,
+0x5b,
+0x2a,
+0x43,
+0x45,
+0x5d,
+0x7d,
+0xb6,
+0x40,
+0x48,
+0xd2,
+0xec,
+0x92,
+0x09,
+0xbf,
+0xd4,
+0xdc,
+0xf0,
+0x3e,
+0x36,
+0x7e,
+0xd4,
+0x70,
+0x9b,
+0x80,
+0xfc,
+0x2e,
+0x8f,
+0xa5,
+0x0b,
+0x88,
+0xee,
+0x92,
+0x92,
+0xc0,
+0x8d,
+0x4f,
+0xbc,
+0xdb,
+0xf1,
+0x2c,
+0x60,
+0x9f,
+0xe1,
+0xc6,
+0x1f,
+0x4b,
+0xff,
+0x5d,
+0x13,
+0x7c,
+0x34,
+0x8d,
+0x65,
+0x5d,
+0xec,
+0xc6,
+0xbf,
+0xe8,
+0x82,
+0xb2,
+0x46,
+0x42,
+0x18,
+0x7f,
+0xce,
+0xc8,
+0xc3,
+0x46,
+0xca,
+0xaf,
+0x75,
+0xaf,
+0x3e,
+0xf0,
+0x14,
+0xbb,
+0x33,
+0x70,
+0x42,
+0x9b,
+0x35,
+0xb9,
+0x12,
+0x90,
+0xf4,
+0xa9,
+0xef,
+0xe6,
+0xf6,
+0x8a,
+0xb1,
+0x03,
+0x48,
+0x5a,
+0xb0,
+0x00,
+0x35,
+0xaf,
+0xc4,
+0x66,
+0xa6,
+0x29,
+0x14,
+0x2d,
+0xe6,
+0xfd,
+0x29,
+0x84,
+0xab,
+0x3f,
+0x07,
+0x11,
+0xd8,
+0x4d,
+0x1e,
+0x01,
+0x79,
+0x67,
+0xcd,
+0x3c,
+0xe3,
+0x20,
+0x61,
+0x95,
+0x5f,
+0x4a,
+0x09,
+0xb3,
+0x60,
+0x13,
+0xb6,
+0x30,
+0xba,
+0x1f,
+0xb7,
+0x28,
+0xa8,
+0x28,
+0xd9,
+0xdd,
+0x1f,
+0x48,
+0x7e,
+0x93,
+0xce,
+0x32,
+0x0f,
+0x14,
+0xbc,
+0x76,
+0x0a,
+0x6e,
+0x7b,
+0x94,
+0x7f,
+0xb4,
+0x03,
+0x25,
+0x62,
+0xc2,
+0xe6,
+0xa5,
+0x62,
+0xf5,
+0xb1,
+0x30,
+0x2a,
+0x08,
+0xd6,
+0x13,
+0x82,
+0x65,
+0xfa,
+0x99,
+0xc6,
+0xc8,
+0x8b,
+0x6e,
+0xc5,
+0x1c,
+0x50,
+0xe6,
+0xb9,
+0x0c,
+0xc2,
+0x1b,
+0xf6,
+0x93,
+0xe3,
+0x9a,
+0x35,
+0x6e,
+0x2f,
+0x90,
+0x01,
+0xd3,
+0xf2,
+0x39,
+0x34,
+0x73,
+0x4c,
+0x38,
+0x5c,
+0xce,
+0x33,
+0x3d,
+0x84,
+0x88,
+0xa4,
+0xb9,
+0x7a,
+0x84,
+0x58,
+0x22,
+0xe9,
+0x37,
+0xd5,
+0xfd,
+0x57,
+0x35,
+0x21,
+0x2a,
+0x02,
+0x04,
+0x79,
+0x02,
+0x1c,
+0xd3,
+0x3b,
+0x1d,
+0x31,
+0xf1,
+0x73,
+0x6e,
+0x5b,
+0xb4,
+0x95,
+0xdd,
+0x63,
+0xe5,
+0x1c,
+0xae,
+0xfb,
+0x80,
+0x62,
+0x90,
+0x5e,
+0x27,
+0x1b,
+0x87,
+0xc0,
+0xa1,
+0xf6,
+0xbb,
+0x19,
+0x6a,
+0x6d,
+0xe4,
+0xb3,
+0x6b,
+0xb7,
+0x36,
+0x99,
+0xc9,
+0x83,
+0xaa,
+0xbc,
+0x95,
+0x56,
+0x80,
+0x81,
+0xc4,
+0xcb,
+0x0c,
+0x2c,
+0x01,
+0x69,
+0x12,
+0xf6,
+0xe3,
+0x65,
+0x47,
+0x53,
+0x40,
+0x23,
+0xf2,
+0x40,
+0xf3,
+0x49,
+0x11,
+0xaf,
+0x5e,
+0x18,
+0x6a,
+0x16,
+0x51,
+0x8a,
+0xa9,
+0x41,
+0x2a,
+0x4d,
+0xa7,
+0xd2,
+0xb8,
+0xdb,
+0xc2,
+0x37,
+0x94,
+0x7e,
+0x27,
+0xfb,
+0x57,
+0x22,
+0x28,
+0x36,
+0xe2,
+0x0f,
+0xe6,
+0x10,
+0x51,
+0xd0,
+0x39,
+0xd9,
+0x5f,
+0x35,
+0x5e,
+0x52,
+0xe8,
+0xbe,
+0xb4,
+0x78,
+0x80,
+0xca,
+0xdb,
+0xb5,
+0xe0,
+0x79,
+0x16,
+0x19,
+0x18,
+0x53,
+0x23,
+0xb6,
+0xa3,
+0xee,
+0x6a,
+0xdf,
+0x66,
+0x0e,
+0xd3,
+0x05,
+0x5f,
+0x4a,
+0xe9,
+0x9e,
+0x51,
+0xa9,
+0xf4,
+0xef,
+0x55,
+0xc1,
+0x1a,
+0x28,
+0x6c,
+0xf8,
+0x09,
+0x8b,
+0xb8,
+0xc7,
+0x14,
+0x7c,
+0x86,
+0x8c,
+0xfd,
+0xd4,
+0x92,
+0xd7,
+0x05,
+0x16,
+0x67,
+0x53,
+0x95,
+0x8d,
+0xe0,
+0xe4,
+0x70,
+0x76,
+0x28,
+0xac,
+0x95,
+0x4a,
+0xc6,
+0x66,
+0x6b,
+0xfd,
+0xd2,
+0x86,
+0x70,
+0xc5,
+0x08,
+0x95,
+0x5c,
+0x69,
+0xdb,
+0xca,
+0x85,
+0x3e,
+0xef,
+0xc1,
+0xcb,
+0xb7,
+0x85,
+0x4f,
+0x65,
+0x28,
+0x6a,
+0x29,
+0x6e,
+0xa9,
+0x51,
+0x4f,
+0xcf,
+0x2a,
+0x56,
+0xfc,
+0xc7,
+0x6e,
+0x69,
+0x00,
+0xaf,
+0x50,
+0xb8,
+0x9b,
+0xe1,
+0x02,
+0xa7,
+0xab,
+0x25,
+0xf1,
+0xbd,
+0x55,
+0x78,
+0x60,
+0x6d,
+0x37,
+0x1d,
+0x7c,
+0xdb,
+0xbc,
+0x0f,
+0x70,
+0xca,
+0xb8,
+0x15,
+0x1f,
+0x26,
+0x12,
+0x0b,
+0x56,
+0x7c,
+0xb2,
+0x48,
+0xa7,
+0xee,
+0x73,
+0x1a,
+0xb5,
+0xb9,
+0x2a,
+0xfb,
+0x37,
+0x04,
+0x4e,
+0x44,
+0x8f,
+0x8c,
+0xc8,
+0xd1,
+0x2b,
+0x83,
+0x6a,
+0xee,
+0xf7,
+0x8a,
+0x2c,
+0x9e,
+0x48,
+0xe5,
+0xf3,
+0xad,
+0xdf,
+0xf2,
+0x9f,
+0x31,
+0xb5,
+0x4c,
+0xb8,
+0x91,
+0x79,
+0xb6,
+0xc0,
+0x92,
+0x42,
+0x01,
+0x05,
+0x63,
+0xa7,
+0xbd,
+0x2b,
+0x31,
+0x43,
+0x07,
+0xfe,
+0x1c,
+0x40,
+0xec,
+0xc9,
+0x24,
+0x0b,
+0x7d,
+0x01,
+0xaa,
+0x43,
+0x0a,
+0x7e,
+0xae,
+0xbf,
+0xb5,
+0x74,
+0x3a,
+0x91,
+0x21,
+0x23,
+0x35,
+0xa0,
+0xeb,
+0xd6,
+0x65,
+0x35,
+0xce,
+0x81,
+0x83,
+0x4d,
+0x46,
+0x70,
+0x44,
+0x33,
+0x9e,
+0x37,
+0xa0,
+0xe0,
+0x76,
+0x9f,
+0x02,
+0x4e,
+0xcd,
+0x00,
+0xb0,
+0x4c,
+0x9c,
+0xf1,
+0x33,
+0x4b,
+0xa0,
+0xc5,
+0x88,
+0x46,
+0x03,
+0xa0,
+0x99,
+0x81,
+0xb0,
+0x43,
+0x02,
+0xb4,
+0xe9,
+0x45,
+0xa8,
+0x37,
+0x85,
+0x64,
+0x7e,
+0xf9,
+0x9a,
+0x8f,
+0xdd,
+0x3f,
+0xb7,
+0xea,
+0xe5,
+0x1d,
+0x9d,
+0x4f,
+0x5e,
+0x80,
+0x14,
+0xe1,
+0xcc,
+0xd5,
+0x2e,
+0x57,
+0xab,
+0x79,
+0x0c,
+0xb5,
+0xc5,
+0x71,
+0x18,
+0xc0,
+0x80,
+0xe4,
+0x87,
+0x7e,
+0xbc,
+0x5c,
+0x43,
+0x8a,
+0x7a,
+0x7b,
+0xbe,
+0x40,
+0x76,
+0xf7,
+0x04,
+0x23,
+0xf3,
+0xd6,
+0x76,
+0xd1,
+0x82,
+0x93,
+0x04,
+0xe1,
+0x29,
+0xf7,
+0x60,
+0xd3,
+0x6d,
+0xee,
+0x85,
+0xc9,
+0x23,
+0xfe,
+0x9a,
+0x93,
+0xe6,
+0x61,
+0xbd,
+0x3d,
+0x3d,
+0x7a,
+0xf8,
+0x48,
+0x1c,
+0xc3,
+0x6c,
+0x61,
+0x42,
+0x66,
+0x9b,
+0x70,
+0x56,
+0xdc,
+0x97,
+0xcd,
+0x75,
+0xb3,
+0x83,
+0x40,
+0xfe,
+0x2b,
+0x81,
+0xa6,
+0x12,
+0xd3,
+0x84,
+0x0b,
+0x9f,
+0x13,
+0x65,
+0x91,
+0xb0,
+0x3e,
+0x75,
+0x13,
+0x0c,
+0xd5,
+0x62,
+0x81,
+0xed,
+0x06,
+0xf8,
+0x96,
+0xf9,
+0x48,
+0xea,
+0x98,
+0x12,
+0x4a,
+0x6e,
+0xf3,
+0x57,
+0xf4,
+0x09,
+0x74,
+0x31,
+0xda,
+0x98,
+0x1e,
+0xdb,
+0x43,
+0x90,
+0xb6,
+0xdb,
+0x78,
+0x88,
+0xdd,
+0x39,
+0xec,
+0xf2,
+0xe5,
+0xc2,
+0x37,
+0x55,
+0x9f,
+0xbb,
+0xb4,
+0x6c,
+0x1e,
+0xd8,
+0x9e,
+0x46,
+0x79,
+0xa7,
+0xa1,
+0x89,
+0x41,
+0xe0,
+0xf6,
+0xa6,
+0x1c,
+0xf2,
+0x12,
+0x0d,
+0xee,
+0xa2,
+0x3b,
+0xcc,
+0x35,
+0x97,
+0x44,
+0x6b,
+0x56,
+0xac,
+0xb6,
+0xdf,
+0x2a,
+0x70,
+0xde,
+0xf2,
+0xf5,
+0xcc,
+0xa5,
+0xc5,
+0xb8,
+0x3b,
+0x32,
+0x62,
+0xcc,
+0xb3,
+0xc9,
+0x5f,
+0x32,
+0x6a,
+0xe5,
+0x85,
+0x03,
+0x2b,
+0xe2,
+0x73,
+0x4f,
+0x00,
+0xba,
+0x21,
+0x6e,
+0x51,
+0xc6,
+0x52,
+0x5c,
+0x34,
+0x16,
+0x58,
+0x5c,
+0xb2,
+0xa5,
+0x70,
+0x84,
+0xbd,
+0xa8,
+0x25,
+0x7f,
+0x04,
+0x6b,
+0xd2,
+0xf8,
+0x35,
+0xa1,
+0xc2,
+0x3b,
+0x52,
+0xce,
+0x29,
+0x8e,
+0x7a,
+0x5f,
+0x5b,
+0x04,
+0x3b,
+0x0b,
+0xc1,
+0x0a,
+0x25,
+0xa8,
+0xd1,
+0x59,
+0x83,
+0x03,
+0x90,
+0x42,
+0x74,
+0xcb,
+0xc8,
+0x5b,
+0x7f,
+0x6a,
+0x2b,
+0xe5,
+0x6a,
+0x56,
+0x77,
+0xe2,
+0x96,
+0xce,
+0xf3,
+0x99,
+0x36,
+0xa5,
+0x8b,
+0x2f,
+0xf5,
+0x17,
+0xf7,
+0x92,
+0x6c,
+0xf5,
+0x62,
+0xd8,
+0x25,
+0x0f,
+0xdb,
+0xfc,
+0x5e,
+0x08,
+0x4b,
+0x27,
+0x13,
+0x2a,
+0x9e,
+0x4c,
+0xb4,
+0xb9,
+0xc8,
+0x62,
+0x4e,
+0xb3,
+0x13,
+0x0d,
+0xc1,
+0x2b,
+0x8e,
+0xff,
+0x8e,
+0x52,
+0x2a,
+0x52,
+0x95,
+0x5a,
+0x09,
+0xd8,
+0x1e,
+0x29,
+0x75,
+0x49,
+0x54,
+0xf3,
+0x30,
+0xd0,
+0x2b,
+0x03,
+0xaf,
+0x53,
+0x6c,
+0xbd,
+0x8a,
+0xdc,
+0x44,
+0x96,
+0x68,
+0x51,
+0xf8,
+0xd8,
+0x7a,
+0xf5,
+0x35,
+0xc3,
+0x11,
+0x5c,
+0x5e,
+0x9a,
+0x70,
+0xf4,
+0xac,
+0x5c,
+0xee,
+0x8d,
+0x68,
+0x53,
+0x0a,
+0xb7,
+0x7b,
+0x16,
+0x6b,
+0xb8,
+0x08,
+0xf0,
+0xd6,
+0xda,
+0x29,
+0x17,
+0xf2,
+0xf3,
+0x35,
+0x8e,
+0x0d,
+0x03,
+0x98,
+0xe8,
+0x53,
+0x33,
+0x84,
+0x79,
+0xa3,
+0x6c,
+0x12,
+0xc5,
+0x76,
+0x04,
+0xd5,
+0x5f,
+0x58,
+0xd5,
+0xac,
+0x57,
+0x5f,
+0xfe,
+0xd0,
+0x22,
+0xf3,
+0x26,
+0x07,
+0x99,
+0x1c,
+0xe8,
+0xa0,
+0x32,
+0x1d,
+0xdb,
+0x0f,
+0xcb,
+0x10,
+0x85,
+0x6f,
+0x22,
+0x7a,
+0x05,
+0x66,
+0x91,
+0x2f,
+0x58,
+0xa0,
+0xc4,
+0xa6,
+0x1c,
+0xa9,
+0x16,
+0x07,
+0x84,
+0x82,
+0x06,
+0xc0,
+0x65,
+0xae,
+0x78,
+0xf2,
+0xcf,
+0xa3,
+0x9e,
+0xf0,
+0xe0,
+0x2e,
+0x92,
+0x6e,
+0xb8,
+0xa6,
+0x77,
+0x59,
+0x01,
+0x8d,
+0xac,
+0xfa,
+0x5e,
+0x29,
+0xd0,
+0xda,
+0x5e,
+0xeb,
+0x8a,
+0x5e,
+0xd6,
+0x9d,
+0x6f,
+0xda,
+0x7d,
+0x25,
+0x30,
+0xf9,
+0xc0,
+0xee,
+0xdf,
+0x14,
+0x6d,
+0x07,
+0x09,
+0x0a,
+0x3f,
+0xfb,
+0x23,
+0x40,
+0x19,
+0xe6,
+0xd2,
+0x3c,
+0xd0,
+0x69,
+0x00,
+0xf6,
+0xaa,
+0x4b,
+0xf7,
+0x0e,
+0xc0,
+0x0e,
+0x4d,
+0x81,
+0x01,
+0x44,
+0xad,
+0xf4,
+0x33,
+0x83,
+0xc5,
+0x45,
+0x3f,
+0x33,
+0xea,
+0x5c,
+0x8d,
+0xa7,
+0xcd,
+0xef,
+0x5c,
+0x40,
+0xbc,
+0xcf,
+0x9b,
+0x04,
+0x30,
+0x12,
+0x90,
+0x83,
+0xbd,
+0xc8,
+0x1a,
+0x9a,
+0x8c,
+0xc0,
+0x56,
+0x39,
+0xc1,
+0x7c,
+0xd8,
+0x66,
+0xa8,
+0x2c,
+0x27,
+0x4d,
+0xe4,
+0xa3,
+0x84,
+0xf6,
+0xae,
+0xbd,
+0x41,
+0x69,
+0xa2,
+0xc3,
+0x69,
+0xd3,
+0x34,
+0x06,
+0x3a,
+0x34,
+0x15,
+0x61,
+0xa5,
+0xb1,
+0xae,
+0xd0,
+0xc5,
+0x01,
+0x28,
+0x82,
+0x7b,
+0x67,
+0x48,
+0x5b,
+0x71,
+0xb8,
+0xe3,
+0x6b,
+0xe3,
+0x51,
+0x35,
+0xac,
+0x18,
+0x4b,
+0x7a,
+0x7e,
+0x88,
+0xc8,
+0x6f,
+0x8d,
+0xe0,
+0x68,
+0x53,
+0x70,
+0x4f,
+0x0a,
+0xa8,
+0x3e,
+0x44,
+0xea,
+0xd2,
+0x9e,
+0x64,
+0xb7,
+0xca,
+0x50,
+0xc5,
+0x08,
+0x46,
+0xc7,
+0x21,
+0x7f,
+0x58,
+0x8d,
+0x3e,
+0x56,
+0x8b,
+0xa9,
+0x4d,
+0x81,
+0x95,
+0xe2,
+0x20,
+0xc6,
+0xd0,
+0x2d,
+0x68,
+0x5e,
+0x99,
+0xbc,
+0x97,
+0x13,
+0xbf,
+0xd6,
+0x48,
+0x53,
+0x7e,
+0x41,
+0x96,
+0x3b,
+0x44,
+0x34,
+0xea,
+0xf6,
+0xbe,
+0x8a,
+0x04,
+0xdf,
+0x40,
+0xa5,
+0x54,
+0x37,
+0xec,
+0x9e,
+0x2f,
+0x37,
+0x52,
+0xa6,
+0xcd,
+0xdd,
+0xc8,
+0x6f,
+0xdd,
+0xc4,
+0x04,
+0x03,
+0xa1,
+0xe7,
+0xb5,
+0xa4,
+0x1c,
+0x7a,
+0xc8,
+0x0b,
+0x8e,
+0x32,
+0x6b,
+0x6f,
+0xc7,
+0x5b,
+0x7a,
+0x79,
+0xe5,
+0xf5,
+0x48,
+0x16,
+0x75,
+0xd8,
+0xb4,
+0x27,
+0xc4,
+0x92,
+0xc3,
+0xcb,
+0x67,
+0xf4,
+0x46,
+0xcf,
+0xed,
+0x2e,
+0x38,
+0x8c,
+0x9d,
+0xab,
+0x78,
+0x80,
+0xc9,
+0x15,
+0x57,
+0x77,
+0x3e,
+0x6d,
+0xf7,
+0xdc,
+0x4f,
+0x3d,
+0xd5,
+0x10,
+0x7e,
+0x64,
+0x2c,
+0xdd,
+0x34,
+0xb1,
+0x8d,
+0xde,
+0x59,
+0x54,
+0x5a,
+0x20,
+0xcb,
+0x7f,
+0xe2,
+0x67,
+0x3d,
+0xe3,
+0x21,
+0xd0,
+0x36,
+0xb3,
+0x10,
+0xdc,
+0x78,
+0xd5,
+0xa1,
+0x64,
+0x69,
+0x75,
+0x8d,
+0xce,
+0x6e,
+0xab,
+0xa7,
+0x79,
+0x8c,
+0xfb,
+0x63,
+0x57,
+0xaf,
+0x9b,
+0xd0,
+0x9b,
+0x6b,
+0x82,
+0xf1,
+0x18,
+0x87,
+0xc1,
+0x8a,
+0x93,
+0xd4,
+0xf8,
+0xfa,
+0xad,
+0xe4,
+0xfd,
+0x73,
+0xfc,
+0x84,
+0x41,
+0x49,
+0x21,
+0x53,
+0x22,
+0xdf,
+0xd9,
+0x06,
+0xd0,
+0xf4,
+0xa1,
+0x1b,
+0x04,
+0xe5,
+0x08,
+0xf3,
+0x3f,
+0x2c,
+0x1c,
+0x73,
+0x07,
+0x0a,
+0x82,
+0xdc,
+0xb9,
+0x0b,
+0x93,
+0xdd,
+0x22,
+0xdb,
+0xe7,
+0x8d,
+0xf7,
+0xe1,
+0xb9,
+0x55,
+0xa4,
+0xeb,
+0xd2,
+0x19,
+0x4b,
+0x7c,
+0x27,
+0x14,
+0x65,
+0x5c,
+0xaa,
+0x31,
+0x6f,
+0x7c,
+0xe8,
+0xfc,
+0x9d,
+0xce,
+0x6d,
+0x0a,
+0x76,
+0xa9,
+0x67,
+0xab,
+0x74,
+0xdc,
+0x75,
+0x31,
+0x58,
+0x0d,
+0x55,
+0xb9,
+0xe3,
+0x01,
+0x96,
+0xf8,
+0xcb,
+0x9b,
+0xb7,
+0x0f,
+0x80,
+0x8f,
+0x6c,
+0x65,
+0xc9,
+0xe1,
+0x4a,
+0x4b,
+0x8d,
+0x36,
+0xd9,
+0x9a,
+0xdd,
+0xe4,
+0xad,
+0x00,
+0x7c,
+0x47,
+0xb5,
+0xb5,
+0x8c,
+0x34,
+0x6a,
+0x50,
+0xb7,
+0x88,
+0xa1,
+0x44,
+0xae,
+0xb3,
+0x84,
+0x13,
+0x70,
+0xa9,
+0x42,
+0x5f,
+0xa8,
+0xb5,
+0xcb,
+0x4f,
+0xe9,
+0x40,
+0x53,
+0x23,
+0x59,
+0x51,
+0xc0,
+0xf1,
+0xc0,
+0x41,
+0xac,
+0x7b,
+0xf9,
+0xbb,
+0x91,
+0xbe,
+0xb5,
+0xfb,
+0x53,
+0x9e,
+0xf4,
+0xe9,
+0xad,
+0xa1,
+0x62,
+0x83,
+0x1b,
+0x24,
+0xb8,
+0x48,
+0xbb,
+0x70,
+0x29,
+0x11,
+0x11,
+0xf1,
+0x08,
+0x1d,
+0x46,
+0x7a,
+0xfb,
+0x52,
+0x3d,
+0x97,
+0x60,
+0xcc,
+0xb1,
+0xed,
+0xa2,
+0xb8,
+0xda,
+0x9e,
+0x44,
+0x5d,
+0x95,
+0x53,
+0x2b,
+0xc4,
+0x8a,
+0x4c,
+0xdd,
+0xb3,
+0xe2,
+0x06,
+0xa6,
+0x81,
+0x36,
+0xf2,
+0x36,
+0xfd,
+0x7d,
+0x46,
+0x20,
+0x31,
+0x73,
+0xc8,
+0xbc,
+0x85,
+0xa2,
+0x89,
+0x74,
+0x56,
+0xaf,
+0x76,
+0xff,
+0xea,
+0xae,
+0xe3,
+0x34,
+0xbc,
+0x72,
+0xfe,
+0x35,
+0xf1,
+0x69,
+0x43,
+0x6a,
+0x83,
+0x25,
+0x3a,
+0x97,
+0x33,
+0xc6,
+0x35,
+0x84,
+0xa5,
+0x6b,
+0x58,
+0x0c,
+0xdc,
+0xd2,
+0xd6,
+0x43,
+0x0d,
+0x43,
+0x77,
+0x11,
+0xc0,
+0x70,
+0x67,
+0x48,
+0x2d,
+0x26,
+0x45,
+0xfc,
+0xe6,
+0x0f,
+0x33,
+0x2a,
+0xe7,
+0xf3,
+0xb6,
+0x63,
+0xf8,
+0xc5,
+0xad,
+0x0b,
+0x0e,
+0x91,
+0x9e,
+0x61,
+0x8a,
+0x08,
+0xde,
+0x4b,
+0x02,
+0x8a,
+0x96,
+0xa1,
+0xb7,
+0x95,
+0xda,
+0x45,
+0xd2,
+0x17,
+0x3c,
+0xf4,
+0xa4,
+0xc6,
+0x35,
+0x52,
+0xc4,
+0x38,
+0x76,
+0x92,
+0x66,
+0x33,
+0x77,
+0x6f,
+0x41,
+0xdc,
+0xc0,
+0x02,
+0x0d,
+0x48,
+0x0a,
+0xfd,
+0x9e,
+0x0e,
+0x42,
+0x1b,
+0xe3,
+0xf6,
+0xb1,
+0xe7,
+0x14,
+0xc8,
+0x58,
+0xff,
+0xdb,
+0xe2,
+0x5f,
+0x08,
+0xba,
+0x98,
+0xcc,
+0x69,
+0x06,
+0xee,
+0x4b,
+0x9c,
+0x4d,
+0xfe,
+0x5f,
+0xb4,
+0x3d,
+0x71,
+0x8f,
+0x90,
+0x63,
+0xb5,
+0xd5,
+0x82,
+0x96,
+0xcc,
+0x60,
+0x5d,
+0x5f,
+0x3e,
+0x78,
+0x50,
+0x89,
+0x1f,
+0xb9,
+0x26,
+0xbf,
+0x40,
+0xa2,
+0x58,
+0xae,
+0xe5,
+0x44,
+0xd3,
+0x4f,
+0x6a,
+0x5a,
+0x9f,
+0x62,
+0x06,
+0x63,
+0x2f,
+0x26,
+0xf3,
+0x6d,
+0x07,
+0xba,
+0xc5,
+0x44,
+0x8b,
+0x03,
+0x10,
+0x9b,
+0xfd,
+0x4a,
+0x82,
+0x45,
+0x83,
+0x80,
+0x89,
+0xeb,
+0x71,
+0x9d,
+0x7b,
+0xa8,
+0x97,
+0xef,
+0xac,
+0x63,
+0x8e,
+0x6a,
+0x34,
+0xca,
+0x18,
+0x82,
+0x82,
+0x8b,
+0xcb,
+0xa2,
+0xe6,
+0x1b,
+0x78,
+0xf7,
+0xb0,
+0x86,
+0xaa,
+0xec,
+0x30,
+0xcf,
+0x31,
+0x67,
+0x80,
+0x54,
+0x45,
+0x25,
+0x37,
+0x26,
+0x20,
+0xcc,
+0xcd,
+0xf9,
+0x01,
+0x7a,
+0xcc,
+0xcf,
+0x9b,
+0x72,
+0x90,
+0x6d,
+0xf6,
+0x1c,
+0x2b,
+0xaa,
+0x9b,
+0x9d,
+0xb0,
+0x03,
+0x00,
+0xcc,
+0xd0,
+0x62,
+0x6f,
+0xa9,
+0x08,
+0x9f,
+0xca,
+0x7a,
+0x20,
+0x0c,
+0xdd,
+0x5d,
+0x0f,
+0x84,
+0xc1,
+0x38,
+0x6f,
+0xc7,
+0x74,
+0xe9,
+0x88,
+0x03,
+0xe1,
+0x5a,
+0x99,
+0x92,
+0xb3,
+0xe8,
+0x68,
+0x28,
+0xac,
+0x10,
+0x67,
+0x76,
+0xda,
+0xfc,
+0x96,
+0xd7,
+0xc5,
+0xfc,
+0x00,
+0xb5,
+0xeb,
+0x03,
+0x28,
+0x15,
+0x03,
+0x04,
+0x5d,
+0xf4,
+0x1b,
+0xdc,
+0x9c,
+0xc1,
+0xa4,
+0x13,
+0xff,
+0xaf,
+0x34,
+0x01,
+0x64,
+0x0b,
+0xe9,
+0x75,
+0xac,
+0x0a,
+0x5f,
+0x3c,
+0x87,
+0x0f,
+0xb5,
+0xe2,
+0x3f,
+0x35,
+0x71,
+0xf4,
+0xf1,
+0x93,
+0x88,
+0x5e,
+0x9c,
+0xdd,
+0x90,
+0x8e,
+0x79,
+0x60,
+0x75,
+0x47,
+0x71,
+0x6a,
+0x0a,
+0xa5,
+0x35,
+0x4c,
+0x0e,
+0xa0,
+0x29,
+0x21,
+0x83,
+0x4a,
+0xf4,
+0x5e,
+0xac,
+0x65,
+0xab,
+0x89,
+0x9a,
+0x41,
+0x4f,
+0x14,
+0x86,
+0x7b,
+0xe4,
+0xe9,
+0xf6,
+0xe4,
+0x83,
+0xda,
+0x2a,
+0x04,
+0x51,
+0xf2,
+0xd4,
+0xae,
+0xd0,
+0x7f,
+0x9f,
+0xbc,
+0xcd,
+0xa5,
+0xe1,
+0xcb,
+0x20,
+0xd4,
+0x4d,
+0x05,
+0x54,
+0x66,
+0x5d,
+0x35,
+0xce,
+0x5a,
+0xe1,
+0xfb,
+0xe7,
+0xbd,
+0xf6,
+0x01,
+0x34,
+0x48,
+0xe7,
+0x60,
+0x73,
+0x58,
+0xaf,
+0x55,
+0xf6,
+0xce,
+0x16,
+0xd4,
+0xf3,
+0x9b,
+0xaa,
+0xe8,
+0xd7,
+0x6c,
+0xea,
+0x1b,
+0x4d,
+0x84,
+0x1f,
+0x4e,
+0x0c,
+0x44,
+0x82,
+0x72,
+0x32,
+0x22,
+0xce,
+0xc2,
+0xfd,
+0xba,
+0x46,
+0x3e,
+0x73,
+0x28,
+0x5f,
+0xa6,
+0x19,
+0x81,
+0xe6,
+0x45,
+0xed,
+0x01,
+0xea,
+0x74,
+0xcd,
+0x64,
+0xa9,
+0x5a,
+0x31,
+0x61,
+0x82,
+0xa1,
+0x9f,
+0x30,
+0x9d,
+0x0c,
+0xa8,
+0x57,
+0x1c,
+0xd5,
+0x50,
+0x4e,
+0x96,
+0x4b,
+0x8c,
+0xa7,
+0xff,
+0x59,
+0x04,
+0x29,
+0x17,
+0x3c,
+0xec,
+0x73,
+0xfd,
+0x81,
+0x5e,
+0x24,
+0xd0,
+0xf7,
+0xd4,
+0x8f,
+0xfc,
+0xa2,
+0x9f,
+0x21,
+0x06,
+0xf1,
+0x4b,
+0x2e,
+0x2c,
+0x3d,
+0x2b,
+0x86,
+0x81,
+0x3f,
+0xa7,
+0xdc,
+0xa7,
+0xf0,
+0x1d,
+0xd0,
+0xaa,
+0xbb,
+0x96,
+0x52,
+0xae,
+0xc9,
+0xda,
+0x22,
+0x83,
+0xc5,
+0xfc,
+0x0d,
+0x1c,
+0xa6,
+0x49,
+0x08,
+0x74,
+0x97,
+0x65,
+0xdb,
+0x23,
+0xf8,
+0x2c,
+0x57,
+0xa9,
+0x60,
+0x82,
+0x11,
+0xc7,
+0xa9,
+0x86,
+0x57,
+0x1f,
+0xb3,
+0xc8,
+0xd8,
+0x51,
+0x99,
+0x3d,
+0x32,
+0x7f,
+0x83,
+0x81,
+0xf6,
+0x64,
+0x7b,
+0x29,
+0x81,
+0x65,
+0x48,
+0x1d,
+0x28,
+0x8c,
+0x99,
+0xf7,
+0xe8,
+0xf4,
+0xfa,
+0xd9,
+0x92,
+0x0c,
+0xad,
+0xb3,
+0x73,
+0x00,
+0x01,
+0x62,
+0x2c,
+0x0c,
+0xb6,
+0x73,
+0xff,
+0xb7,
+0xf0,
+0x8a,
+0x6c,
+0x56,
+0x7f,
+0xa0,
+0x23,
+0xf3,
+0x28,
+0xf1,
+0x49,
+0x1c,
+0x9a,
+0x50,
+0xd6,
+0x37,
+0x58,
+0xfe,
+0x1d,
+0x94,
+0xf5,
+0x95,
+0x89,
+0x65,
+0xed,
+0x3e,
+0x47,
+0x33,
+0xa6,
+0x36,
+0xdc,
+0x69,
+0xc7,
+0x85,
+0x0b,
+0xa2,
+0x35,
+0x17,
+0x01,
+0xa6,
+0x7b,
+0x31,
+0xd7,
+0xda,
+0xed,
+0x18,
+0xdf,
+0x07,
+0x4a,
+0xd2,
+0x4a,
+0xa1,
+0x7f,
+0x8a,
+0xca,
+0xe7,
+0x8a,
+0x65,
+0xb8,
+0x3d,
+0x88,
+0x9e,
+0x61,
+0x3b,
+0xda,
+0x05,
+0x30,
+0xae,
+0x9f,
+0x83,
+0x16,
+0x7f,
+0xc9,
+0x75,
+0xdd,
+0x0a,
+0x17,
+0xea,
+0xc4,
+0x27,
+0xe8,
+0x21,
+0x42,
+0x48,
+0x2d,
+0x91,
+0xd4,
+0xd8,
+0x59,
+0x03,
+0x50,
+0xb9,
+0xbc,
+0xc9,
+0x55,
+0x3d,
+0xc2,
+0x4b,
+0x5f,
+0xe3,
+0xca,
+0x28,
+0xe4,
+0x63,
+0x69,
+0xfd,
+0xa5,
+0xf6,
+0x1b,
+0xb2,
+0x01,
+0xb3,
+0x76,
+0x54,
+0x95,
+0xe2,
+0x9d,
+0xd4,
+0x4b,
+0xa6,
+0x00,
+0xb4,
+0xe8,
+0x41,
+0x80,
+0xe0,
+0x0b,
+0xfd,
+0xe6,
+0xbf,
+0x0a,
+0xb8,
+0xba,
+0xd2,
+0x57,
+0xff,
+0xc7,
+0x77,
+0xe4,
+0x13,
+0xd9,
+0x75,
+0x8f,
+0xee,
+0x21,
+0x66,
+0x21,
+0xd6,
+0xb9,
+0x1b,
+0x42,
+0x93,
+0xe6,
+0x27,
+0x38,
+0xcb,
+0x10,
+0x11,
+0xdc,
+0xe9,
+0xfd,
+0x5a,
+0xf8,
+0x00,
+0xc2,
+0x22,
+0xa7,
+0x95,
+0xa7,
+0xe9,
+0x78,
+0x3d,
+0x31,
+0x29,
+0x7d,
+0xfa,
+0x9a,
+0x6f,
+0x5a,
+0xe4,
+0xb6,
+0x9d,
+0x45,
+0x30,
+0xbb,
+0x7f,
+0xff,
+0xca,
+0x56,
+0xb2,
+0xfd,
+0x20,
+0x57,
+0xa9,
+0xee,
+0x2b,
+0x9e,
+0xd7,
+0xfc,
+0x53,
+0x93,
+0x1a,
+0x60,
+0x91,
+0xf6,
+0xf7,
+0x0d,
+0x7d,
+0xf0,
+0x04,
+0x02,
+0xd1,
+0x8e,
+0x26,
+0x56,
+0x1e,
+0xda,
+0xcc,
+0x55,
+0xc3,
+0x30,
+0x1b,
+0x1c,
+0xad,
+0xa8,
+0xa6,
+0x71,
+0xf2,
+0x00,
+0x94,
+0xfe,
+0x17,
+0xf9,
+0x04,
+0x08,
+0xdf,
+0xb2,
+0x07,
+0x10,
+0x33,
+0x94,
+0x28,
+0xe7,
+0x9a,
+0x43,
+0x1e,
+0xad,
+0x6f,
+0x49,
+0xfd,
+0xd1,
+0xec,
+0xf9,
+0x7b,
+0x40,
+0x8a,
+0x9a,
+0x49,
+0xbc,
+0xfe,
+0x6b,
+0xb4,
+0xce,
+0x0f,
+0xf3,
+0x85,
+0xad,
+0x90,
+0xb7,
+0xf1,
+0x6e,
+0x16,
+0x41,
+0xb5,
+0xca,
+0x5b,
+0x1a,
+0x8b,
+0x74,
+0xa7,
+0x94,
+0xdc,
+0x8f,
+0x21,
+0xbd,
+0x68,
+0xbf,
+0xf4,
+0x8f,
+0x67,
+0x2b,
+0x3a,
+0xce,
+0x97,
+0x8f,
+0x55,
+0xa5,
+0xf6,
+0xf1,
+0xf2,
+0x6a,
+0xed,
+0x82,
+0xe5,
+0x91,
+0xab,
+0xf1,
+0xb3,
+0xbe,
+0x9f,
+0x9a,
+0x99,
+0x79,
+0x4a,
+0x25,
+0x9f,
+0x99,
+0xad,
+0x16,
+0x0f,
+0x5c,
+0x0e,
+0x66,
+0x9d,
+0x4c,
+0x58,
+0x21,
+0xcf,
+0xcc,
+0x19,
+0x09,
+0x46,
+0x76,
+0x0b,
+0x93,
+0x46,
+0x34,
+0x85,
+0xee,
+0x58,
+0x70,
+0x24,
+0x45,
+0xc1,
+0xed,
+0x0f,
+0xed,
+0x1e,
+0xb8,
+0xc9,
+0xf3,
+0x22,
+0x90,
+0xe2,
+0x26,
+0x2f,
+0x15,
+0x19,
+0xf2,
+0xb8,
+0x10,
+0x26,
+0x0b,
+0x35,
+0xde,
+0xda,
+0x29,
+0x8f,
+0x45,
+0x46,
+0xec,
+0x00,
+0xe6,
+0x03,
+0x4e,
+0xb9,
+0x54,
+0x3a,
+0x9d,
+0x15,
+0xd7,
+0x27,
+0xd6,
+0x04,
+0x0c,
+0x3e,
+0x3e,
+0xf7,
+0xcf,
+0xcc,
+0x9b,
+0x1d,
+0xa5,
+0x93,
+0xf8,
+0xd2,
+0xe3,
+0x1b,
+0x7c,
+0xce,
+0xfc,
+0x5f,
+0x0a,
+0x35,
+0xa9,
+0x99,
+0x37,
+0x83,
+0xec,
+0x2d,
+0x7a,
+0x6d,
+0x07,
+0x76,
+0x35,
+0x02,
+0xfd,
+0x3c,
+0xa8,
+0x9e,
+0x47,
+0x2b,
+0xcc,
+0x78,
+0x52,
+0xd3,
+0x8d,
+0x57,
+0xb6,
+0x79,
+0x3c,
+0x09,
+0x33,
+0x14,
+0x61,
+0x47,
+0x84,
+0xd9,
+0x3d,
+0x63,
+0xe8,
+0xdb,
+0x29,
+0xb6,
+0x45,
+0xe1,
+0x9c,
+0x54,
+0x0d,
+0x23,
+0xf7,
+0x8d,
+0x00,
+0x9a,
+0x03,
+0x4b,
+0xaf,
+0x19,
+0xf5,
+0x45,
+0xa5,
+0x97,
+0x5d,
+0x47,
+0x78,
+0x8a,
+0x99,
+0x0d,
+0x40,
+0xac,
+0xf6,
+0xa5,
+0xf5,
+0x3b,
+0x4e,
+0xdc,
+0xd9,
+0x36,
+0xed,
+0xd0,
+0x38,
+0xee,
+0xa0,
+0x24,
+0x71,
+0xbc,
+0x69,
+0x6d,
+0xa4,
+0x36,
+0xe8,
+0xe9,
+0xb7,
+0x21,
+0xd1,
+0xb9,
+0xe1,
+0xd6,
+0xa7,
+0x4d,
+0xf8,
+0x7b,
+0x12,
+0x63,
+0x56,
+0x8b,
+0x21,
+0xdd,
+0x05,
+0x34,
+0xed,
+0x7c,
+0xc9,
+0x04,
+0x75,
+0xfd,
+0xd0,
+0x38,
+0xf0,
+0xa7,
+0x23,
+0x06,
+0x77,
+0x7c,
+0xd5,
+0xc2,
+0x46,
+0xd6,
+0xe4,
+0x03,
+0x5b,
+0x13,
+0x64,
+0x6b,
+0xf9,
+0xac,
+0x5e,
+0x6e,
+0xd5,
+0x55,
+0xca,
+0xfe,
+0x5f,
+0x5e,
+0x44,
+0xa3,
+0x88,
+0x17,
+0xdb,
+0x36,
+0x93,
+0xe6,
+0x02,
+0x02,
+0x89,
+0x39,
+0x86,
+0x3f,
+0x5e,
+0x7e,
+0x91,
+0xff,
+0xca,
+0xbe,
+0x95,
+0xbd,
+0x06,
+0xe4,
+0xf9,
+0x5f,
+0x99,
+0xa9,
+0x1a,
+0xc0,
+0x5d,
+0xd1,
+0x29,
+0x38,
+0x77,
+0xb4,
+0x00,
+0xf3,
+0x57,
+0x16,
+0x34,
+0x52,
+0x04,
+0xe7,
+0xf4,
+0x6c,
+0xf4,
+0xb1,
+0x75,
+0x6d,
+0x7a,
+0x63,
+0x14,
+0x5b,
+0x9f,
+0x56,
+0x69,
+0x6b,
+0x13,
+0x8e,
+0x76,
+0x9c,
+0x36,
+0xa6,
+0x2f,
+0xd5,
+0x57,
+0xcf,
+0x20,
+0xbc,
+0x78,
+0x0a,
+0x59,
+0x0f,
+0xdc,
+0xf4,
+0xde,
+0x5d,
+0xf1,
+0x99,
+0x91,
+0xd8,
+0xf8,
+0x82,
+0x90,
+0x28,
+0xe3,
+0x2e,
+0x75,
+0x3d,
+0xf1,
+0x10,
+0xf6,
+0xe8,
+0xfb,
+0x09,
+0x84,
+0x7a,
+0xf0,
+0xab,
+0xa4,
+0x08,
+0xf5,
+0xfb,
+0x62,
+0xfe,
+0x6a,
+0x70,
+0x79,
+0x52,
+0x97,
+0xea,
+0x13,
+0xf0,
+0x3b,
+0xe1,
+0xf6,
+0xab,
+0x64,
+0x89,
+0x23,
+0x34,
+0x20,
+0x69,
+0xd2,
+0xf5,
+0x8d,
+0x0b,
+0x62,
+0x5e,
+0x04,
+0x1c,
+0xf1,
+0x2b,
+0xe9,
+0x79,
+0x98,
+0xa1,
+0x34,
+0x87,
+0x41,
+0xe1,
+0x83,
+0x70,
+0x30,
+0x94,
+0xdc,
+0x7a,
+0x34,
+0x10,
+0xea,
+0x6c,
+0xe2,
+0xfb,
+0xc8,
+0x29,
+0x97,
+0x26,
+0xa0,
+0x64,
+0x2e,
+0x56,
+0x90,
+0xf2,
+0xc6,
+0x7d,
+0x86,
+0x76,
+0x41,
+0xe3,
+0x62,
+0x5e,
+0x03,
+0xda,
+0x33,
+0x8f,
+0x4b,
+0x44,
+0x7a,
+0x9c,
+0x16,
+0x2d,
+0xc5,
+0x16,
+0xfa,
+0x5a,
+0x99,
+0xa2,
+0x51,
+0x60,
+0x66,
+0x01,
+0xa9,
+0x75,
+0x05,
+0x44,
+0x7e,
+0x3b,
+0xc0,
+0x97,
+0xbc,
+0x58,
+0xe0,
+0x0a,
+0x18,
+0x8f,
+0x5a,
+0xd1,
+0x83,
+0x7f,
+0x15,
+0x18,
+0xae,
+0xbb,
+0x75,
+0xb0,
+0x5b,
+0x9c,
+0xa2,
+0x12,
+0x2d,
+0x16,
+0x4a,
+0x8e,
+0xc7,
+0x69,
+0xd5,
+0xc3,
+0xa1,
+0x39,
+0x5d,
+0x2d,
+0x88,
+0xb8,
+0x29,
+0xa4,
+0x5d,
+0x4a,
+0x76,
+0x7c,
+0x8b,
+0x3b,
+0xf0,
+0x55,
+0xa2,
+0x6c,
+0xb2,
+0xb6,
+0xf7,
+0xaa,
+0x05,
+0xb3,
+0x19,
+0x4c,
+0x7e,
+0x7d,
+0xdf,
+0x30,
+0xed,
+0x1c,
+0x61,
+0x7b,
+0x08,
+0x30,
+0x03,
+0x22,
+0x57,
+0x15,
+0x15,
+0x7b,
+0x93,
+0x5e,
+0xbf,
+0x30,
+0x07,
+0xa2,
+0xd6,
+0xf6,
+0xc8,
+0xb4,
+0xce,
+0xb7,
+0x96,
+0x62,
+0x33,
+0x83,
+0xcd,
+0xd4,
+0x92,
+0x00,
+0xc7,
+0x4e,
+0xf8,
+0x5e,
+0xb6,
+0x1d,
+0xcf,
+0x04,
+0x4a,
+0x13,
+0x11,
+0x7d,
+0x2a,
+0xa5,
+0xd6,
+0xb0,
+0xee,
+0xfd,
+0xe8,
+0x4f,
+0xc0,
+0x9d,
+0x70,
+0xe0,
+0x27,
+0x41,
+0x80,
+0x2a,
+0x92,
+0x79,
+0x23,
+0xf1,
+0x70,
+0x5f,
+0x1b,
+0xcf,
+0x89,
+0xd5,
+0xa5,
+0xdb,
+0x5c,
+0x65,
+0xc6,
+0x86,
+0x37,
+0x44,
+0xbe,
+0xdf,
+0x90,
+0xd0,
+0x2c,
+0xe2,
+0x24,
+0x0e,
+0x3e,
+0xdb,
+0xe3,
+0x4f,
+0xdd,
+0xdf,
+0xe1,
+0x40,
+0x2b,
+0x64,
+0x58,
+0x70,
+0x4f,
+0xae,
+0xc9,
+0x8a,
+0x67,
+0x03,
+0x34,
+0xba,
+0xf4,
+0x06,
+0x8c,
+0xfb,
+0x6f,
+0x1c,
+0x1c,
+0xa0,
+0x70,
+0x72,
+0xdd,
+0xbc,
+0x12,
+0xd7,
+0xe2,
+0xc0,
+0x6d,
+0x31,
+0x19,
+0x5c,
+0xe2,
+0x14,
+0x4f,
+0x1a,
+0x50,
+0xb6,
+0xb0,
+0x1d,
+0x26,
+0x54,
+0x8e,
+0xfa,
+0x70,
+0xf7,
+0xeb,
+0x58,
+0x21,
+0x6f,
+0xf0,
+0x9f,
+0x79,
+0x4d,
+0x7b,
+0x32,
+0x0a,
+0x6b,
+0xbe,
+0xbd,
+0x02,
+0x6c,
+0xe8,
+0xec,
+0x48,
+0x60,
+0x76,
+0x98,
+0xed,
+0x86,
+0xb7,
+0x1e,
+0xf9,
+0x68,
+0x89,
+0x71,
+0x09,
+0x66,
+0x8d,
+0xa5,
+0x17,
+0x77,
+0x05,
+0xb6,
+0xac,
+0x6c,
+0x80,
+0xb3,
+0x19,
+0x49,
+0xf1,
+0x6d,
+0x28,
+0xe3,
+0x87,
+0x33,
+0x67,
+0xb1,
+0x2f,
+0x2a,
+0xd8,
+0xf1,
+0x54,
+0x22,
+0x88,
+0x19,
+0x55,
+0xc4,
+0x08,
+0xa6,
+0x16,
+0x4b,
+0x68,
+0x9e,
+0xd0,
+0x88,
+0xaa,
+0x03,
+0x47,
+0xd2,
+0x3b,
+0xf4,
+0xb2,
+0x38,
+0x3d,
+0x6a,
+0x6d,
+0xac,
+0x1e,
+0x15,
+0x09,
+0xa3,
+0xcf,
+0x46,
+0x85,
+0x12,
+0xfe,
+0xe2,
+0x89,
+0xaf,
+0xff,
+0x49,
+0x78,
+0x56,
+0xbc,
+0x8b,
+0x56,
+0xdd,
+0xd4,
+0x06,
+0xb8,
+0x71,
+0xa8,
+0x9a,
+0xb8,
+0xd1,
+0xdc,
+0x7b,
+0xdb,
+0xbf,
+0x31,
+0xb0,
+0xc8,
+0x7e,
+0x4f,
+0x3b,
+0xa4,
+0x6c,
+0x8d,
+0xdf,
+0xb7,
+0x7c,
+0xf3,
+0xac,
+0xb6,
+0x81,
+0x6f,
+0xf3,
+0xff,
+0xbe,
+0x5a,
+0xa5,
+0x56,
+0x3a,
+0x42,
+0x8b,
+0x75,
+0xff,
+0xd5,
+0x32,
+0x42,
+0x2b,
+0x61,
+0x37,
+0xf5,
+0x60,
+0xe8,
+0xe0,
+0x66,
+0x33,
+0x65,
+0x56,
+0x46,
+0x86,
+0x89,
+0x38,
+0xb8,
+0x34,
+0xad,
+0xfb,
+0x5c,
+0x8b,
+0xc1,
+0x64,
+0xa3,
+0x59,
+0xbd,
+0xb6,
+0x17,
+0xb7,
+0x55,
+0x9c,
+0xf2,
+0x12,
+0xcc,
+0xef,
+0x38,
+0xd3,
+0xc8,
+0xee,
+0xbf,
+0x80,
+0x6e,
+0x5b,
+0x57,
+0xfc,
+0xef,
+0x52,
+0xac,
+0x67,
+0xe1,
+0xd6,
+0x17,
+0xca,
+0x6b,
+0x45,
+0x18,
+0xde,
+0x3f,
+0x8a,
+0x06,
+0x3f,
+0x04,
+0x3d,
+0x6c,
+0xe2,
+0xe1,
+0xc6,
+0x7a,
+0x06,
+0xbd,
+0x1a,
+0x0d,
+0xbf,
+0xf2,
+0x1b,
+0x99,
+0x62,
+0x62,
+0xf3,
+0x85,
+0x90,
+0x5b,
+0xd1,
+0xf2,
+0x14,
+0x49,
+0x9a,
+0x4c,
+0x85,
+0x15,
+0xf1,
+0xc0,
+0xd5,
+0xf0,
+0x13,
+0x07,
+0x77,
+0x67,
+0x68,
+0xf4,
+0x4e,
+0x04,
+0xfe,
+0x56,
+0x44,
+0xd6,
+0x8e,
+0x7a,
+0xce,
+0xe8,
+0x6d,
+0xcb,
+0x6b,
+0xcf,
+0x4a,
+0x36,
+0x8f,
+0x93,
+0x31,
+0x59,
+0xda,
+0x61,
+0x60,
+0xc3,
+0xdd,
+0xe2,
+0x94,
+0xb8,
+0xd3,
+0xa1,
+0xdc,
+0x7c,
+0x1c,
+0x24,
+0x42,
+0xa6,
+0x3c,
+0x8f,
+0x17,
+0x01,
+0xc8,
+0x64,
+0xbd,
+0xfd,
+0x6b,
+0x5a,
+0xdb,
+0x10,
+0x08,
+0xf7,
+0x7c,
+0x50,
+0xdc,
+0x88,
+0x2e,
+0xf7,
+0xf8,
+0x02,
+0x26,
+0x67,
+0x0d,
+0xe6,
+0x9c,
+0xf6,
+0x3d,
+0xb6,
+0x9c,
+0x48,
+0x3f,
+0x4c,
+0x27,
+0x70,
+0x14,
+0xe7,
+0x84,
+0x3a,
+0x94,
+0x72,
+0x9a,
+0x42,
+0x54,
+0xb7,
+0x6f,
+0xb2,
+0x14,
+0xf1,
+0xe6,
+0x4c,
+0xd7,
+0x17,
+0xe1,
+0x70,
+0x5a,
+0x50,
+0x39,
+0x2b,
+0xf6,
+0x7a,
+0xcb,
+0x46,
+0xd5,
+0x53,
+0x2f,
+0x27,
+0xf0,
+0x12,
+0xd8,
+0xba,
+0xb6,
+0x39,
+0x1d,
+0x0b,
+0xb3,
+0xbf,
+0xe9,
+0x0d,
+0xf5,
+0x36,
+0x8f,
+0xfa,
+0x99,
+0xc3,
+0xef,
+0x6c,
+0x12,
+0x14,
+0x97,
+0xa5,
+0x7d,
+0x71,
+0x95,
+0x68,
+0x6f,
+0x0f,
+0x52,
+0x10,
+0x4d,
+0x06,
+0x8f,
+0xe4,
+0x01,
+0x75,
+0x7e,
+0x06,
+0xf5,
+0xe6,
+0xdd,
+0xd4,
+0x34,
+0x2d,
+0x74,
+0x3c,
+0x5e,
+0x20,
+0xba,
+0xcf,
+0x33,
+0x09,
+0x9f,
+0xf2,
+0x5e,
+0x23,
+0x17,
+0x35,
+0x41,
+0xd7,
+0xe7,
+0x3d,
+0x3d,
+0x8a,
+0x24,
+0xbb,
+0x45,
+0x4e,
+0xa9,
+0x03,
+0x2c,
+0x09,
+0xd7,
+0x4f,
+0xaa,
+0x70,
+0xb6,
+0x1c,
+0xe2,
+0xcb,
+0xf6,
+0xef,
+0x0e,
+0x08,
+0x90,
+0x0c,
+0x23,
+0xe8,
+0x44,
+0xd5,
+0x5d,
+0x83,
+0xb2,
+0xed,
+0x80,
+0xdf,
+0xf2,
+0xa7,
+0xaa,
+0xfe,
+0xa9,
+0x43,
+0x0e,
+0xe8,
+0x48,
+0xff,
+0x0c,
+0xc4,
+0xd8,
+0x79,
+0x7c,
+0x94,
+0x20,
+0x83,
+0x04,
+0xc4,
+0x85,
+0xc2,
+0x46,
+0x1a,
+0x99,
+0x2d,
+0xff,
+0x65,
+0x56,
+0x81,
+0x11,
+0x54,
+0x5a,
+0xab,
+0xe1,
+0xc5,
+0xad,
+0xb3,
+0xfe,
+0xc3,
+0xfd,
+0xd4,
+0x28,
+0xb0,
+0xa8,
+0xab,
+0x1d,
+0x5e,
+0x37,
+0x7c,
+0xdb,
+0x11,
+0xa8,
+0x14,
+0x57,
+0x13,
+0x86,
+0x6a,
+0x81,
+0x7e,
+0x9c,
+0x31,
+0x94,
+0x89,
+0xb5,
+0x8c,
+0x35,
+0xbd,
+0xe1,
+0xae,
+0xf9,
+0x5f,
+0x74,
+0x88,
+0x58,
+0x7a,
+0x0e,
+0x2b,
+0x7e,
+0x4c,
+0xd8,
+0x38,
+0x10,
+0xab,
+0xfc,
+0x22,
+0xae,
+0xc9,
+0x3f,
+0x32,
+0x85,
+0xff,
+0x90,
+0xd8,
+0xb0,
+0x2b,
+0x5b,
+0x1f,
+0x92,
+0x18,
+0x28,
+0x5f,
+0xf4,
+0x58,
+0x37,
+0x8c,
+0xc5,
+0xe6,
+0xfd,
+0x69,
+0x27,
+0x95,
+0xc6,
+0x64,
+0x0c,
+0x0a,
+0x0b,
+0xe9,
+0xbb,
+0xf7,
+0xb9,
+0x10,
+0x68,
+0xca,
+0x34,
+0x36,
+0xa3,
+0x85,
+0xb1,
+0xd8,
+0xdb,
+0x48,
+0x32,
+0x6f,
+0xbf,
+0x50,
+0x63,
+0x83,
+0x72,
+0xb3,
+0x87,
+0xbd,
+0x6b,
+0x39,
+0xd2,
+0x6c,
+0xaf,
+0x0a,
+0x1f,
+0xaf,
+0x21,
+0x87,
+0xc7,
+0xcc,
+0x32,
+0x7a,
+0xf3,
+0xed,
+0xce,
+0xcd,
+0x9c,
+0x12,
+0x6c,
+0x00,
+0x93,
+0xde,
+0x93,
+0x52,
+0x86,
+0x44,
+0x6f,
+0xdd,
+0xbd,
+0x8d,
+0x67,
+0xf2,
+0xe6,
+0x3d,
+0xce,
+0xaf,
+0xc8,
+0x05,
+0xc9,
+0x16,
+0xb7,
+0xe6,
+0x29,
+0x4d,
+0xc8,
+0xc3,
+0x76,
+0xbb,
+0x83,
+0x90,
+0x91,
+0xd1,
+0x68,
+0x9c,
+0x9a,
+0x46,
+0xa6,
+0xaa,
+0x05,
+0x8a,
+0x92,
+0x88,
+0x49,
+0x35,
+0x5e,
+0x19,
+0x2e,
+0xda,
+0xa9,
+0xf7,
+0xbe,
+0x31,
+0x99,
+0xad,
+0xcf,
+0x2f,
+0xcd,
+0x65,
+0x61,
+0xc9,
+0xda,
+0xc1,
+0x91,
+0xfd,
+0xac,
+0x7f,
+0xb9,
+0xec,
+0xdc,
+0x50,
+0x74,
+0x68,
+0x7f,
+0x18,
+0x74,
+0xae,
+0xe8,
+0x46,
+0xdf,
+0x18,
+0x30,
+0xf9,
+0xb4,
+0x96,
+0x17,
+0x2d,
+0x78,
+0x89,
+0xef,
+0x8b,
+0xc7,
+0x68,
+0x8f,
+0x3f,
+0xc6,
+0xec,
+0xd8,
+0x33,
+0xa5,
+0xd7,
+0xf1,
+0xbb,
+0x75,
+0x9a,
+0x53,
+0x9d,
+0x49,
+0x83,
+0x49,
+0x01,
+0x4b,
+0x58,
+0x93,
+0xa8,
+0xd9,
+0xae,
+0x1d,
+0xcb,
+0xb0,
+0x54,
+0xa7,
+0xfd,
+0x3e,
+0x60,
+0x38,
+0x4e,
+0x7c,
+0xcc,
+0x98,
+0x12,
+0x7c,
+0xe9,
+0x2b,
+0xde,
+0x7e,
+0x4f,
+0x96,
+0x87,
+0x35,
+0xd3,
+0xf0,
+0x6d,
+0x8d,
+0x7a,
+0x61,
+0x50,
+0xa5,
+0xad,
+0x51,
+0x9a,
+0x6c,
+0x04,
+0x52,
+0x84,
+0x2c,
+0xfa,
+0x90,
+0x78,
+0x7f,
+0xb1,
+0x45,
+0x8c,
+0xb1,
+0x09,
+0x7e,
+0xe9,
+0x9f,
+0xe6,
+0xa0,
+0xee,
+0x0c,
+0x03,
+0x13,
+0x29,
+0xa2,
+0x69,
+0xc8,
+0x74,
+0xe9,
+0x23,
+0x07,
+0x3c,
+0xd8,
+0xa5,
+0xff,
+0x70,
+0xa9,
+0x20,
+0x83,
+0xf0,
+0x65,
+0x21,
+0x89,
+0x53,
+0x64,
+0x0d,
+0x25,
+0x06,
+0xd4,
+0xd2,
+0x10,
+0xc2,
+0x86,
+0x80,
+0xd6,
+0x31,
+0x69,
+0xe9,
+0x0a,
+0xd3,
+0x2f,
+0xc2,
+0xc9,
+0xb6,
+0x00,
+0xb4,
+0xd6,
+0x09,
+0xe2,
+0x4c,
+0x92,
+0x41,
+0x7a,
+0x30,
+0xef,
+0xe3,
+0xf2,
+0xa8,
+0x1e,
+0xe7,
+0x00,
+0x3e,
+0x96,
+0x39,
+0xbc,
+0x90,
+0x2c,
+0x84,
+0xa1,
+0x1e,
+0x79,
+0x96,
+0xe1,
+0x19,
+0x7e,
+0x69,
+0x3a,
+0x51,
+0x5c,
+0x04,
+0xf3,
+0xfe,
+0x44,
+0x5c,
+0x50,
+0x42,
+0x28,
+0x6e,
+0x61,
+0xed,
+0xc5,
+0xa5,
+0xe1,
+0x04,
+0xf9,
+0x07,
+0x3f,
+0xf9,
+0xdd,
+0xc3,
+0x6d,
+0x50,
+0x16,
+0x20,
+0x31,
+0x0f,
+0xa5,
+0x89,
+0xfd,
+0x6d,
+0xce,
+0x66,
+0xfd,
+0xd0,
+0x6c,
+0xe8,
+0x6b,
+0x1a,
+0x6e,
+0x6b,
+0x81,
+0x98,
+0x24,
+0x83,
+0x30,
+0x20,
+0x33,
+0x78,
+0x5b,
+0xf9,
+0x98,
+0xe8,
+0x06,
+0xff,
+0x0e,
+0x45,
+0x65,
+0x35,
+0x78,
+0x5e,
+0x7f,
+0xe8,
+0xce,
+0xc8,
+0x2c,
+0xcf,
+0xca,
+0x4a,
+0xa1,
+0x0d,
+0x71,
+0x3e,
+0xa0,
+0x15,
+0xbf,
+0xa9,
+0x6a,
+0x61,
+0x02,
+0xb5,
+0x88,
+0x90,
+0xa6,
+0xe8,
+0xd0,
+0x5d,
+0xdb,
+0x5f,
+0x7b,
+0xc9,
+0xde,
+0x4f,
+0x0b,
+0x6f,
+0x84,
+0x39,
+0x9b,
+0xbc,
+0x79,
+0x30,
+0x5c,
+0x2e,
+0x19,
+0x8f,
+0x03,
+0x9b,
+0x2f,
+0x68,
+0x17,
+0xd3,
+0x4d,
+0x15,
+0xfc,
+0xea,
+0x44,
+0xb6,
+0x98,
+0xd4,
+0x05,
+0x3f,
+0xd7,
+0x1b,
+0x35,
+0x1a,
+0x30,
+0x4b,
+0xbe,
+0x1c,
+0xcb,
+0x94,
+0x24,
+0xac,
+0x45,
+0x14,
+0xdf,
+0x90,
+0xbf,
+0x0e,
+0x98,
+0xe9,
+0x84,
+0xab,
+0xb5,
+0xfc,
+0xe6,
+0xec,
+0x11,
+0x22,
+0x75,
+0xf3,
+0x8e,
+0x27,
+0xd5,
+0x6a,
+0xaf,
+0xe6,
+0x15,
+0x8f,
+0xbe,
+0x3a,
+0x63,
+0x81,
+0xde,
+0xe1,
+0x41,
+0xec,
+0x14,
+0x8e,
+0x24,
+0xdc,
+0x79,
+0xf2,
+0x9d,
+0xaf,
+0xdb,
+0x09,
+0x3d,
+0x99,
+0xbb,
+0x66,
+0xf2,
+0xa4,
+0x24,
+0xb9,
+0x86,
+0xaa,
+0xc6,
+0x5b,
+0x55,
+0xba,
+0xb9,
+0x28,
+0xd1,
+0x6a,
+0x13,
+0x48,
+0x68,
+0xd1,
+0x3c,
+0xd8,
+0xef,
+0xb7,
+0x68,
+0x99,
+0xeb,
+0xed,
+0x63,
+0x6d,
+0x09,
+0xf9,
+0xdf,
+0xde,
+0x30,
+0x72,
+0x13,
+0xcf,
+0x02,
+0xba,
+0x15,
+0x8a,
+0x1d,
+0xfe,
+0x40,
+0x02,
+0x9f,
+0x7b,
+0xb5,
+0x07,
+0xcf,
+0x9a,
+0x09,
+0xb2,
+0x32,
+0x1f,
+0xf6,
+0x67,
+0x2a,
+0x4a,
+0x57,
+0x33,
+0x63,
+0x22,
+0xaa,
+0x5f,
+0x94,
+0xd1,
+0x30,
+0x68,
+0x48,
+0xed,
+0x2c,
+0xde,
+0x66,
+0x0c,
+0x2c,
+0x38,
+0xdf,
+0x9c,
+0x40,
+0x03,
+0xef,
+0x07,
+0xac,
+0x6c,
+0x42,
+0xeb,
+0xa0,
+0x21,
+0xc4,
+0x9a,
+0x04,
+0x46,
+0xee,
+0x8c,
+0x88,
+0xb0,
+0xfd,
+0xa9,
+0x1e,
+0x7b,
+0x11,
+0x2d,
+0x97,
+0x11,
+0x00,
+0xe5,
+0x3a,
+0x41,
+0x25,
+0x0d,
+0xc0,
+0xff,
+0xe2,
+0x67,
+0x7b,
+0xf3,
+0xf6,
+0x4d,
+0xa1,
+0x71,
+0xcd,
+0xe7,
+0x5a,
+0xa2,
+0x48,
+0x0b,
+0x0a,
+0x48,
+0xe3,
+0x9a,
+0x35,
+0x7a,
+0x35,
+0x11,
+0x4c,
+0x0e,
+0x9b,
+0x51,
+0xe2,
+0xdd,
+0x1c,
+0x78,
+0x40,
+0x67,
+0x8b,
+0x13,
+0x8d,
+0x6b,
+0xec,
+0xb3,
+0xfa,
+0x5c,
+0xfe,
+0xcd,
+0xbc,
+0xa0,
+0x34,
+0x4a,
+0x2e,
+0xc6,
+0x3b,
+0x55,
+0x80,
+0xbd,
+0xb9,
+0x71,
+0xe0,
+0x65,
+0xe1,
+0x11,
+0x31,
+0xf8,
+0x40,
+0xe4,
+0x8d,
+0x49,
+0x2e,
+0x7c,
+0x00,
+0x06,
+0xbc,
+0x4c,
+0xaf,
+0xd7,
+0x23,
+0x66,
+0xf1,
+0x96,
+0xeb,
+0x55,
+0x54,
+0x13,
+0x65,
+0x11,
+0x1a,
+0x17,
+0x21,
+0x17,
+0xd6,
+0xdd,
+0x3f,
+0xd1,
+0x32,
+0xda,
+0x3a,
+0x2e,
+0x8f,
+0xe3,
+0xa8,
+0xd4,
+0x7e,
+0xe8,
+0xca,
+0xf3,
+0x99,
+0x69,
+0x4f,
+0x56,
+0xd9,
+0xf5,
+0x8d,
+0x67,
+0x52,
+0xaa,
+0xa5,
+0x58,
+0x5e,
+0x21,
+0xf8,
+0x28,
+0xef,
+0x96,
+0x6c,
+0x4f,
+0x49,
+0x4d,
+0xd7,
+0x1a,
+0xda,
+0xd9,
+0x2f,
+0xf9,
+0x85,
+0x75,
+0x24,
+0x4e,
+0x1a,
+0xa4,
+0x98,
+0x71,
+0x7e,
+0x1c,
+0xed,
+0xb7,
+0x31,
+0x3e,
+0x5d,
+0x9c,
+0x2a,
+0xed,
+0x16,
+0xaf,
+0x04,
+0x4b,
+0x5a,
+0x5e,
+0xb0,
+0xa0,
+0x58,
+0xe7,
+0x31,
+0xec,
+0xb0,
+0x64,
+0xb9,
+0x60,
+0x6f,
+0x83,
+0xce,
+0xef,
+0x8f,
+0x39,
+0x05,
+0x3c,
+0xca,
+0xc6,
+0x6d,
+0xb0,
+0xc6,
+0xf3,
+0x79,
+0xb9,
+0x24,
+0x6a,
+0xc8,
+0xd7,
+0xb5,
+0x15,
+0xa7,
+0xa7,
+0xb9,
+0x2c,
+0xed,
+0x5c,
+0x1c,
+0xd4,
+0xd4,
+0x89,
+0xdb,
+0xf6,
+0x40,
+0x73,
+0xa2,
+0x09,
+0xbe,
+0x96,
+0x65,
+0xd0,
+0x72,
+0xc2,
+0x94,
+0x17,
+0x7b,
+0x03,
+0x8b,
+0xd0,
+0x2f,
+0x67,
+0x19,
+0x82,
+0x56,
+0x39,
+0xa8,
+0x9d,
+0x8c,
+0x11,
+0xa2,
+0x68,
+0xd4,
+0x84,
+0x79,
+0xba,
+0x76,
+0x09,
+0x27,
+0x2a,
+0x12,
+0xa4,
+0x53,
+0x32,
+0xda,
+0x71,
+0x98,
+0x36,
+0x08,
+0x9b,
+0x73,
+0x55,
+0x39,
+0xc5,
+0xd8,
+0xe7,
+0x4a,
+0xb8,
+0x0c,
+0x0d,
+0x83,
+0x11,
+0xe9,
+0x8d,
+0xbc,
+0x06,
+0x60,
+0xff,
+0x08,
+0xe7,
+0x14,
+0x73,
+0xd5,
+0x2d,
+0x4a,
+0xb6,
+0x5a,
+0x3d,
+0x43,
+0xea,
+0xd1,
+0xde,
+0x98,
+0xf4,
+0xe3,
+0xc7,
+0x6d,
+0xbc,
+0x59,
+0xae,
+0xfc,
+0x00,
+0xf3,
+0x2f,
+0x57,
+0xa6,
+0xb0,
+0xb2,
+0x74,
+0xfc,
+0x67,
+0x9d,
+0x9e,
+0x46,
+0xa0,
+0xaa,
+0x05,
+0x84,
+0x10,
+0x47,
+0x50,
+0xc2,
+0xc0,
+0xda,
+0x07,
+0xc7,
+0x4d,
+0xe1,
+0xf1,
+0x9b,
+0x23,
+0xd5,
+0xf4,
+0x77,
+0xd5,
+0x64,
+0xc2,
+0xf0,
+0x16,
+0x63,
+0x24,
+0x3b,
+0x30,
+0x15,
+0xfa,
+0x68,
+0x83,
+0x40,
+0xf2,
+0x61,
+0xe2,
+0x9c,
+0xdf,
+0xa7,
+0x91,
+0xf7,
+0x39,
+0x3d,
+0xe8,
+0x38,
+0x78,
+0x28,
+0xf3,
+0xb5,
+0x6a,
+0x78,
+0x51,
+0x74,
+0x53,
+0x16,
+0xed,
+0x81,
+0x06,
+0x0d,
+0xda,
+0x51,
+0xb0,
+0x3f,
+0xc5,
+0x1a,
+0x56,
+0xae,
+0xd0,
+0xc2,
+0x14,
+0x5c,
+0xad,
+0xfa,
+0xd8,
+0x2f,
+0xef,
+0x5d,
+0x47,
+0x4f,
+0x95,
+0x8a,
+0x27,
+0x80,
+0x11,
+0xd1,
+0x6b,
+0x43,
+0xab,
+0xea,
+0x57,
+0xad,
+0xe4,
+0xc6,
+0x5f,
+0x5a,
+0x46,
+0x66,
+0x9f,
+0x67,
+0x90,
+0xf1,
+0x3e,
+0x7d,
+0x02,
+0xcc,
+0x1d,
+0x4e,
+0x8c,
+0xb2,
+0x78,
+0x10,
+0x68,
+0x7f,
+0x23,
+0x14,
+0xea,
+0xa0,
+0x6c,
+0x48,
+0xce,
+0x23,
+0xf2,
+0x4c,
+0x25,
+0x94,
+0x80,
+0xda,
+0x9f,
+0xe5,
+0x70,
+0x4b,
+0x1a,
+0x94,
+0xf3,
+0xb3,
+0x67,
+0xfd,
+0x47,
+0x0e,
+0x87
+};
+
+
+
+/* -----------EXPECTED OUTPUTS----------- */
+uint8_t exp_outp_0 [] = {
+0xf7,
+0x15,
+0x51,
+0x51,
+0x24,
+0x15,
+0x55,
+0x55,
+0x51,
+0x5c,
+0x50,
+0x45,
+0x55,
+0x11,
+0xc0,
+0x05,
+0x55,
+0x15,
+0x55,
+0x45,
+0x55,
+0x55,
+0x51,
+0x45,
+0x51,
+0xc4,
+0x55,
+0x55,
+0x55,
+0x54,
+0x55,
+0x55,
+0x55,
+0x14,
+0x54,
+0x50,
+0x51,
+0x0c,
+0xd5,
+0x11,
+0x54,
+0x55,
+0x5d,
+0x75,
+0x47,
+0x50,
+0x44,
+0x14,
+0x10,
+0x50,
+0x54,
+0x35,
+0x57,
+0x14,
+0x44,
+0x45,
+0x54,
+0x55,
+0x50,
+0x1d,
+0x55,
+0x11,
+0x15,
+0x05,
+0x54,
+0x51,
+0xd5,
+0x45,
+0x51,
+0xd5,
+0x55,
+0x55,
+0x54,
+0x51,
+0x50,
+0x45,
+0xd4,
+0x47,
+0x10,
+0x4d,
+0x55,
+0x55,
+0x51,
+0x51,
+0xf4,
+0x54,
+0xd8,
+0x59,
+0x05,
+0xb3,
+0x4f,
+0x15,
+0x54,
+0x50,
+0x15,
+0xd4,
+0xc5,
+0x93,
+0x45,
+0x15,
+0x54,
+0x55,
+0xb5,
+0x1c,
+0x54,
+0x50,
+0x55,
+0x55,
+0x55,
+0x41,
+0xc5,
+0x55,
+0x15,
+0xc5,
+0x10,
+0x55,
+0x54,
+0x51,
+0x75,
+0x55,
+0x55,
+0x17,
+0x55,
+0x55,
+0x55,
+0x5c,
+0x05,
+0x55,
+0x55,
+0x11,
+0x55,
+0x45,
+0xc7,
+0x4c,
+0x57,
+0x55,
+0x15,
+0x55,
+0x54,
+0x51,
+0x55,
+0x51,
+0x55,
+0x51,
+0x41,
+0x70,
+0x55,
+0x15,
+0x45,
+0x0d,
+0x1d,
+0x55,
+0x4c,
+0x51,
+0x05,
+0x51,
+0x55,
+0x45,
+0xc5,
+0x51,
+0x51,
+0x51,
+0x55,
+0x15,
+0x5d,
+0x50,
+0x55,
+0x55,
+0x41,
+0x44,
+0x15,
+0x55,
+0x45,
+0x54,
+0x55,
+0x10,
+0x55,
+0x55,
+0x15,
+0x55,
+0x11,
+0x55,
+0xf5,
+0x55,
+0x15,
+0x55,
+0x55,
+0x51,
+0x45,
+0x05,
+0x51,
+0x55,
+0x54,
+0x05,
+0x44,
+0x71,
+0x5c,
+0x55,
+0x44,
+0x55,
+0x35,
+0x45,
+0x1c,
+0x50,
+0x55,
+0x45,
+0x55,
+0x15,
+0x5e,
+0x15,
+0x11,
+0x55,
+0x54,
+0x55,
+0x45,
+0x54,
+0x54,
+0x55,
+0x55,
+0x55,
+0x51,
+0x15,
+0x05,
+0x51,
+0x55,
+0x51,
+0x50,
+0x51,
+0x45,
+0x00,
+0x54,
+0x75,
+0x14,
+0x45,
+0x5d,
+0x44,
+0x51,
+0x55,
+0x15,
+0x40,
+0xd4,
+0x45,
+0x45,
+0x17,
+0x45,
+0x55,
+0x71,
+0x05,
+0x40,
+0x11,
+0x13,
+0xd4,
+0x15,
+0x51,
+0x55,
+0x01,
+0x57,
+0x50,
+0x55,
+0x74,
+0x15,
+0x55,
+0x51,
+0x47,
+0x15,
+0x55,
+0x17,
+0x55,
+0x4d,
+0x45,
+0x41,
+0x04,
+0x15,
+0x14,
+0x51,
+0x51,
+0x5d,
+0x55,
+0x31,
+0x55,
+0x47,
+0x55,
+0x15,
+0x41,
+0x55,
+0x41,
+0x04,
+0x55,
+0x54,
+0x55,
+0x54,
+0x5e,
+0x55,
+0x15,
+0x44,
+0x45,
+0xd1,
+0x15,
+0x55,
+0x15,
+0x75,
+0x54,
+0x54,
+0x34,
+0x15,
+0x15,
+0x54,
+0x55,
+0x54,
+0x05,
+0x55,
+0x14,
+0x55,
+0x14,
+0x15,
+0x11,
+0x51,
+0x50,
+0x45,
+0x55,
+0x1d,
+0x51,
+0x15,
+0x10,
+0x55,
+0x55,
+0xd4,
+0x45,
+0x55,
+0x55,
+0x55,
+0x35,
+0x55,
+0x15,
+0x75,
+0x54,
+0x5c,
+0x54,
+0x45,
+0x74,
+0x51,
+0x55,
+0x55,
+0x75,
+0x40,
+0x41,
+0xd0,
+0x5d,
+0x51,
+0x54,
+0x55,
+0x41,
+0x45,
+0x57,
+0x75,
+0x55,
+0x50,
+0x05,
+0x54,
+0x05,
+0x74,
+0x51,
+0x51,
+0x75,
+0x13,
+0x34,
+0x15,
+0x41,
+0x54,
+0x5d,
+0x5c,
+0x5d,
+0x53,
+0x04,
+0x44,
+0x41,
+0x55,
+0x55,
+0x54,
+0x55,
+0x54,
+0x01,
+0x41,
+0x91,
+0x45,
+0x40,
+0xc1,
+0x44,
+0x50,
+0x55,
+0x51,
+0x51,
+0x55,
+0x3c,
+0x50,
+0x55,
+0x51,
+0x55,
+0x50,
+0xd5,
+0x15,
+0x55,
+0x5d,
+0xc5,
+0x55,
+0x45,
+0x55,
+0x41,
+0x55,
+0x41,
+0x54,
+0x54,
+0x57,
+0x75,
+0xd0,
+0x71,
+0x01,
+0x75,
+0x15,
+0x54,
+0x45,
+0x45,
+0x54,
+0x71,
+0x51,
+0x55,
+0xc5,
+0x5d,
+0x55,
+0x44,
+0x15,
+0x47,
+0x1c,
+0x5d,
+0x55,
+0x53,
+0x55,
+0x55,
+0x15,
+0x55,
+0x45,
+0x55,
+0x01,
+0x55,
+0x15,
+0x55,
+0x44,
+0x51,
+0xd1,
+0xd5,
+0x14,
+0x41,
+0x15,
+0x54,
+0x55,
+0x55,
+0xd4,
+0x15,
+0x45,
+0x01,
+0x71,
+0x44,
+0x54,
+0xd5,
+0x54,
+0x4d,
+0x14,
+0x51,
+0x01,
+0x15,
+0x4d,
+0x55,
+0x55,
+0x45,
+0x44,
+0x75,
+0x54,
+0x0d,
+0x45,
+0x01,
+0x55,
+0xd5,
+0x15,
+0x45,
+0x55,
+0x55,
+0x55,
+0x54,
+0xd4,
+0x54,
+0x55,
+0x17,
+0x55,
+0x55,
+0x49,
+0x65,
+0x14,
+0x74,
+0x40,
+0x5d,
+0x55,
+0x34,
+0x54,
+0x55,
+0x44,
+0xd5,
+0x55,
+0x55,
+0x55,
+0x40,
+0x54,
+0x51,
+0xc5,
+0x01,
+0x15,
+0x51,
+0x55,
+0x54,
+0x50,
+0x48,
+0x45,
+0x5d,
+0x55,
+0x04,
+0x54,
+0xd1,
+0x54,
+0x57,
+0x55,
+0x54,
+0x05,
+0x11,
+0x41,
+0x54,
+0x35,
+0xd5,
+0x55,
+0x56,
+0x15,
+0x45,
+0x15,
+0x50,
+0x74,
+0x50,
+0x40,
+0x54,
+0x55,
+0x55,
+0x55,
+0x15,
+0x6d,
+0x51,
+0x54,
+0x54,
+0x75,
+0x55,
+0x51,
+0x51,
+0xf1,
+0x54,
+0x44,
+0x51,
+0x10,
+0x05,
+0x00,
+0xc7,
+0x55,
+0x05,
+0x5c,
+0x74,
+0x11,
+0x5d,
+0x4d,
+0x44,
+0x34,
+0x44,
+0x51,
+0xd5,
+0x55,
+0xcd,
+0x51,
+0x45,
+0x47,
+0x55,
+0x54,
+0x11,
+0x55,
+0x55,
+0x05,
+0x45,
+0x45,
+0x55,
+0x54,
+0x51,
+0x45,
+0x51,
+0x01,
+0x15,
+0x50,
+0x40,
+0x44,
+0x54,
+0x14,
+0x55,
+0x41,
+0x41,
+0x01,
+0x55,
+0x55,
+0x55,
+0x15,
+0x55,
+0x44,
+0x55,
+0x15,
+0x07,
+0xc4,
+0x15,
+0x45,
+0x5d,
+0x55,
+0x15,
+0x54,
+0x51,
+0x55,
+0x55,
+0x51,
+0x51,
+0x55,
+0x10,
+0x55,
+0x5d,
+0x54,
+0x11,
+0x55,
+0x4c,
+0xd5,
+0x51,
+0x45,
+0x50,
+0x55,
+0x45,
+0x41,
+0x15,
+0x57,
+0x75,
+0x33,
+0x55,
+0x54,
+0x10,
+0x15,
+0x50,
+0x55,
+0xc4,
+0x7d,
+0x55,
+0x45,
+0x15,
+0x45,
+0x15,
+0x45,
+0x55,
+0x15,
+0x55,
+0x51,
+0x45,
+0x51,
+0x55,
+0x51,
+0x41,
+0x15,
+0x44,
+0x45,
+0x55,
+0x47,
+0x14,
+0x54,
+0x53,
+0x54,
+0x51,
+0x05,
+0x55,
+0x55,
+0x55,
+0x55,
+0x55,
+0x51,
+0x51,
+0x15,
+0x51,
+0x14,
+0x55,
+0x55,
+0x45,
+0x54,
+0x50,
+0x55,
+0x55,
+0x45,
+0x4d,
+0x51,
+0x50,
+0x55,
+0x55,
+0x50,
+0x55,
+0x57,
+0x41,
+0x54,
+0x55,
+0x04,
+0x1d,
+0x41,
+0x55,
+0xd4,
+0xdd,
+0x4d,
+0x14,
+0x11,
+0x55,
+0x51,
+0x74,
+0x45,
+0x01,
+0x44,
+0x55,
+0x55,
+0x40,
+0x55,
+0x51,
+0x51,
+0x55,
+0x55,
+0x55,
+0x45,
+0x75,
+0x53,
+0x5c,
+0x54,
+0x55,
+0x51,
+0x15,
+0x55,
+0x01,
+0x55,
+0x55,
+0x54,
+0x55,
+0x15,
+0x50,
+0x11,
+0x50,
+0x14,
+0x53,
+0x55,
+0x55,
+0x54,
+0x15,
+0x5c,
+0x45,
+0x7d,
+0x18,
+0x14,
+0x55,
+0x44,
+0x55,
+0x1d,
+0x55,
+0x54,
+0x05,
+0x54,
+0x50,
+0x05,
+0x47,
+0x54,
+0xdc,
+0x51,
+0x51,
+0x51,
+0x14,
+0x54,
+0x54,
+0x55,
+0x45,
+0x55,
+0x50,
+0x55,
+0x55,
+0x15,
+0x44,
+0x54,
+0x51,
+0x15,
+0x55,
+0x51,
+0x11,
+0x14,
+0x54,
+0x55,
+0x45,
+0x41,
+0x15,
+0x04,
+0x55,
+0x70,
+0x10,
+0x54,
+0x55,
+0x75,
+0x55,
+0x55,
+0x45,
+0x54,
+0x45,
+0x01,
+0x55,
+0x15,
+0x74,
+0x35,
+0x15,
+0x14,
+0x51,
+0x55,
+0xd1,
+0x55,
+0x55,
+0x74,
+0x51,
+0x11,
+0x45,
+0x41,
+0xd5,
+0x15,
+0xd0,
+0x55,
+0x55,
+0x05,
+0x05,
+0x45,
+0x4d,
+0x1c,
+0x55,
+0x15,
+0x01,
+0x75,
+0x55,
+0x54,
+0x15,
+0x55,
+0x50,
+0x51,
+0x55,
+0x51,
+0x55,
+0x14,
+0xd5,
+0xd5,
+0x57,
+0x15,
+0x0c,
+0x47,
+0x55,
+0x05,
+0x5c,
+0x55,
+0x51,
+0x55,
+0x75,
+0x5c,
+0x4d,
+0x45,
+0xd5,
+0x50,
+0x5d,
+0x45,
+0x55,
+0x54,
+0x55,
+0x5d,
+0x05,
+0x51,
+0x5d,
+0x15,
+0x51,
+0x54,
+0x52,
+0x54,
+0x55,
+0x55,
+0x55,
+0x14,
+0xd4,
+0x40,
+0x5d,
+0x54,
+0x45,
+0x10,
+0x50,
+0x55,
+0x51,
+0x51,
+0x11,
+0x41,
+0x57,
+0x55,
+0x55,
+0xd0,
+0x55,
+0x13,
+0x40,
+0x04,
+0xd5,
+0x45,
+0x57,
+0x55,
+0x51,
+0x10,
+0x05,
+0x55,
+0x54,
+0x54,
+0x51,
+0x45,
+0xd4,
+0x75,
+0x55,
+0x11,
+0x54,
+0x45,
+0x55,
+0x15,
+0x55,
+0x45,
+0x1d,
+0x15,
+0x55,
+0x4d,
+0x45,
+0x15,
+0x11,
+0xd4,
+0x55,
+0x55,
+0x55,
+0x50,
+0x55,
+0x45,
+0x04,
+0x64,
+0x50,
+0x55,
+0x11,
+0x14,
+0x45,
+0x45,
+0x50,
+0x55,
+0x54,
+0x45,
+0x15,
+0x41,
+0x45,
+0x60,
+0x54,
+0x15,
+0x55,
+0x41,
+0x55,
+0x50,
+0x15,
+0x55,
+0x0d,
+0x55,
+0x51,
+0x45,
+0x55,
+0x34,
+0x15,
+0x55,
+0x57,
+0x50,
+0x10,
+0x55,
+0x50,
+0x50,
+0x41,
+0x05,
+0x04,
+0x55,
+0x55,
+0x55,
+0x05,
+0x01,
+0x11,
+0x41,
+0x05,
+0x71,
+0x41,
+0x54,
+0xd4,
+0x55,
+0x11,
+0x55,
+0x45,
+0x15,
+0x55,
+0x45,
+0x85,
+0x05,
+0x35,
+0x51,
+0x45,
+0x41,
+0x14,
+0x40,
+0x55,
+0x55,
+0x15,
+0xc5,
+0xc5,
+0x55,
+0x45,
+0x55,
+0x41,
+0x05,
+0x05,
+0x41,
+0x45,
+0x50,
+0x54,
+0x43,
+0x11,
+0x05,
+0x45,
+0x54,
+0x51,
+0x55,
+0x11,
+0x41,
+0xd1,
+0x45,
+0x44,
+0x55,
+0x55,
+0x55,
+0x55,
+0x47,
+0x05,
+0xd1,
+0x45,
+0x15,
+0x51,
+0x54,
+0x55,
+0x55,
+0x14,
+0x45,
+0x55,
+0x51,
+0x50,
+0x44,
+0x50,
+0x74,
+0x17,
+0x51,
+0x55,
+0x57,
+0x05,
+0x11,
+0x11,
+0x55,
+0x75,
+0x51,
+0x55,
+0x71,
+0x05,
+0x11,
+0x45,
+0x57,
+0x5c,
+0x51,
+0x54,
+0xd0,
+0x54,
+0x45,
+0x45,
+0x54,
+0x11,
+0x41,
+0x5c,
+0x50,
+0x55,
+0xd5,
+0x51,
+0x17,
+0x45,
+0x17,
+0x05,
+0x55,
+0x41,
+0x51,
+0x11,
+0x45,
+0x54,
+0x14,
+0x55,
+0x54,
+0x51,
+0x55,
+0x15,
+0x55,
+0x5c,
+0xd5,
+0x71,
+0x34,
+0x51,
+0x15,
+0x54,
+0x15,
+0x55,
+0x55,
+0xc4,
+0x15,
+0x05,
+0x55,
+0xe1,
+0x54,
+0x04,
+0x05,
+0x54,
+0x54,
+0x51,
+0x55,
+0x0d,
+0x95,
+0x55,
+0x54,
+0x5d,
+0x55,
+0x55,
+0x51,
+0x35,
+0x01,
+0x51,
+0x50,
+0x04,
+0x17,
+0x55,
+0x5d,
+0x55,
+0x15,
+0x50,
+0x51,
+0x45,
+0x41,
+0x55,
+0x41,
+0x54,
+0xd5,
+0x55,
+0x55,
+0xc4,
+0xc5,
+0x55,
+0x54,
+0x41,
+0x40,
+0x50,
+0x14,
+0x55,
+0x45,
+0x11,
+0x54,
+0x05,
+0xd5,
+0x5d,
+0xd4,
+0x75,
+0x0c,
+0x11,
+0x55,
+0x45,
+0x14,
+0xc7,
+0x55,
+0x55,
+0x75,
+0x15,
+0x55,
+0x45,
+0x55,
+0x51,
+0x41,
+0x55,
+0x45,
+0x54,
+0x51,
+0x55,
+0x50,
+0x75,
+0x41,
+0x58,
+0x5d,
+0x55,
+0x45,
+0x40,
+0x51,
+0x54,
+0x55,
+0x55,
+0x55,
+0x51,
+0x45,
+0x55,
+0x47,
+0xd5,
+0x45,
+0x51,
+0x15,
+0x44,
+0x54,
+0x54,
+0xd1,
+0x55,
+0x50,
+0x55,
+0x55,
+0x41,
+0x55,
+0x11,
+0x45,
+0x55,
+0x55,
+0x15,
+0xcd,
+0x51,
+0x11,
+0x45,
+0x54,
+0x57,
+0x05,
+0x04,
+0x55,
+0x44,
+0x54,
+0x59,
+0x51,
+0x51,
+0xd5,
+0x55,
+0x51,
+0x15,
+0x55,
+0x14,
+0x44,
+0x17,
+0x41,
+0x15,
+0x55,
+0x45,
+0x55,
+0x65,
+0x45,
+0x54,
+0x44,
+0x31,
+0xd5,
+0x10,
+0x51,
+0x54,
+0x54,
+0x51,
+0x07,
+0x55,
+0x11,
+0x55,
+0x45,
+0x55,
+0x05,
+0x55,
+0x55,
+0x15,
+0x14,
+0x51,
+0x54,
+0x54,
+0x11,
+0x54,
+0x45,
+0x75,
+0x75,
+0x05,
+0x05,
+0x51,
+0x04,
+0xd5,
+0x54,
+0x45,
+0x57,
+0x55,
+0x5d,
+0x55,
+0x15,
+0x55,
+0x54,
+0x14,
+0x65,
+0x05,
+0x45,
+0x53,
+0x45,
+0x54,
+0x54,
+0x17,
+0x45,
+0x55,
+0x01,
+0xd5,
+0x14,
+0x10,
+0x54,
+0x5d,
+0x15,
+0x55,
+0x55,
+0x51,
+0x45,
+0x10,
+0x55,
+0xc5,
+0x55,
+0x45,
+0x51,
+0x55,
+0x55,
+0x35,
+0x40,
+0x54,
+0x15,
+0x55,
+0x55,
+0x17,
+0x4c,
+0x54,
+0x55,
+0x57,
+0x72,
+0x45,
+0x44,
+0x5f,
+0x07,
+0x11,
+0x55,
+0x75,
+0x55,
+0x41,
+0x45,
+0xd7,
+0x45,
+0xdd,
+0x40,
+0xd5,
+0x40,
+0x04,
+0x57,
+0x5c,
+0x55,
+0x54,
+0x55,
+0x55,
+0x35,
+0x14,
+0x47,
+0x31,
+0x45,
+0x51,
+0x45,
+0x05,
+0x51,
+0x44,
+0x1c,
+0x14,
+0x51,
+0x11,
+0x10,
+0x00,
+0x45,
+0x14,
+0x51,
+0x5d,
+0x55,
+0x5d,
+0x54,
+0x45,
+0x54,
+0x55,
+0x15,
+0x44,
+0xf9,
+0x44,
+0x54,
+0x55,
+0x10,
+0x45,
+0x55,
+0x55,
+0x51,
+0x15,
+0x11,
+0x55,
+0x55,
+0x15,
+0x14,
+0x1c,
+0x51,
+0x17,
+0x41,
+0x44,
+0x51,
+0x41,
+0x51,
+0x54,
+0x75,
+0x74,
+0x57,
+0x17,
+0x51,
+0xc5,
+0x5d,
+0x54,
+0x15,
+0x55,
+0x55,
+0x75,
+0xd5,
+0x55,
+0x47,
+0x31,
+0x05,
+0x54,
+0x11,
+0x55,
+0x55,
+0x15,
+0x55,
+0xd1,
+0x7d,
+0x1d,
+0x55,
+0x15,
+0x5d,
+0x45,
+0x75,
+0x1d,
+0x5d,
+0x51,
+0x15,
+0x47,
+0x05,
+0x55,
+0x51,
+0x54,
+0x55,
+0x15,
+0x55,
+0x11,
+0x51,
+0x43,
+0x55,
+0x11,
+0x15,
+0x19,
+0x05,
+0x40,
+0x55,
+0x55,
+0x15,
+0x05,
+0x15,
+0x14,
+0x57,
+0x05,
+0x54,
+0x34,
+0x1d,
+0x55,
+0xd4,
+0x14,
+0x54,
+0x1d,
+0x05,
+0x57,
+0x05,
+0x15,
+0xd1,
+0x55,
+0x55,
+0x11,
+0xf5,
+0x55,
+0x78,
+0x51,
+0x54,
+0x55,
+0x45,
+0x55,
+0x57,
+0x55,
+0x51,
+0x55,
+0x05,
+0x5d,
+0x55,
+0x54,
+0x55,
+0x51,
+0x35,
+0x05,
+0x55,
+0xd5,
+0x51,
+0x55,
+0x50,
+0x15,
+0x55,
+0x55,
+0x45,
+0x14,
+0x55,
+0x15,
+0x15,
+0x04,
+0x55,
+0x14,
+0x51,
+0x14,
+0x51,
+0x55,
+0xc5,
+0x54,
+0x15,
+0x5d,
+0x14,
+0x41,
+0x55,
+0x40,
+0x55,
+0x55,
+0x55,
+0x11,
+0x15,
+0x15,
+0x55,
+0x55,
+0x04,
+0x04,
+0x55,
+0x55,
+0x55,
+0x40,
+0x51,
+0x55,
+0x54,
+0x70,
+0x15,
+0x55,
+0x51,
+0x44,
+0x50,
+0x54,
+0x11,
+0x00,
+0x55,
+0x15,
+0x15,
+0x75,
+0x55,
+0xd5,
+0x11,
+0x54,
+0x55,
+0x45,
+0x10,
+0x50,
+0x54,
+0x15,
+0x55,
+0x55,
+0x44,
+0x55,
+0x15,
+0x5d,
+0x14,
+0x55,
+0x71,
+0x55,
+0x54,
+0x51,
+0x54,
+0x1c,
+0x45,
+0x55,
+0x55,
+0x1c,
+0x50,
+0x55,
+0x14,
+0x55,
+0x35,
+0x41,
+0x45,
+0x55,
+0x51,
+0x50,
+0x55,
+0x1c,
+0xd5,
+0xcb,
+0x57,
+0x10,
+0x54,
+0x71,
+0x75,
+0x57,
+0x55,
+0x5d,
+0x53,
+0x54,
+0x44,
+0x51,
+0x51,
+0x35,
+0xc1,
+0x51,
+0x15,
+0x55,
+0x14,
+0x57,
+0x45,
+0x55,
+0x70,
+0x74,
+0x40,
+0x54,
+0x54,
+0x11,
+0x15,
+0x45,
+0x05,
+0x55,
+0x55,
+0x44,
+0x45,
+0x55,
+0x15,
+0x55,
+0x45,
+0x15,
+0x55,
+0x55,
+0x15,
+0x45,
+0x51,
+0x71,
+0x54,
+0x54,
+0x40,
+0x15,
+0x55,
+0x54,
+0x04,
+0x44,
+0x45,
+0x54,
+0x51,
+0x15,
+0x41,
+0x14,
+0x55,
+0x45,
+0x45,
+0x55,
+0x51,
+0x11,
+0x55,
+0xd1,
+0x14,
+0x55,
+0x5d,
+0x55,
+0x15,
+0x94,
+0xc5,
+0x51,
+0x51,
+0x55,
+0x74,
+0x41,
+0x45,
+0x53,
+0x14,
+0x45,
+0x57,
+0xd5,
+0x54,
+0x34,
+0x41,
+0x54,
+0x75,
+0x50,
+0x15,
+0x14,
+0x75,
+0x45,
+0x55,
+0x55,
+0x55,
+0x05,
+0x01,
+0x54,
+0x43,
+0x51,
+0x41,
+0xd4,
+0x14,
+0x17,
+0x55,
+0x54,
+0xd5,
+0x55,
+0x41,
+0x55,
+0x55,
+0x00,
+0x5c,
+0x55,
+0x50,
+0x14,
+0x05,
+0xd5,
+0x45,
+0x44,
+0x15,
+0x11,
+0x05,
+0xd1,
+0x55,
+0x55,
+0xd1,
+0x55,
+0x55,
+0x51,
+0x54,
+0x55,
+0x55,
+0x71,
+0x14,
+0x55,
+0x54,
+0x17,
+0x54,
+0x45,
+0x55,
+0x51,
+0x75,
+0xd5,
+0x51,
+0x75,
+0x59,
+0x50,
+0x71,
+0x50,
+0x40,
+0x5d,
+0x45,
+0x54,
+0x15,
+0x55,
+0x5c,
+0x14,
+0x51,
+0x04,
+0x54,
+0x55,
+0x54,
+0x15,
+0x04,
+0x1d,
+0x75,
+0x35,
+0x15,
+0x15,
+0x55,
+0x50,
+0x44,
+0x14,
+0x45,
+0xf3,
+0x55,
+0x51,
+0x44,
+0x41,
+0x13,
+0x55,
+0x15,
+0x71,
+0x15,
+0x54,
+0x51,
+0x50,
+0x51,
+0x50,
+0x55,
+0x51,
+0x51,
+0x55,
+0x45,
+0x41,
+0x55,
+0x15,
+0x05,
+0x45,
+0x15,
+0xd5,
+0xd5,
+0x51,
+0x51,
+0x4c,
+0x54,
+0x55,
+0x54,
+0x44,
+0x11,
+0x05,
+0x10,
+0x11,
+0x45,
+0x55,
+0x35,
+0x5d,
+0x51,
+0x51,
+0x0d,
+0x41,
+0x45,
+0x71,
+0x15,
+0xdc,
+0x55,
+0x54,
+0x51,
+0x50,
+0x55,
+0x44,
+0xd5,
+0x14,
+0x41,
+0x54,
+0x4d,
+0x15,
+0x01,
+0x15,
+0x54,
+0x41,
+0x45,
+0x55,
+0x54,
+0x55,
+0x15,
+0x55,
+0x13,
+0x55,
+0x5c,
+0x45,
+0x54,
+0xd4,
+0x11,
+0x4c,
+0x55,
+0x14,
+0x51,
+0x45,
+0x43,
+0x55,
+0x54,
+0x15,
+0x51,
+0x51,
+0x55,
+0x55,
+0x05,
+0x55,
+0x04,
+0x75,
+0x11,
+0x51,
+0x75,
+0xd1,
+0x65,
+0x14,
+0x45,
+0x05,
+0x11,
+0x17,
+0x50,
+0x55,
+0x15,
+0x0d,
+0x50,
+0x55,
+0x45,
+0x55,
+0x15,
+0x55,
+0x51,
+0x50,
+0x51,
+0x05,
+0xc1,
+0x15,
+0x54,
+0x15,
+0x15,
+0x55,
+0x05,
+0x11,
+0x54,
+0x45,
+0x57,
+0x14,
+0x55,
+0x54,
+0x54,
+0x45,
+0x11,
+0x70,
+0x55,
+0x55,
+0x44,
+0x54,
+0x45,
+0x51,
+0xdd,
+0x44,
+0x41,
+0x15,
+0x15,
+0x14,
+0x54,
+0x51,
+0x55,
+0x15,
+0x15,
+0x11,
+0x45,
+0x4c,
+0x45,
+0x45,
+0x4d,
+0x01,
+0x45,
+0x55,
+0x57,
+0x55,
+0x51,
+0x55,
+0x15,
+0x50,
+0x47,
+0x55,
+0x55,
+0x51,
+0x51,
+0x54,
+0x40,
+0x44,
+0x55,
+0x50,
+0x55,
+0xc0,
+0xd1,
+0x05,
+0x45,
+0x51,
+0x45,
+0x01,
+0x51,
+0x15,
+0x55,
+0x44,
+0x44,
+0x15,
+0x15,
+0x55,
+0x55,
+0x11,
+0x54,
+0x31,
+0x55,
+0x54,
+0x55,
+0x50,
+0x15,
+0x71,
+0xd5,
+0x46,
+0x41,
+0x51,
+0x51,
+0x55,
+0x54,
+0x55,
+0x15,
+0x55,
+0x64,
+0x50,
+0x55,
+0x44,
+0x41,
+0x55,
+0x45,
+0x54,
+0x51,
+0x15,
+0x05,
+0x5d,
+0x51,
+0x46,
+0x55,
+0x15,
+0x47,
+0x15,
+0x40,
+0x50,
+0x54,
+0x41,
+0x70,
+0x50,
+0x54,
+0x05,
+0xd4,
+0xc4,
+0x11,
+0x55,
+0x35,
+0x44,
+0x47,
+0x55
+};
+
+
+
+L1_DATA uint32_t threshs_l1 [32] = {0};
+L1_DATA int32_t kappa_l1 [32] = {0};
+L1_DATA int32_t lambda_l1 [32] = {0};
+L1_DATA uint8_t inp_l1 [8192] = {0};
+L1_DATA uint8_t outp_l1 [2048] = {0};
+L1_DATA uint8_t im2col_l1 [IM2COL_DIM] = {0};
+
+#endif
diff --git a/rt_nn_tests/xpnn_maxpool_2b/pulp_nn_kernels.h b/rt_nn_tests/xpnn_maxpool_2b/pulp_nn_kernels.h
new file mode 100644
index 0000000..f4ec977
--- /dev/null
+++ b/rt_nn_tests/xpnn_maxpool_2b/pulp_nn_kernels.h
@@ -0,0 +1,26528 @@
+/*
+ * pulp_nn_kernels.h
+ * Nazareno Bruschi <nazareno.bruschi@unibo.it>
+ *
+ * Copyright (C) 2019-2020 University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef __XPULPNN_KERNELS__
+#define __XPULPNN_KERNELS__
+
+void xpulp_nn_conv_u8_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+#ifndef PROFILE
+                        uint8_t flag_batch_norm);
+#else
+  uint8_t flag_batch_norm,
+    uint32_t *im2col_cycles,
+    uint32_t *hotloop_prep_cycles,
+    uint32_t *hotloop_cycles,
+    uint32_t *requant_cycles,
+    uint32_t *hotloop_leftover_cycles,
+    uint32_t *matmul4x2_leftover_cycles);
+#endif
+                        
+
+void xpulp_nn_conv_i2_i2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_i8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_i8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_i8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_i4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_i4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_i4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_i2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_i2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_i2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_i8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_i8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_i8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_i4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_i4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_i4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_i2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_i2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_i2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_i8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_i8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_i8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_i4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_i4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_i4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_i2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_i2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_i2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u8_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i8_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u8_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i8_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u8_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i8_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u4_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i4_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u4_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i4_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u4_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i4_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u2_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i2_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u2_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i2_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u2_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i2_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u8_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i8_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u8_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i8_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u8_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i8_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u4_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i4_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u4_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i4_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u4_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i4_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u2_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i2_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u2_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i2_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u2_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i2_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u8_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i8_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u8_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i8_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u8_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i8_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u4_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i4_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u4_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i4_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u4_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i4_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u2_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i2_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u2_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i2_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+#ifndef PROFILE
+                        uint8_t flag_batch_norm);
+#else
+  uint8_t flag_batch_norm,
+    uint32_t *hotloop_prep_cycles,
+    uint32_t *hotloop_cycles,
+    uint32_t *requant_cycles,
+    uint32_t *hotloop_leftover_cycles);
+#endif
+
+uint8_t *xpulp_nn_matmul_u2_u2_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i2_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_i8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_i8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_i8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_i4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_i4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_i4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_i2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_i2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_i2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_i8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_i8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_i8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_i4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_i4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_i4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_i2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_i2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_i2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_i8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_i8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_i8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_i4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_i4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_i4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_i2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_i2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_i2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_i32_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_i8_i32_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_u8_i32_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_i8_i32_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_u8_i32_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_i8_i32_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_u4_i32_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_i4_i32_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_u4_i32_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_i4_i32_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_u4_i32_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_i4_i32_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_u2_i32_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_i2_i32_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_u2_i32_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_i2_i32_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_u2_i32_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_i2_i32_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_u8_u8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_i8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_u8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_i8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_u8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_i8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_u4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_i4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_u4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_i4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_u4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_i4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_u2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_i2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_u2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_i2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_u2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_i2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_u8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_i8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_u8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_i8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_u8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_i8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_u4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_i4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_u4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_i4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_u4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_i4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_u2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_i2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_u2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_i2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_u2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_i2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_u8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_i8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_u8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_i8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_u8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_i8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_u4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_i4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_u4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_i4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_u4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_i4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_u2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_i2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_u2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_i2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_u2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_i2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_maxpool_u8(
+                        uint8_t * pIn,
+                        uint8_t * pOut,
+                        uint16_t  dim_im_in_x,
+                        uint16_t  dim_im_in_y,
+                        uint16_t  ch_im_in,
+                        uint16_t  dim_im_out_x,
+                        uint16_t  dim_im_out_y,
+                        uint16_t  dim_kernel_x,
+                        uint16_t  dim_kernel_y,
+                        uint16_t  padding_t,
+                        uint16_t  padding_b,
+                        uint16_t  padding_l,
+                        uint16_t  padding_r,
+                        uint16_t  stride_x,
+                        uint16_t  stride_y);
+
+void xpulp_nn_maxpool_i8(
+                        int8_t * pIn,
+                        int8_t * pOut,
+                        uint16_t  dim_im_in_x,
+                        uint16_t  dim_im_in_y,
+                        uint16_t  ch_im_in,
+                        uint16_t  dim_im_out_x,
+                        uint16_t  dim_im_out_y,
+                        uint16_t  dim_kernel_x,
+                        uint16_t  dim_kernel_y,
+                        uint16_t  padding_t,
+                        uint16_t  padding_b,
+                        uint16_t  padding_l,
+                        uint16_t  padding_r,
+                        uint16_t  stride_x,
+                        uint16_t  stride_y);
+
+void xpulp_nn_maxpool_u4(
+                        uint8_t * pIn,
+                        uint8_t * pOut,
+                        uint16_t  dim_im_in_x,
+                        uint16_t  dim_im_in_y,
+                        uint16_t  ch_im_in,
+                        uint16_t  dim_im_out_x,
+                        uint16_t  dim_im_out_y,
+                        uint16_t  dim_kernel_x,
+                        uint16_t  dim_kernel_y,
+                        uint16_t  padding_t,
+                        uint16_t  padding_b,
+                        uint16_t  padding_l,
+                        uint16_t  padding_r,
+                        uint16_t  stride_x,
+                        uint16_t  stride_y);
+
+void xpulp_nn_maxpool_i4(
+                        int8_t * pIn,
+                        int8_t * pOut,
+                        uint16_t  dim_im_in_x,
+                        uint16_t  dim_im_in_y,
+                        uint16_t  ch_im_in,
+                        uint16_t  dim_im_out_x,
+                        uint16_t  dim_im_out_y,
+                        uint16_t  dim_kernel_x,
+                        uint16_t  dim_kernel_y,
+                        uint16_t  padding_t,
+                        uint16_t  padding_b,
+                        uint16_t  padding_l,
+                        uint16_t  padding_r,
+                        uint16_t  stride_x,
+                        uint16_t  stride_y);
+
+void xpulp_nn_maxpool_u2(
+                        uint8_t * pIn,
+                        uint8_t * pOut,
+                        uint16_t  dim_im_in_x,
+                        uint16_t  dim_im_in_y,
+                        uint16_t  ch_im_in,
+                        uint16_t  dim_im_out_x,
+                        uint16_t  dim_im_out_y,
+                        uint16_t  dim_kernel_x,
+                        uint16_t  dim_kernel_y,
+                        uint16_t  padding_t,
+                        uint16_t  padding_b,
+                        uint16_t  padding_l,
+                        uint16_t  padding_r,
+                        uint16_t  stride_x,
+                        uint16_t  stride_y);
+
+void xpulp_nn_maxpool_i2(
+                        int8_t * pIn,
+                        int8_t * pOut,
+                        uint16_t  dim_im_in_x,
+                        uint16_t  dim_im_in_y,
+                        uint16_t  ch_im_in,
+                        uint16_t  dim_im_out_x,
+                        uint16_t  dim_im_out_y,
+                        uint16_t  dim_kernel_x,
+                        uint16_t  dim_kernel_y,
+                        uint16_t  padding_t,
+                        uint16_t  padding_b,
+                        uint16_t  padding_l,
+                        uint16_t  padding_r,
+                        uint16_t  stride_x,
+                        uint16_t  stride_y);
+
+void xpulp_nn_avgpool_u8_u8(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u8_i8(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i8_u8(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i8_i8(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u8_u4(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u8_i4(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i8_u4(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i8_i4(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u8_u2(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u8_i2(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i8_u2(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i8_i2(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u4_u8(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u4_i8(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i4_u8(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i4_i8(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u4_u4(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u4_i4(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i4_u4(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i4_i4(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u4_u2(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u4_i2(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i4_u2(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i4_i2(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u2_u8(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u2_i8(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i2_u8(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i2_i8(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u2_u4(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u2_i4(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i2_u4(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i2_i4(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u2_u2(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u2_i2(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i2_u2(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i2_i2(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_add_u8_u8(
+                        uint8_t * pIn1,
+                        uint8_t * pIn2,
+                        uint8_t * pOut,
+                        uint16_t out_mult1,
+                        uint16_t out_mult2,
+                        uint16_t out_shift,
+                        uint16_t dim_im_in_x,
+                        uint16_t dim_im_in_y,
+                        uint16_t ch_im_in);
+
+void xpulp_nn_add_u8_u8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u4(
+                        uint8_t * pIn1,
+                        uint8_t * pIn2,
+                        uint8_t * pOut,
+                        uint16_t out_mult1,
+                        uint16_t out_mult2,
+                        uint16_t out_shift,
+                        uint16_t dim_im_in_x,
+                        uint16_t dim_im_in_y,
+                        uint16_t ch_im_in);
+
+void xpulp_nn_add_u8_u4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u2(
+                        uint8_t * pIn1,
+                        uint8_t * pIn2,
+                        uint8_t * pOut,
+                        uint16_t out_mult1,
+                        uint16_t out_mult2,
+                        uint16_t out_shift,
+                        uint16_t dim_im_in_x,
+                        uint16_t dim_im_in_y,
+                        uint16_t ch_im_in);
+
+void xpulp_nn_add_u8_u2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u4(
+                        uint8_t * pIn1,
+                        uint8_t * pIn2,
+                        uint8_t * pOut,
+                        uint16_t out_mult1,
+                        uint16_t out_mult2,
+                        uint16_t out_shift,
+                        uint16_t dim_im_in_x,
+                        uint16_t dim_im_in_y,
+                        uint16_t ch_im_in);
+
+void xpulp_nn_add_u4_u4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u2(
+                        uint8_t * pIn1,
+                        uint8_t * pIn2,
+                        uint8_t * pOut,
+                        uint16_t out_mult1,
+                        uint16_t out_mult2,
+                        uint16_t out_shift,
+                        uint16_t dim_im_in_x,
+                        uint16_t dim_im_in_y,
+                        uint16_t ch_im_in);
+
+void xpulp_nn_add_u4_u2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u2(
+                        uint8_t * pIn1,
+                        uint8_t * pIn2,
+                        uint8_t * pOut,
+                        uint16_t out_mult1,
+                        uint16_t out_mult2,
+                        uint16_t out_shift,
+                        uint16_t dim_im_in_x,
+                        uint16_t dim_im_in_y,
+                        uint16_t ch_im_in);
+
+void xpulp_nn_add_u2_u2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+
+
+#endif
diff --git a/rt_nn_tests/xpnn_maxpool_2b/pulp_nn_mix_kernels.h b/rt_nn_tests/xpnn_maxpool_2b/pulp_nn_mix_kernels.h
new file mode 100644
index 0000000..8b2a3c6
--- /dev/null
+++ b/rt_nn_tests/xpnn_maxpool_2b/pulp_nn_mix_kernels.h
@@ -0,0 +1,7093 @@
+/*
+ * pulp_nn_kernels.h
+ * Nazareno Bruschi <nazareno.bruschi@unibo.it>
+ *
+ * Copyright (C) 2019-2020 University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef __XPULPNN-MIXED_KERNELS__
+#define __XPULPNN-MIXED_KERNELS__
+
+void xpulp_nn_mix_conv_u8_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u8_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u8_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u8_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u4_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u4_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u4_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u2_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u2_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u2_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u8_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u8_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u8_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u4_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u4_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u4_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u2_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u2_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u2_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u8_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u8_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u8_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u4_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u4_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u4_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u2_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u2_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u2_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_avgpool_u8_u8(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u8_i8(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i8_u8(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i8_i8(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u8_u4(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u8_i4(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i8_u4(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i8_i4(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u8_u2(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u8_i2(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i8_u2(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i8_i2(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u4_u8(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u4_i8(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i4_u8(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i4_i8(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u4_u4(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u4_i4(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i4_u4(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i4_i4(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u4_u2(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u4_i2(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i4_u2(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i4_i2(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u2_u8(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u2_i8(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i2_u8(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i2_i8(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u2_u4(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u2_i4(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i2_u4(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i2_i4(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u2_u2(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u2_i2(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i2_u2(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i2_i2(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_add_u8_u8(
+                        uint8_t * pIn1,
+                        uint8_t * pIn2,
+                        uint8_t * pOut,
+                        uint16_t out_mult1,
+                        uint16_t out_mult2,
+                        uint16_t out_shift,
+                        uint16_t dim_im_in_x,
+                        uint16_t dim_im_in_y,
+                        uint16_t ch_im_in);
+
+void xpulp_nn_add_u8_u8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u4(
+                        uint8_t * pIn1,
+                        uint8_t * pIn2,
+                        uint8_t * pOut,
+                        uint16_t out_mult1,
+                        uint16_t out_mult2,
+                        uint16_t out_shift,
+                        uint16_t dim_im_in_x,
+                        uint16_t dim_im_in_y,
+                        uint16_t ch_im_in);
+
+void xpulp_nn_add_u8_u4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u2(
+                        uint8_t * pIn1,
+                        uint8_t * pIn2,
+                        uint8_t * pOut,
+                        uint16_t out_mult1,
+                        uint16_t out_mult2,
+                        uint16_t out_shift,
+                        uint16_t dim_im_in_x,
+                        uint16_t dim_im_in_y,
+                        uint16_t ch_im_in);
+
+void xpulp_nn_add_u8_u2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u4(
+                        uint8_t * pIn1,
+                        uint8_t * pIn2,
+                        uint8_t * pOut,
+                        uint16_t out_mult1,
+                        uint16_t out_mult2,
+                        uint16_t out_shift,
+                        uint16_t dim_im_in_x,
+                        uint16_t dim_im_in_y,
+                        uint16_t ch_im_in);
+
+void xpulp_nn_add_u4_u4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u2(
+                        uint8_t * pIn1,
+                        uint8_t * pIn2,
+                        uint8_t * pOut,
+                        uint16_t out_mult1,
+                        uint16_t out_mult2,
+                        uint16_t out_shift,
+                        uint16_t dim_im_in_x,
+                        uint16_t dim_im_in_y,
+                        uint16_t ch_im_in);
+
+void xpulp_nn_add_u4_u2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u2(
+                        uint8_t * pIn1,
+                        uint8_t * pIn2,
+                        uint8_t * pOut,
+                        uint16_t out_mult1,
+                        uint16_t out_mult2,
+                        uint16_t out_shift,
+                        uint16_t dim_im_in_x,
+                        uint16_t dim_im_in_y,
+                        uint16_t ch_im_in);
+
+void xpulp_nn_add_u2_u2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+
+
+#endif
\ No newline at end of file
diff --git a/rt_nn_tests/xpnn_maxpool_2b/pulp_nn_utils.h b/rt_nn_tests/xpnn_maxpool_2b/pulp_nn_utils.h
new file mode 100644
index 0000000..44d2251
--- /dev/null
+++ b/rt_nn_tests/xpnn_maxpool_2b/pulp_nn_utils.h
@@ -0,0 +1,2079 @@
+/*
+ * pulp_nn_utils.h
+ * Nazareno   Bruschi  <nazareno.bruschi@unibo.it>
+ * Alessandro Nadalini <alessandro.nadalini3@unibo.it>
+ * Georg Rutishauser   <georgr@iis.ee.ethz.ch>
+ *
+ * Copyright (C) 2019-2020 ETH Zurich & University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef __PULPNN_UTILS__
+#define __PULPNN_UTILS__
+
+#include <stdint.h>
+
+typedef   signed short v2s __attribute__((vector_size (4)));
+
+
+
+#define min(a,b)                                             ((a)<(b)?(a):(b))
+#define log2(x)                                              __builtin_pulp_fl1(x)
+#define PACK_INT2_SIZE(x)                                    ((x) >> 2)
+
+#define MemoryFence()                                        asm volatile("":::"memory")
+
+#define CHANS_DECOMPR(x)                                     (5*x >> 2) // equivalent to division by 0.8
+
+/* Functions for Compressed MAC */
+#define CompressedMAC(sum, ptr, config) asm volatile(                \
+    "pv.smlsdotsp.t %[shum], %[phtr], %[chonfig];"                    \
+    : [shum] "+r" (sum), [phtr] "+r" (ptr): [chonfig] "I" (config))
+
+#define CompressedMACUnsigned(sum, ptr, config) asm volatile(                   \
+    "pv.smlsdotsup.t %[shum], %[phtr], %[chonfig];" \
+    : [shum] "+r" (sum), [phtr] "+r" (ptr): [chonfig] "I" (config))
+
+#define InitNNRF(ptr, config) asm volatile(        \
+    "pv.smlsdotsp.t x0, %[phtr], %[chonfig];"       \
+    : [phtr] "+r" (ptr) : [chonfig] "I" (config))
+
+#define ThresholdCompress(res, val, thrs) asm volatile(                                                \
+    "pv.thrc %[rhes], %[vhal], %[thhrs];" : [rhes] "+r" (res) : [vhal] "r" (val), [thhrs] "r" (thrs))
+
+#define GetConfig(a_update, b_update, a_reg, b_reg) a_update << 4 | b_update << 3 | a_reg << 1 | b_reg
+
+/* Functions for threshold&compress */
+#define check_store(res, pOut)            \
+  if ((res & 0xe0000000) == 0x00000000) { \
+    *pOut = res & 0xff;                   \
+    pOut++;                               \
+    incr_val=ch_out_r; }
+
+#define check_store_4x1(res, pOut)        \
+  if ((res & 0xe0000000) == 0x00000000) { \
+    *pOut = res & 0xff;                   \
+    pOut++; }
+
+#define reset_currThr()                           \
+  if ((uint32_t *) currThr == (uint32_t *) (pThr + ch_out)) { \
+    currThr = (v2s *) pThr;                                    \
+  }
+
+#define MacLoads20(a_update, b_update, a_reg, b_reg, ptr, sum)  __builtin_pulp_mlsdotsp20_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+
+#define MacLoad20(a_update, b_update, a_reg, b_reg, ptr, sum)  __builtin_pulp_mlsdotsup20_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+
+/* Functions for compressed min/max */
+#define CompressedMax(res, in1, in2) asm volatile( \
+                                                   "pv.max.t %[res_], %[in1_], %[in2_];" : [res_] "=r" (res) : [in1_] "r" (in1), [in2_] "r" (in2))
+
+#define CompressedMin(res, in1, in2) asm volatile(                      \
+                                                   "pv.min.t %[res_], %[in1_], %[in2_];" : [res_] "=r" (res) : [in1_] "r" (in1), [in2_] "r" (in2))
+
+#define Max16(res, in1, in2) asm volatile(                      \
+                                                   "pv.max.c %[res_], %[in1_], %[in2_];" : [res_] "=r" (res) : [in1_] "r" (in1), [in2_] "r" (in2))
+
+#define Min16(res, in1, in2) asm volatile(                      \
+                                                                        "pv.min.c %[res_], %[in1_], %[in2_];" : [res_] "=r" (res) : [in1_] "r" (in1), [in2_] "r" (in2))
+
+#define thr_cmp(state, val, threshs)  __builtin_pulp_thresh_compr(state, val, threshs)
+
+typedef unsigned char  v4u __attribute__((vector_size (4)));
+typedef   signed char  v4s __attribute__((vector_size (4)));
+
+#define bitext(x,size,off)                                      __builtin_pulp_bextract(x,size,off)
+#define bitextu(x,size,off)                                     __builtin_pulp_bextractu(x,size,off)
+#ifdef __clang__
+#define bitins(dst,not_mask_imm,src,mask_imm,off)               __builtin_binsert(dst,not_mask_imm,src,mask_imm,off)
+#else
+#define bitins(dst,not_mask_imm,src,mask_imm,off)               __builtin_pulp_binsert(dst,not_mask_imm,src,mask_imm,off)
+#endif
+#define pack(x,y,z,t)                                           __builtin_pulp_pack4(x,y,z,t)
+#define max4(a,b)                                               __builtin_pulp_maxu4(a,b)
+#define maxs4(a, b)                                             __builtin_pulp_max4(a, b)
+#define max8(a, b)                                              __builtin_pulp_maxu8(a, b)
+#define maxs8(a, b)                                             __builtin_pulp_max8(a, b)
+#define max16(a, b)                                             __builtin_pulp_maxu16(a, b)
+#define maxs16(a, b)                                            __builtin_pulp_max16(a, b)
+#define maxs20(a, b)                                            __builtin_pulp_max20(a, b)
+#define max32(a,b)                                              __builtin_pulp_maxusi(a,b)
+#define maxs32(a,b)                                             __builtin_pulp_maxsi(a,b)
+#define min32(a,b)                                              __builtin_pulp_minusi(a,b)
+#define mins32(a,b)                                             __builtin_pulp_minsi(a,b)
+#define min4(a, b)                                              __builtin_pulp_minu4(a, b)
+#define mins4(a, b)                                             __builtin_pulp_min4(a, b)
+#define min8(a, b)                                              __builtin_pulp_minu8(a, b)
+#define mins8(a, b)                                             __builtin_pulp_min8(a, b)
+#define min16(a, b)                                             __builtin_pulp_minu16(a, b)
+#define mins16(a, b)                                            __builtin_pulp_min16(a, b)
+#define mins20(a, b)                                            __builtin_pulp_min20(a, b)
+#define avg4(a,b)                                               __builtin_pulp_avgu4(a,b)
+#define avg8(a,b)                                               __builtin_pulp_avgu8(a,b)
+#define avg16(a,b)                                              __builtin_pulp_avgu16(a,b)
+#define log2(x)                                                 __builtin_pulp_fl1(x)
+#define min(a,b)                                                ((a)<(b)?(a):(b))
+#define SumDotp4(a, b, c)                                       __builtin_pulp_sdotusp4(a, b, c)
+#define SumDotp8(a, b, c)                                       __builtin_pulp_sdotusp8(a, b, c)
+#define SumDotp16(a, b, c)                                      __builtin_pulp_sdotusp16(a, b, c)
+#define SumDotps4(a, b, c)                                      __builtin_pulp_sdotsp4(a, b, c)
+#define SumDotps8(a, b, c)                                      __builtin_pulp_sdotsp8(a, b, c)
+#define SumDotps16(a, b, c)                                     __builtin_pulp_sdotsp16(a, b, c)
+#define clip4(x)                                                __builtin_pulp_clipu_r(x, 15)
+#define clip2(x)                                                __builtin_pulp_clipu_r(x, 3)
+#define clip8(x)                                                __builtin_pulp_clipu_r(x, 255)
+
+#define clips4(x)                                               __builtin_pulp_clip_r(x, 7)
+#define clips2(x)                                               __builtin_pulp_clip_r(x, 1)
+#define clips8(x)                                               __builtin_pulp_clip_r(x, 127)
+#define MacLoadInit(a_update, b_update, a_reg, b_reg, ptr)      __builtin_pulp_mlinitspr_v3(a_update, b_update, a_reg, b_reg, ptr)
+#define MacLoadUpdate(ptr)                                      __builtin_pulp_mlupdatespr_v3(ptr)
+#define MacLoadAssign(ptr)                                      __builtin_pulp_mlassignspr_v3(ptr)
+#define MacLoad4(a_update, b_update, a_reg, b_reg, ptr, sum)    __builtin_pulp_mlsdotsup4_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+#define MacLoad8(a_update, b_update, a_reg, b_reg, ptr, sum)    __builtin_pulp_mlsdotsup8_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+#define MacLoad16(a_update, b_update, a_reg, b_reg, ptr, sum)   __builtin_pulp_mlsdotsup16_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+#define MacLoads4(a_update, b_update, a_reg, b_reg, ptr, sum)   __builtin_pulp_mlsdotsp4_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+#define MacLoads8(a_update, b_update, a_reg, b_reg, ptr, sum)   __builtin_pulp_mlsdotsp8_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+#define MacLoads16(a_update, b_update, a_reg, b_reg, ptr, sum)  __builtin_pulp_mlsdotsp16_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+
+#define PACK_INT8_SIZE(x)                                    (x)
+#define PACK_INT4_SIZE(x)                                    ((x) >> 1)
+#define PACK_INT2_SIZE(x)                                    ((x) >> 2)
+
+#define MemoryFence()                                        asm volatile("":::"memory")
+
+#define LEGACY_MODE(x)                                       asm volatile ("csrwi 0x010," x)
+#define IVEC_FMT(x)                                          asm volatile ("csrwi 0x00D," x)
+#define MIXED_SKIP(x)                                        asm volatile ("csrwi 0x00F," x)
+#define A_ADDRESS(x)                                         asm volatile ("csrw 0x100, %0" :: "r" (x))
+#define W_ADDRESS(x)                                         asm volatile ("csrw 0x101, %0" :: "r" (x))
+#define A_STRIDE(x)                                          asm volatile ("csrw 0x102, %0":: "r" (x))
+#define W_STRIDE(x)                                          asm volatile ("csrw 0x103, %0":: "r" (x))
+#define A_ROLLBACK(x)                                        asm volatile ("csrw 0x104, %0":: "r" (x))
+#define W_ROLLBACK(x)                                        asm volatile ("csrw 0x105, %0":: "r" (x))
+#define A_SKIP(x)                                            asm volatile ("csrwi 0x106," x)
+#define W_SKIP(x)                                            asm volatile ("csrwi 0x107," x)
+
+static uint8_t __attribute__((noinline)) pulp_nn_quant_u2 (
+  int32_t phi,
+  int16_t m,
+  int8_t d
+  ) {
+  int32_t x = (m * phi) >> d;
+  uint8_t res = clip2(x);
+  return res;
+}
+static uint8_t __attribute__((noinline)) pulp_nn_bn_quant_u2 (
+  int32_t phi,
+  int32_t k,
+  int32_t lambda,
+  int8_t d
+  ) {
+  int32_t integer_image_phi = (k * phi) + lambda;
+  int32_t x = (integer_image_phi) >> d;
+  uint8_t res = clip2(x);
+  return res;
+}
+static uint8_t __attribute__((noinline)) pulp_nn_add_quant_u2 (
+  uint8_t pix1,
+  uint8_t pix2,
+  int16_t m1,
+  int16_t m2,
+  int8_t d
+  ) {
+  uint32_t integer_image = pix1*m1 + pix2*m2;
+  uint32_t x = (integer_image) >> d;
+  uint8_t res = clip2(x);
+  return res;
+}
+
+static int8_t __attribute__((noinline)) pulp_nn_quant_i2 (
+  int32_t phi,
+  int16_t m,
+  int8_t d
+  ) {
+  int32_t x = (m * phi) >> d;
+  int8_t res = clips2(x);
+  return res;
+}
+static int8_t __attribute__((noinline)) pulp_nn_bn_quant_i2 (
+  int32_t phi,
+  int32_t k,
+  int32_t lambda,
+  int8_t d
+  ) {
+  int32_t integer_image_phi = (k * phi) + lambda;
+  int32_t x = (integer_image_phi) >> d;
+  int8_t res = clips2(x);
+  return res;
+}
+static int8_t __attribute__((noinline)) pulp_nn_add_quant_i2 (
+  int8_t pix1,
+  int8_t pix2,
+  int16_t m1,
+  int16_t m2,
+  int8_t d
+  ) {
+  int32_t integer_image = pix1*m1 + pix2*m2;
+  int32_t x = (integer_image) >> d;
+  int8_t res = clips2(x);
+  return res;
+}
+
+static uint8_t __attribute__((noinline)) pulp_nn_quant_u4 (
+  int32_t phi,
+  int16_t m,
+  int8_t d
+  ) {
+  int32_t x = (m * phi) >> d;
+  uint8_t res = clip4(x);
+  return res;
+}
+static uint8_t __attribute__((noinline)) pulp_nn_bn_quant_u4 (
+  int32_t phi,
+  int32_t k,
+  int32_t lambda,
+  int8_t d
+  ) {
+  int32_t integer_image_phi = (k * phi) + lambda;
+  int32_t x = (integer_image_phi) >> d;
+  uint8_t res = clip4(x);
+  return res;
+}
+static uint8_t __attribute__((noinline)) pulp_nn_add_quant_u4 (
+  uint8_t pix1,
+  uint8_t pix2,
+  int16_t m1,
+  int16_t m2,
+  int8_t d
+  ) {
+  uint32_t integer_image = pix1*m1 + pix2*m2;
+  uint32_t x = (integer_image) >> d;
+  uint8_t res = clip4(x);
+  return res;
+}
+
+static int8_t __attribute__((noinline)) pulp_nn_quant_i4 (
+  int32_t phi,
+  int16_t m,
+  int8_t d
+  ) {
+  int32_t x = (m * phi) >> d;
+  int8_t res = clips4(x);
+  return res;
+}
+static int8_t __attribute__((noinline)) pulp_nn_bn_quant_i4 (
+  int32_t phi,
+  int32_t k,
+  int32_t lambda,
+  int8_t d
+  ) {
+  int32_t integer_image_phi = (k * phi) + lambda;
+  int32_t x = (integer_image_phi) >> d;
+  int8_t res = clips4(x);
+  return res;
+}
+static int8_t __attribute__((noinline)) pulp_nn_add_quant_i4 (
+  int8_t pix1,
+  int8_t pix2,
+  int16_t m1,
+  int16_t m2,
+  int8_t d
+  ) {
+  int32_t integer_image = pix1*m1 + pix2*m2;
+  int32_t x = (integer_image) >> d;
+  int8_t res = clips4(x);
+  return res;
+}
+
+static uint8_t __attribute__((noinline)) pulp_nn_quant_u8 (
+  int32_t phi,
+  int16_t m,
+  int8_t d
+  ) {
+  int32_t x = (m * phi) >> d;
+  uint8_t res = clip8(x);
+  return res;
+}
+static uint8_t __attribute__((noinline)) pulp_nn_bn_quant_u8 (
+  int32_t phi,
+  int32_t k,
+  int32_t lambda,
+  int8_t d
+  ) {
+  int32_t integer_image_phi = (k * phi) + lambda;
+  int32_t x = (integer_image_phi) >> d;
+  uint8_t res = clip8(x);
+  return res;
+}
+static uint8_t __attribute__((noinline)) pulp_nn_add_quant_u8 (
+  uint8_t pix1,
+  uint8_t pix2,
+  int16_t m1,
+  int16_t m2,
+  int8_t d
+  ) {
+  uint32_t integer_image = pix1*m1 + pix2*m2;
+  uint32_t x = (integer_image) >> d;
+  uint8_t res = clip8(x);
+  return res;
+}
+
+static int8_t __attribute__((noinline)) pulp_nn_quant_i8 (
+  int32_t phi,
+  int16_t m,
+  int8_t d
+  ) {
+  int32_t x = (m * phi) >> d;
+  int8_t res = clips8(x);
+  return res;
+}
+static int8_t __attribute__((noinline)) pulp_nn_bn_quant_i8 (
+  int32_t phi,
+  int32_t k,
+  int32_t lambda,
+  int8_t d
+  ) {
+  int32_t integer_image_phi = (k * phi) + lambda;
+  int32_t x = (integer_image_phi) >> d;
+  int8_t res = clips8(x);
+  return res;
+}
+static int8_t __attribute__((noinline)) pulp_nn_add_quant_i8 (
+  int8_t pix1,
+  int8_t pix2,
+  int16_t m1,
+  int16_t m2,
+  int8_t d
+  ) {
+  int32_t integer_image = pix1*m1 + pix2*m2;
+  int32_t x = (integer_image) >> d;
+  int8_t res = clips8(x);
+  return res;
+}
+
+
+static uint8_t __attribute__((noinline)) pulp_nn_u4_quant(int input, int16_t * pThr)
+{
+  if(input <= pThr[7] )
+  {
+    if( input <= pThr[3])
+    {
+      if( input <= pThr[1])
+      {
+        if( input <= pThr[0])
+          return 0;
+        else
+          return 1;
+      }
+      else
+      {
+        if( input <= pThr[2])
+          return 2;
+        else
+          return 3;
+      }
+    }
+    else
+    {
+      if( input <= pThr[5])
+      {
+        if( input <= pThr[4])
+          return 4;
+        else
+          return 5;
+      }
+      else
+      {
+        if( input <= pThr[6])
+          return 6;
+        else
+          return 7;
+      }
+    }
+  }
+  else
+  {
+    if( input <= pThr[11])
+    {
+      if( input <= pThr[9])
+      {
+        if( input <= pThr[8])
+          return 8;
+        else
+          return 9;
+      }
+      else
+      {
+        if( input <= pThr[10])
+          return 10;
+        else
+          return 11;
+      }
+    }
+    else
+    {
+      if( input <= pThr[13])
+      {
+        if( input <= pThr[12])
+          return 12;
+        else
+          return 13;
+      }
+      else
+      {
+        if( input <= pThr[14])
+          return 14;
+        else
+          return 15;
+      }
+    }
+  }
+}
+
+static uint8_t __attribute__((noinline)) pulp_nn_u2_quant(int input, int16_t * pThr)
+{
+  if( input <= pThr[1])
+  {
+    if( input <= pThr[0])
+        {
+      return 0;
+    }
+        else
+        {
+      return 1;
+    }
+  }
+    else
+    {
+    if( input <= pThr[2])
+        {
+      return 2;
+    }
+        else
+        {
+      return 3;
+    }
+  }
+}
+
+/*
+ * Common
+ */
+
+
+static v4s __attribute__((noinline)) pulp_nn_i4_to_i8_r( int8_t *pSrc)
+{
+  v4s Src = *((v4s*) pSrc);
+  int8_t bext1, bext2, bext3, bext4;
+
+  bext1 = (int8_t) bitext((int) Src, 4, 0);
+  bext2 = (int8_t) bitext((int) Src, 4, 4);
+  bext3 = (int8_t) bitext((int) Src, 4, 8);
+  bext4 = (int8_t) bitext((int) Src, 4, 12);
+  v4s res = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+
+  return res;
+}
+
+static v4u __attribute__((noinline)) pulp_nn_u4_to_u8_r(uint8_t *pSrc)
+{
+  v4u Src = *((v4u*) pSrc);
+  uint8_t bext1, bext2, bext3, bext4;
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 4, 0);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 4, 4);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 4, 8);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 4, 12);
+  v4u res = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+
+  return res;
+}
+
+static v4s __attribute__((noinline)) pulp_nn_i2_to_i8_r( int8_t *pSrc)
+{
+  v4s Src = *((v4s*) pSrc);
+  int8_t bext1, bext2, bext3, bext4;
+
+  bext1 = (int8_t) bitext((int) Src, 2, 0);
+  bext2 = (int8_t) bitext((int) Src, 2, 2);
+  bext3 = (int8_t) bitext((int) Src, 2, 4);
+  bext4 = (int8_t) bitext((int) Src, 2, 6);
+  v4s res = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+
+  return res;
+}
+
+static v4u __attribute__((noinline)) pulp_nn_u2_to_u8_r(uint8_t *pSrc)
+{
+  v4u Src = *((v4u*) pSrc);
+  uint8_t bext1, bext2, bext3, bext4;
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 0);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 2);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 4);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 6);
+  v4u res = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+
+  return res;
+}
+
+static v4s __attribute__((noinline)) pulp_nn_i2_to_i4_r(int8_t *pSrc)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  v4s Src = *((v4s*) pSrc);
+  int8_t bext1, bext2, bext3, bext4;
+  int8_t out1, out2, out3, out4;
+
+  bext1 = (int8_t) bitextu((int) Src, 2, 0);
+  bext2 = (int8_t) bitextu((int) Src, 2, 2);
+  bext3 = (int8_t) bitextu((int) Src, 2, 4);
+  bext4 = (int8_t) bitextu((int) Src, 2, 6);
+
+  out1 = bitins(bext1, n_mask, bext2, mask, off);
+  out2 = bitins(bext3, n_mask, bext4, mask, off);
+
+  bext1 = (int8_t) bitextu((int) Src, 2, 8);
+  bext2 = (int8_t) bitextu((int) Src, 2, 10);
+  bext3 = (int8_t) bitextu((int) Src, 2, 12);
+  bext4 = (int8_t) bitextu((int) Src, 2, 14);
+
+  out3 = bitins(bext1, n_mask, bext2, mask, off);
+  out4 = bitins(bext3, n_mask, bext4, mask, off);
+
+  v4s res = pack((int8_t) out1, (int8_t) out2, (int8_t) out3, (int8_t) out4);
+
+  return res;
+}
+
+static v4u __attribute__((noinline)) pulp_nn_u2_to_u4_r(uint8_t *pSrc)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  v4u Src = *((v4u*) pSrc);
+  uint8_t bext1, bext2, bext3, bext4;
+  uint8_t out1, out2, out3, out4;
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 0);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 2);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 4);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 6);
+
+  out1 = bitins(bext1, n_mask, bext2, mask, off);
+  out2 = bitins(bext3, n_mask, bext4, mask, off);
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 8);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 10);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 12);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 14);
+
+  out3 = bitins(bext1, n_mask, bext2, mask, off);
+  out4 = bitins(bext3, n_mask, bext4, mask, off);
+
+  v4u res = pack((uint8_t) out1, (uint8_t) out2, (uint8_t) out3, (uint8_t) out4);
+
+  return res;
+}
+
+static int8_t *__attribute__((always_inline)) pulp_nn_i4_to_i8( int8_t *pSrc, int8_t *pDst)
+{
+  v4s Src = *((v4s*) pSrc);
+  int8_t bext1, bext2, bext3, bext4;
+
+  pSrc+=4;
+
+  bext1 = (int8_t) bitext((int) Src, 4, 0);
+  bext2 = (int8_t) bitext((int) Src, 4, 4);
+  bext3 = (int8_t) bitext((int) Src, 4, 8);
+  bext4 = (int8_t) bitext((int) Src, 4, 12);
+  *((v4s*)pDst) = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+   MemoryFence();
+  pDst+=4;
+  bext1 = (int8_t) bitext((int) Src, 4, 16);
+  bext2 = (int8_t) bitext((int) Src, 4, 20);
+  bext3 = (int8_t) bitext((int) Src, 4, 24);
+  bext4 = (int8_t) bitext((int) Src, 4, 28);
+  *((v4s*)pDst) = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+
+  return pSrc;
+}
+
+static uint8_t *__attribute__((always_inline)) pulp_nn_u4_to_u8(uint8_t *pSrc, uint8_t *pDst)
+{
+  v4u Src = *((v4u*) pSrc);
+  uint8_t bext1, bext2, bext3, bext4;
+
+  pSrc+=4;
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 4, 0);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 4, 4);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 4, 8);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 4, 12);
+  *((v4u*)pDst) = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 4, 16);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 4, 20);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 4, 24);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 4, 28);
+  *((v4u*)pDst) = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+
+  return pSrc;
+}
+
+static int8_t *__attribute__((always_inline)) pulp_nn_i2_to_i8( int8_t * pSrc, int8_t * pDst)
+{
+  v4s Src = *((v4s*) pSrc);
+  int8_t bext1, bext2, bext3, bext4;
+
+  pSrc+=4;
+
+  bext1 = (int8_t) bitext((int) Src, 2, 0);
+  bext2 = (int8_t) bitext((int) Src, 2, 2);
+  bext3 = (int8_t) bitext((int) Src, 2, 4);
+  bext4 = (int8_t) bitext((int) Src, 2, 6);
+  *((v4s*)pDst) = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (int8_t) bitext((int) Src, 2, 8);
+  bext2 = (int8_t) bitext((int) Src, 2, 10);
+  bext3 = (int8_t) bitext((int) Src, 2, 12);
+  bext4 = (int8_t) bitext((int) Src, 2, 14);
+  *((v4s*)pDst) = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (int8_t) bitext((int) Src, 2, 16);
+  bext2 = (int8_t) bitext((int) Src, 2, 18);
+  bext3 = (int8_t) bitext((int) Src, 2, 20);
+  bext4 = (int8_t) bitext((int) Src, 2, 22);
+  *((v4s*)pDst) = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (int8_t) bitext((int) Src, 2, 24);
+  bext2 = (int8_t) bitext((int) Src, 2, 26);
+  bext3 = (int8_t) bitext((int) Src, 2, 28);
+  bext4 = (int8_t) bitext((int) Src, 2, 30);
+  *((v4s*)pDst) = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+
+  return pSrc;
+}
+
+static uint8_t *__attribute__((always_inline)) pulp_nn_u2_to_u8(uint8_t * pSrc, uint8_t * pDst)
+{
+  v4u Src = *((v4u*) pSrc);
+  uint8_t bext1, bext2, bext3, bext4;
+
+  pSrc+=4;
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 0);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 2);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 4);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 6);
+  *((v4u*)pDst) = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 8);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 10);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 12);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 14);
+  *((v4u*)pDst) = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 16);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 18);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 20);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 22);
+  *((v4u*)pDst) = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 24);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 26);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 28);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 30);
+  *((v4u*)pDst) = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+
+  return pSrc;
+}
+
+static int8_t *__attribute__((always_inline)) pulp_nn_i2_to_i4( int8_t * pSrc, int8_t * pDst)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  v4s Src = *((v4s*) pSrc);
+  int8_t bext1, bext2, bext3, bext4;
+  int8_t out1, out2, out3, out4;
+
+  pSrc+=4;
+
+  bext1 = (int8_t) bitext((int) Src, 2, 0);
+  bext2 = (int8_t) bitext((int) Src, 2, 2);
+  bext3 = (int8_t) bitext((int) Src, 2, 4);
+  bext4 = (int8_t) bitext((int) Src, 2, 6);
+
+  out1 = bitins(bext1, n_mask, bext2, mask, off);
+  out2 = bitins(bext3, n_mask, bext4, mask, off);
+  MemoryFence();
+
+  bext1 = (int8_t) bitext((int) Src, 2, 8);
+  bext2 = (int8_t) bitext((int) Src, 2, 10);
+  bext3 = (int8_t) bitext((int) Src, 2, 12);
+  bext4 = (int8_t) bitext((int) Src, 2, 14);
+
+  out3 = bitins(bext1, n_mask, bext2, mask, off);
+  out4 = bitins(bext3, n_mask, bext4, mask, off);
+
+  *((v4s*)pDst) = pack((int8_t) out1, (int8_t) out2, (int8_t) out3, (int8_t) out4);
+  MemoryFence();
+
+  pDst+=4;
+  bext1 = (int8_t) bitext((int) Src, 2, 16);
+  bext2 = (int8_t) bitext((int) Src, 2, 18);
+  bext3 = (int8_t) bitext((int) Src, 2, 20);
+  bext4 = (int8_t) bitext((int) Src, 2, 22);
+
+  out1 = bitins(bext1, n_mask, bext2, mask, off);
+  out2 = bitins(bext3, n_mask, bext4, mask, off);
+  MemoryFence();
+
+  bext1 = (int8_t) bitext((int) Src, 2, 24);
+  bext2 = (int8_t) bitext((int) Src, 2, 26);
+  bext3 = (int8_t) bitext((int) Src, 2, 28);
+  bext4 = (int8_t) bitext((int) Src, 2, 30);
+
+  out3 = bitins(bext1, n_mask, bext2, mask, off);
+  out4 = bitins(bext3, n_mask, bext4, mask, off);
+
+  *((v4s*)pDst) = pack((int8_t) out1, (int8_t) out2, (int8_t) out3, (int8_t) out4);
+
+  return pSrc;
+}
+
+static uint8_t *__attribute__((always_inline)) pulp_nn_u2_to_u4( uint8_t * pSrc, uint8_t * pDst)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  v4u Src = *((v4u*) pSrc);
+  uint8_t bext1, bext2, bext3, bext4;
+  uint8_t out1, out2, out3, out4;
+
+  pSrc+=4;
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 0);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 2);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 4);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 6);
+
+  out1 = bitins(bext1, n_mask, bext2, mask, off);
+  out2 = bitins(bext3, n_mask, bext4, mask, off);
+  MemoryFence();
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 8);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 10);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 12);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 14);
+
+  out3 = bitins(bext1, n_mask, bext2, mask, off);
+  out4 = bitins(bext3, n_mask, bext4, mask, off);
+
+  *((v4u*)pDst) = pack((uint8_t) out1, (uint8_t) out2, (uint8_t) out3, (uint8_t) out4);
+  MemoryFence();
+
+  pDst+=4;
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 16);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 18);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 20);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 22);
+
+  out1 = bitins(bext1, n_mask, bext2, mask, off);
+  out2 = bitins(bext3, n_mask, bext4, mask, off);
+  MemoryFence();
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 24);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 26);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 28);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 30);
+
+  out3 = bitins(bext1, n_mask, bext2, mask, off);
+  out4 = bitins(bext3, n_mask, bext4, mask, off);
+  
+  *((v4u*)pDst) = pack((uint8_t) out1, (uint8_t) out2, (uint8_t) out3, (uint8_t) out4);
+
+  return pSrc;
+}
+
+/*
+ * XpulpV2
+ */
+
+static void __attribute__((noinline)) pulp_zero_mem(uint8_t * pBuffer, unsigned int size)
+{
+  int lfover = size &0x3;
+  for (int i=0; i<(size>>2); i++)
+  {
+    *((v4u *)pBuffer) = (v4u){0,0,0,0};
+    MemoryFence();
+    pBuffer+=4;
+  }
+  while(lfover)
+  {
+    *pBuffer++=0;
+    lfover--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_im2col_u2_to_u8(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_u2_to_u8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 16;
+  }
+  while (lfover)
+  {
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 0);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 2);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 4);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 6);
+    pOutput++;
+    pInput++;
+    lfover -= 4;
+  }
+}
+static void __attribute__((noinline)) pulp_nn_im2col_i2_to_i8(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_i2_to_i8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 16;
+  }
+  while (lfover)
+  {
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 0);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 2);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 4);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 6);
+    pOutput++;
+    pInput++;
+    lfover -= 4;
+  }
+}
+static void __attribute__((noinline)) pulp_nn_im2col_u4_to_u8(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 3u;
+  int lfover = blockSize & 0x07;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_u4_to_u8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 8;
+  }
+  while (lfover)
+  {
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 4, 0);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 4, 4);
+    pOutput++;
+    pInput++;
+    lfover -= 2;
+  }
+}
+static void __attribute__((noinline)) pulp_nn_im2col_i4_to_i8(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 3u;
+  int lfover = blockSize & 0x07;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_i4_to_i8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 8;
+  }
+  while (lfover)
+  {
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 4, 0);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 4, 4);
+    pOutput++;
+    pInput++;
+    lfover -= 2;
+  }
+}
+static void __attribute__((noinline)) pulp_nn_im2col_u8_to_u8(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 2u;
+  int lfover = blockSize & 0x03;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4u*)pOutput) = *((v4u*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    *((uint8_t*)pOutput) = *((uint8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover--;
+  }
+}
+static void __attribute__((noinline)) pulp_nn_im2col_i8_to_i8(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 2u;
+  int lfover = blockSize & 0x03;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4s*)pOutput) = *((v4s*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    *((int8_t*)pOutput) = *((int8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover--;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_im2col_u2_to_u2(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4u*)pOutput) = *((v4u*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    uint8_t extr;
+    *((uint8_t*)pOutput) = *((uint8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover -= 4;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_i2_to_i2(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4s*)pOutput) = *((v4s*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    int8_t extr;
+    *((int8_t*)pOutput) = *((int8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover -= 4;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_u2_to_u4(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_u2_to_u4(pInput, pOutput);
+    MemoryFence();
+    pOutput += 8;
+  }
+  while (lfover)
+  {
+    uint8_t extr;
+    // TODO: this is potentially dangerous/wrong if the number of channels
+    // is not a multiple of 8!
+    *((v4u *)pOutput) = pulp_nn_u2_to_u4_r(pInput);
+    pInput += 2;
+    pOutput += 4;
+    lfover -= 8;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_i2_to_i4(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_i2_to_i4(pInput, pOutput);
+    MemoryFence();
+    pOutput += 8;
+  }
+  while (lfover)
+  {
+    int8_t extr;
+    // TODO: this is potentially dangerous/wrong if the number of channels
+    // is not a multiple of 8!
+    *((v4s *)pOutput) = pulp_nn_i2_to_i4_r(pInput);
+    pInput += 2;
+    pOutput += 4;
+    lfover -= 8;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_u2_to_u8(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_u2_to_u8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 16;
+  }
+  while (lfover)
+  {
+    uint8_t extr;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 0);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 2);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 4);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 6);
+    pOutput++;
+    pInput++;
+    lfover -= 4;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_i2_to_i8(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_i2_to_i8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 16;
+  }
+  while (lfover)
+  {
+    int8_t extr;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 0);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 2);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 4);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 6);
+    pOutput++;
+    pInput++;
+    lfover -= 4;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_u4_to_u4(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 3u;
+  int lfover = blockSize & 0x07;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4u*)pOutput) = *((v4u*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    uint8_t extr;
+    *((uint8_t*)pOutput) = *((uint8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover -= 2;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_i4_to_i4(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 3u;
+  int lfover = blockSize & 0x07;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4s*)pOutput) = *((v4s*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    int8_t extr;
+    *((int8_t*)pOutput) = *((int8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover -= 2;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_u4_to_u8(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 3u;
+  int lfover = blockSize & 0x07;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_u4_to_u8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 8;
+  }
+  while (lfover)
+  {
+    uint8_t extr;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 4, 0);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 4, 4);
+    pOutput++;
+    pInput++;
+    lfover -= 2;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_i4_to_i8(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 3u;
+  int lfover = blockSize & 0x07;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_i4_to_i8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 8;
+  }
+  while (lfover)
+  {
+    int8_t extr;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 4, 0);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 4, 4);
+    pOutput++;
+    pInput++;
+    lfover -= 2;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_u8_to_u8(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 2u;
+  int lfover = blockSize & 0x03;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4u*)pOutput) = *((v4u*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    uint8_t extr;
+    *((uint8_t*)pOutput) = *((uint8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover -= 1;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_i8_to_i8(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 2u;
+  int lfover = blockSize & 0x03;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4s*)pOutput) = *((v4s*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    int8_t extr;
+    *((int8_t*)pOutput) = *((int8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover -= 1;
+  }
+}
+
+
+
+static void __attribute__((noinline)) pulp_nn_compare_and_replace_if_larger_u8(uint8_t * base,
+                                                uint8_t * target,
+                                                uint16_t length)
+{
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+  v4u inp;
+  v4u com;
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    inp = *((v4u*)pIn);
+    com = *((v4u*)pCom);
+
+    *((v4u*)pIn) = max4(inp, com);
+
+    pCom+=4;
+    pIn+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    if(*pIn<*pCom)
+      *pIn=*pCom;
+    
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_compare_and_replace_if_larger_i8(
+    int8_t * base, int8_t * target, uint16_t length) {
+  int8_t *pIn = base;
+  int8_t *pCom = target;
+  v4s inp;
+  v4s com;
+  int cnt = length >> 2;
+
+  while (cnt > 0u) {
+    inp = *((v4s *)pIn);
+    com = *((v4s *)pCom);
+
+    *((v4s *)pIn) = maxs4(inp, com);
+
+    pCom += 4;
+    pIn += 4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left > 0u) {
+    if (*pIn < *pCom)
+      *pIn = *pCom;
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_avg_and_replace_u8(uint8_t * base,
+                                  uint8_t * target,
+                                  uint16_t length)
+{
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+
+  while (length>0u)
+  {
+    *pIn = ((*pIn + *pCom) >> 1);
+
+    pIn++;
+    pCom++;
+    length--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_compare_and_replace_if_larger_u4(uint8_t * base,
+                                                uint8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+  v4u inp[2];
+  v4u com[2];
+  uint8_t *out;
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    pulp_nn_u4_to_u8(pIn, (uint8_t *)inp);
+    pulp_nn_u4_to_u8(pCom, (uint8_t *)com);
+
+    *((v4u *)out) = max4(inp[0], com[0]);
+
+    *((uint8_t*)pIn) = bitins(*out, n_mask, *(out + 1), mask, off);
+    pIn++;
+    *((uint8_t*)pIn) = bitins(*(out + 2), n_mask, *(out + 3), mask, off);
+    pIn++;
+
+    *((v4u *)out) = max4(inp[1], com[1]);
+
+    *((uint8_t*)pIn) = bitins(*out, n_mask, *(out + 1), mask, off);
+    pIn++;
+    *((uint8_t*)pIn) = bitins(*(out + 2), n_mask, *(out + 3), mask, off);
+    pIn++;
+
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 4, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 4, 4);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 4, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 4, 4);
+
+    if(inA0<inB0)
+      inA0=inB0;
+
+    if(inA1<inB1)
+      inA1=inB1;
+
+    *((uint8_t*)pIn) = bitins(inA0, n_mask, inA1, mask, off);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_compare_and_replace_if_larger_i4(int8_t * base,
+                                                int8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  int8_t *pIn = base;
+  int8_t *pCom = target;
+  v4s inp[2];
+  v4s com[2];
+  int8_t *out;
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    pulp_nn_i4_to_i8(pIn, (int8_t *)inp);
+    pulp_nn_i4_to_i8(pCom, (int8_t *)com);
+
+    *((v4s *)out) = maxs4(inp[0], com[0]);
+
+    *((int8_t*)pIn) = bitins(*out, n_mask, *(out + 1), mask, off);
+    pIn++;
+    *((int8_t*)pIn) = bitins(*(out + 2), n_mask, *(out + 3), mask, off);
+    pIn++;
+
+    *((v4s *)out) = maxs4(inp[1], com[1]);
+
+    *((int8_t*)pIn) = bitins(*out, n_mask, *(out + 1), mask, off);
+    pIn++;
+    *((int8_t*)pIn) = bitins(*(out + 2), n_mask, *(out + 3), mask, off);
+    pIn++;
+
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    int8_t inA0 = (int8_t) bitext((int) *pIn, 4, 0);
+    int8_t inA1 = (int8_t) bitext((int) *pIn, 4, 4);
+    int8_t inB0 = (int8_t) bitext((int) *pCom, 4, 0);
+    int8_t inB1 = (int8_t) bitext((int) *pCom, 4, 4);
+
+    if(inA0<inB0)
+      inA0=inB0;
+
+    if(inA1<inB1)
+      inA1=inB1;
+
+    *((int8_t*)pIn) = bitins(inA0, n_mask, inA1, mask, off);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_avg_and_replace_u4(uint8_t * base,
+                                  uint8_t * target,
+                                  uint16_t length)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+
+  while (length>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 4, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 4, 4);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 4, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 4, 4);
+
+    inA0 = ((inA0 + inB0) >> 1);
+    inA1 = ((inA1 + inB1) >> 1);
+
+    *((uint8_t*)pIn) = bitins(inA0, n_mask, inA1, mask, off);
+
+    pIn++;
+    pCom++;
+    length--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_compare_and_replace_if_larger_u2(uint8_t * base,
+                                                uint8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask2 = 0x0c;
+  int8_t n_mask2 = ~ mask2;
+  int8_t mask4 = 0x30;
+  int8_t n_mask4 = ~ mask4;
+  int8_t mask6 = 0xc0;
+  int8_t n_mask6 = ~ mask6;
+  int8_t off2 = 2;
+  int8_t off4 = 4;
+  int8_t off6 = 6;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+  v4u inp[4];
+  v4u com[4];
+  uint8_t *out;
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    pulp_nn_u2_to_u8(pIn, inp);
+    pulp_nn_u2_to_u8(pCom, com);
+
+    *((v4u*)out) = max4(inp[0], com[0]);
+
+    uint8_t inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    *((v4u*)out) = max4(inp[1], com[1]);
+
+    inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    *((v4u*)out) = max4(inp[2], com[2]);
+
+    inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    *((v4u*)out) = max4(inp[3], com[3]);
+
+    inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 2, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 2, 2);
+    uint8_t inA2 = (uint8_t) bitextu((unsigned int) *pIn, 2, 4);
+    uint8_t inA3 = (uint8_t) bitextu((unsigned int) *pIn, 2, 6);
+    v4u inA4 = pack((uint8_t) inA0, (uint8_t) inA1, (uint8_t) inA2, (uint8_t) inA3);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 2, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 2, 2);
+    uint8_t inB2 = (uint8_t) bitextu((unsigned int) *pCom, 2, 4);
+    uint8_t inB3 = (uint8_t) bitextu((unsigned int) *pCom, 2, 6);
+    v4u inB4 = pack((uint8_t) inB0, (uint8_t) inB1, (uint8_t) inB2, (uint8_t) inB3);
+
+    *((v4u*)out) = max4(inA4, inB4);
+
+    uint8_t inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+
+static void __attribute__((noinline)) pulp_nn_compare_and_replace_if_larger_i2(int8_t * base,
+                                                int8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask2 = 0x0c;
+  int8_t n_mask2 = ~ mask2;
+  int8_t mask4 = 0x30;
+  int8_t n_mask4 = ~ mask4;
+  int8_t mask6 = 0xc0;
+  int8_t n_mask6 = ~ mask6;
+  int8_t off2 = 2;
+  int8_t off4 = 4;
+  int8_t off6 = 6;
+
+  int8_t *pIn = base;
+  int8_t *pCom = target;
+  v4s inp[4];
+  v4s com[4];
+  int8_t *out;
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    pulp_nn_i2_to_i8(pIn, inp);
+    pulp_nn_i2_to_i8(pCom, com);
+
+    *((v4s*)out) = maxs4(inp[0], com[0]);
+
+    int8_t inA = (int8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((int8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    *((v4s*)out) = maxs4(inp[1], com[1]);
+
+    inA = (int8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((int8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    *((v4s*)out) = maxs4(inp[2], com[2]);
+
+    inA = (int8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((int8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    *((v4s*)out) = maxs4(inp[3], com[3]);
+
+    inA = (int8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((int8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    int8_t inA0 = (int8_t) bitext((unsigned int) *pIn, 2, 0);
+    int8_t inA1 = (int8_t) bitext((unsigned int) *pIn, 2, 2);
+    int8_t inA2 = (int8_t) bitext((unsigned int) *pIn, 2, 4);
+    int8_t inA3 = (int8_t) bitext((unsigned int) *pIn, 2, 6);
+    v4s inA4 = pack((int8_t) inA0, (int8_t) inA1, (int8_t) inA2, (int8_t) inA3);
+    int8_t inB0 = (int8_t) bitext((unsigned int) *pCom, 2, 0);
+    int8_t inB1 = (int8_t) bitext((unsigned int) *pCom, 2, 2);
+    int8_t inB2 = (int8_t) bitext((unsigned int) *pCom, 2, 4);
+    int8_t inB3 = (int8_t) bitext((unsigned int) *pCom, 2, 6);
+    v4s inB4 = pack((int8_t) inB0, (int8_t) inB1, (int8_t) inB2, (int8_t) inB3);
+
+    *((v4s*)out) = maxs4(inA4, inB4);
+
+    int8_t inA = (int8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((int8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_avg_and_replace_u2(uint8_t * base,
+                                  uint8_t * target,
+                                  uint16_t length)
+{
+  int8_t mask2 = 0x0c;
+  int8_t n_mask2 = ~ mask2;
+  int8_t mask4 = 0x30;
+  int8_t n_mask4 = ~ mask4;
+  int8_t mask6 = 0xc0;
+  int8_t n_mask6 = ~ mask6;
+  int8_t off2 = 2;
+  int8_t off4 = 4;
+  int8_t off6 = 6;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+
+  while (length>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 2, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 2, 2);
+    uint8_t inA2 = (uint8_t) bitextu((unsigned int) *pIn, 2, 4);
+    uint8_t inA3 = (uint8_t) bitextu((unsigned int) *pIn, 2, 6);
+
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 2, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 2, 2);
+    uint8_t inB2 = (uint8_t) bitextu((unsigned int) *pCom, 2, 4);
+    uint8_t inB3 = (uint8_t) bitextu((unsigned int) *pCom, 2, 6);
+
+    inA0 = ((inA0 + inB0) >> 1);
+    inA1 = ((inA1 + inB1) >> 1);
+    inA2 = ((inA2 + inB2) >> 1);
+    inA3 = ((inA3 + inB3) >> 1);
+
+    uint8_t inA = (uint8_t) bitins(inA0, n_mask2, inA1, mask2, off2);
+    inA = bitins(inA, n_mask4, inA2, mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, inA3, mask6, off6);
+
+    pIn++;
+    pCom++;
+    length--;
+  }
+}
+
+/*
+ * XpulpNN
+ */
+
+static void __attribute__((noinline)) xpulp_nn_zero_mem_u8(uint8_t * pBuffer, unsigned int size)
+{
+  int lfover = size &0x3;
+  for (int i=0; i<(size>>2); i++)
+  {
+    *((v4u *)pBuffer) = (v4u){0,0,0,0};
+    MemoryFence();
+    pBuffer+=4;
+  }
+  while(lfover)
+  {
+    *pBuffer++=0;
+    lfover--;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_zero_mem_u4(uint8_t * pBuffer, unsigned int size)
+{
+  int lfover = size &0x7;
+  for (int i=0; i<(size>>3); i++)
+  {
+    *((v4u *)pBuffer) = (v4u){0,0,0,0};
+    MemoryFence();
+    pBuffer+=4;
+  }
+  while(lfover)
+  {
+    *pBuffer++=0;
+    lfover-=2;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_zero_mem_u2(uint8_t * pBuffer, unsigned int size)
+{
+  int lfover = size &0xf;
+  for (int i=0; i<(size>>4); i++)
+  {
+    *((v4u *)pBuffer) = (v4u){0,0,0,0};
+    MemoryFence();
+    pBuffer+=4;
+  }
+  while(lfover)
+  {
+    *pBuffer++=0;
+    lfover-=4;
+  }
+}
+
+
+static void __attribute__((noinline)) xpulp_tnn_zero_mem_ternary(uint8_t * pBuffer, unsigned int size, unsigned int uns)
+{
+  uint8_t pad_val = 0xd9;
+  uint32_t pad_vec = 0xd9d9d9d9;
+  if (uns) {
+    // if we are using an unsigned kernel, we need to pad with -1 because the hardware will add a +1 to ALL values!
+    pad_val = 0xff;
+    pad_vec = 0xffffffff;
+  }
+  int lfover = size &0xf;
+  for (int i=0; i<(size>>4); i++)
+    {
+      *((v4u *)pBuffer) = (v4u)pad_vec;
+      MemoryFence();
+      pBuffer+=4;
+    }
+  while(lfover)
+    {
+      *pBuffer++=pad_val;
+      lfover-=4;
+    }
+}
+
+static void __attribute__((noinline)) xpulp_nn_compare_and_replace_if_larger_u4(uint8_t * base,
+                                                uint8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    *((uint32_t *)pIn) = max8(*((uint32_t *)pIn), *((uint32_t *)pCom));
+
+    pIn+=4;
+    pCom+=4;
+
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 4, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 4, 4);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 4, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 4, 4);
+
+    if(inA0<inB0)
+      inA0=inB0;
+
+    if(inA1<inB1)
+      inA1=inB1;
+
+    *((uint8_t*)pIn) = bitins(inA0, n_mask, inA1, mask, off);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_compare_and_replace_if_larger_i4(
+    int8_t * base, int8_t * target, uint16_t length) {
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~mask;
+  int8_t off = 0x04;
+
+  int8_t *pIn = base;
+  int8_t *pCom = target;
+
+  int cnt = length >> 2;
+
+  while (cnt > 0u) {
+    *((int32_t *)pIn) = maxs8(*((int32_t *)pIn), *((int32_t *)pCom));
+
+    pIn += 4;
+    pCom += 4;
+
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left > 0u) {
+    int8_t inA0 = (int8_t)bitext((int)*pIn, 4, 0);
+    int8_t inA1 = (int8_t)bitext((int)*pIn, 4, 4);
+    int8_t inB0 = (int8_t)bitext((int)*pCom, 4, 0);
+    int8_t inB1 = (int8_t)bitext((int)*pCom, 4, 4);
+
+    if (inA0 < inB0)
+      inA0 = inB0;
+
+    if (inA1 < inB1)
+      inA1 = inB1;
+
+    *((int8_t *)pIn) = bitins(inA0, n_mask, inA1, mask, off);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_avg_and_replace_u4(uint8_t * base,
+                                  uint8_t * target,
+                                  uint16_t length)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+
+  int cnt = length >> 2;
+
+  while (cnt > 0u)
+  {
+    *((uint32_t *)pIn) = avg8(*((uint32_t *)pIn), *((uint32_t *)pCom));
+
+    pIn+=4;
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+
+  while (left>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 4, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 4, 4);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 4, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 4, 4);
+
+    inA0 = ((inA0 + inB0) >> 1);
+    inA1 = ((inA1 + inB1) >> 1);
+
+    *((uint8_t*)pIn) = bitins(inA0, n_mask, inA1, mask, off);
+
+    pIn++;
+    pCom++;
+    length--;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_compare_and_replace_if_larger_u2(uint8_t * base,
+                                                uint8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask2 = 0x0c;
+  int8_t n_mask2 = ~ mask2;
+  int8_t mask4 = 0x30;
+  int8_t n_mask4 = ~ mask4;
+  int8_t mask6 = 0xc0;
+  int8_t n_mask6 = ~ mask6;
+  int8_t off2 = 2;
+  int8_t off4 = 4;
+  int8_t off6 = 6;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+  uint8_t *out;
+
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    *((uint32_t *)pIn) = max16(*((uint32_t *)pIn), *((uint32_t *)pCom));
+
+    pIn+=4;
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 2, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 2, 2);
+    uint8_t inA2 = (uint8_t) bitextu((unsigned int) *pIn, 2, 4);
+    uint8_t inA3 = (uint8_t) bitextu((unsigned int) *pIn, 2, 6);
+    v4u inA4 = pack((uint8_t) inA0, (uint8_t) inA1, (uint8_t) inA2, (uint8_t) inA3);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 2, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 2, 2);
+    uint8_t inB2 = (uint8_t) bitextu((unsigned int) *pCom, 2, 4);
+    uint8_t inB3 = (uint8_t) bitextu((unsigned int) *pCom, 2, 6);
+    v4u inB4 = pack((uint8_t) inB0, (uint8_t) inB1, (uint8_t) inB2, (uint8_t) inB3);
+
+    *((v4u*)out) = max4(inA4, inB4);
+
+    uint8_t inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_compare_and_replace_if_larger_i2(int8_t * base,
+                                                int8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask2 = 0x0c;
+  int8_t n_mask2 = ~ mask2;
+  int8_t mask4 = 0x30;
+  int8_t n_mask4 = ~ mask4;
+  int8_t mask6 = 0xc0;
+  int8_t n_mask6 = ~ mask6;
+  int8_t off2 = 2;
+  int8_t off4 = 4;
+  int8_t off6 = 6;
+
+  int8_t *pIn = base;
+  int8_t *pCom = target;
+  int8_t *out;
+
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    *((int32_t *)pIn) = maxs16(*((int32_t *)pIn), *((int32_t *)pCom));
+
+    pIn+=4;
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    int8_t inA0 = (int8_t) bitext((int) *pIn, 2, 0);
+    int8_t inA1 = (int8_t) bitext((int) *pIn, 2, 2);
+    int8_t inA2 = (int8_t) bitext((int) *pIn, 2, 4);
+    int8_t inA3 = (int8_t) bitext((int) *pIn, 2, 6);
+    v4s inA4 = pack((int8_t) inA0, (int8_t) inA1, (int8_t) inA2, (int8_t) inA3);
+    int8_t inB0 = (int8_t) bitext((int) *pCom, 2, 0);
+    int8_t inB1 = (int8_t) bitext((int) *pCom, 2, 2);
+    int8_t inB2 = (int8_t) bitext((int) *pCom, 2, 4);
+    int8_t inB3 = (int8_t) bitext((int) *pCom, 2, 6);
+    v4s inB4 = pack((int8_t) inB0, (int8_t) inB1, (int8_t) inB2, (int8_t) inB3);
+
+    *((v4s*)out) = maxs4(inA4, inB4);
+
+    int8_t inA = (int8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((int8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_tnn_compare_and_replace_if_larger_ternary(int8_t * base,
+                                                int8_t * target,
+                                                uint16_t length)
+{
+  uint8_t mask2 = 0x0c;
+  uint8_t n_mask2 = ~ mask2;
+  uint8_t mask4 = 0x30;
+  uint8_t n_mask4 = ~ mask4;
+  uint8_t mask6 = 0xc0;
+  uint8_t n_mask6 = ~ mask6;
+  uint8_t off2 = 2;
+  uint8_t off4 = 4;
+  uint8_t off6 = 6;
+
+  uint8_t *pIn = (uint8_t *) base;
+  uint8_t *pCom = (uint8_t *) target;
+  uint8_t *out;
+
+  int cnt = length >> 2;
+  uint32_t result;
+
+  while(cnt > 0u)
+  {
+    uint32_t in1 = *((uint32_t *)pIn);
+    uint32_t in2 = *((int32_t *)pCom);
+    result = maxs20(in1, in2);
+    *((uint32_t *)pIn) = result;
+
+    pIn+=4;
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  if (left>0u)
+  {
+    // do the vector max on the whole word - we won't use the leftover bytes
+    uint32_t in1 = *((uint32_t *)pIn);
+    uint32_t in2 = *((int32_t *)pCom);
+    result = maxs20(in1, in2);
+
+    // ...and copy back the relevant bytes of the result to pIn
+    for (int i=0; i<left; i++)
+      *((uint8_t*)(pIn + i)) = (uint8_t) (result >> (8*i));
+
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_avg_and_replace_u2(uint8_t * base,
+                                  uint8_t * target,
+                                  uint16_t length)
+{
+  int8_t mask2 = 0x0c;
+  int8_t n_mask2 = ~ mask2;
+  int8_t mask4 = 0x30;
+  int8_t n_mask4 = ~ mask4;
+  int8_t mask6 = 0xc0;
+  int8_t n_mask6 = ~ mask6;
+  int8_t off2 = 2;
+  int8_t off4 = 4;
+  int8_t off6 = 6;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+  uint8_t *out;
+
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    *((uint32_t *)pIn) = avg16(*((uint32_t *)pIn), *((uint32_t *)pCom));
+
+    pIn+=4;
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 2, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 2, 2);
+    uint8_t inA2 = (uint8_t) bitextu((unsigned int) *pIn, 2, 4);
+    uint8_t inA3 = (uint8_t) bitextu((unsigned int) *pIn, 2, 6);
+    v4u inA4 = pack((uint8_t) inA0, (uint8_t) inA1, (uint8_t) inA2, (uint8_t) inA3);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 2, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 2, 2);
+    uint8_t inB2 = (uint8_t) bitextu((unsigned int) *pCom, 2, 4);
+    uint8_t inB3 = (uint8_t) bitextu((unsigned int) *pCom, 2, 6);
+    v4u inB4 = pack((uint8_t) inB0, (uint8_t) inB1, (uint8_t) inB2, (uint8_t) inB3);
+
+    *((v4u*)out) = avg4(inA4, inB4);
+
+    uint8_t inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+#endif
diff --git a/rt_nn_tests/xpnn_maxpool_2b/pulp_nn_utils_xpnn.h b/rt_nn_tests/xpnn_maxpool_2b/pulp_nn_utils_xpnn.h
new file mode 100644
index 0000000..0c783ae
--- /dev/null
+++ b/rt_nn_tests/xpnn_maxpool_2b/pulp_nn_utils_xpnn.h
@@ -0,0 +1,1937 @@
+/*
+ * pulp_nn_utils.h
+ * Nazareno   Bruschi  <nazareno.bruschi@unibo.it>
+ * Alessandro Nadalini <alessandro.nadalini3@unibo.it>
+ * Georg Rutishauser   <georgr@iis.ee.ethz.ch>
+ *
+ * Copyright (C) 2019-2020 ETH Zurich & University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef __PULPNN_UTILS__
+#define __PULPNN_UTILS__
+
+#include "pmsis.h"
+#ifdef GAP_SDK
+#include "pulp.h"
+#endif
+
+#define bitext(x,size,off)                                      __builtin_pulp_bextract(x,size,off)
+#define bitextu(x,size,off)                                     __builtin_pulp_bextractu(x,size,off)
+#define bitins(dst,not_mask_imm,src,mask_imm,off)               __builtin_pulp_binsert(dst,not_mask_imm,src,mask_imm,off)
+#define pack(x,y,z,t)                                           __builtin_pulp_pack4(x,y,z,t)
+#define max4(a,b)                                               __builtin_pulp_maxu4(a,b)
+#define maxs4(a, b)                                             __builtin_pulp_max4(a, b)
+#define max8(a, b)                                              __builtin_pulp_maxu8(a, b)
+#define maxs8(a, b)                                             __builtin_pulp_max8(a, b)
+#define max16(a, b)                                             __builtin_pulp_maxu16(a, b)
+#define maxs16(a, b)                                            __builtin_pulp_max16(a, b)
+#define max32(a,b)                                              __builtin_pulp_maxusi(a,b)
+#define maxs32(a,b)                                             __builtin_pulp_maxsi(a,b)
+#define min32(a,b)                                              __builtin_pulp_minusi(a,b)
+#define mins32(a,b)                                             __builtin_pulp_minsi(a,b)
+#define min4(a, b)                                              __builtin_pulp_minu4(a, b)
+#define mins4(a, b)                                             __builtin_pulp_min4(a, b)
+#define min8(a, b)                                              __builtin_pulp_minu8(a, b)
+#define mins8(a, b)                                             __builtin_pulp_min8(a, b)
+#define min16(a, b)                                             __builtin_pulp_minu16(a, b)
+#define mins16(a, b)                                            __builtin_pulp_min16(a, b)
+#define avg4(a,b)                                               __builtin_pulp_avgu4(a,b)
+#define avg8(a,b)                                               __builtin_pulp_avgu8(a,b)
+#define avg16(a,b)                                              __builtin_pulp_avgu16(a,b)
+#define log2(x)                                                 __builtin_pulp_fl1(x)
+#define min(a,b)                                                ((a)<(b)?(a):(b))
+#define SumDotp4(a, b, c)                                       __builtin_pulp_sdotusp4(a, b, c)
+#define SumDotp8(a, b, c)                                       __builtin_pulp_sdotusp8(a, b, c)
+#define SumDotp16(a, b, c)                                      __builtin_pulp_sdotusp16(a, b, c)
+#define SumDotps4(a, b, c)                                      __builtin_pulp_sdotsp4(a, b, c)
+#define SumDotps8(a, b, c)                                      __builtin_pulp_sdotsp8(a, b, c)
+#define SumDotps16(a, b, c)                                     __builtin_pulp_sdotsp16(a, b, c)
+#define clip4(x)                                                __builtin_pulp_clipu_r(x, 15)
+#define clip2(x)                                                __builtin_pulp_clipu_r(x, 3)
+#define clip8(x)                                                __builtin_pulp_clipu_r(x, 255)
+
+#define clips4(x)                                               __builtin_pulp_clip_r(x, 7)
+#define clips2(x)                                               __builtin_pulp_clip_r(x, 1)
+#define clips8(x)                                               __builtin_pulp_clip_r(x, 127)
+#define MacLoadInit(a_update, b_update, a_reg, b_reg, ptr)      __builtin_pulp_mlinitspr_v3(a_update, b_update, a_reg, b_reg, ptr)
+#define MacLoadUpdate(ptr)                                      __builtin_pulp_mlupdatespr_v3(ptr)
+#define MacLoadAssign(ptr)                                      __builtin_pulp_mlassignspr_v3(ptr)
+#define MacLoad4(a_update, b_update, a_reg, b_reg, ptr, sum)    __builtin_pulp_mlsdotsup4_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+#define MacLoad8(a_update, b_update, a_reg, b_reg, ptr, sum)    __builtin_pulp_mlsdotsup8_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+#define MacLoad16(a_update, b_update, a_reg, b_reg, ptr, sum)   __builtin_pulp_mlsdotsup16_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+#define MacLoads4(a_update, b_update, a_reg, b_reg, ptr, sum)   __builtin_pulp_mlsdotsp4_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+#define MacLoads8(a_update, b_update, a_reg, b_reg, ptr, sum)   __builtin_pulp_mlsdotsp8_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+#define MacLoads16(a_update, b_update, a_reg, b_reg, ptr, sum)  __builtin_pulp_mlsdotsp16_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+
+#define PACK_INT8_SIZE(x)                                    (x)
+#define PACK_INT4_SIZE(x)                                    ((x) >> 1)
+#define PACK_INT2_SIZE(x)                                    ((x) >> 2)
+
+#define MemoryFence()                                        asm volatile("":::"memory")
+
+#define LEGACY_MODE(x)                                       asm volatile ("csrwi 0x010," x)
+#define IVEC_FMT(x)                                          asm volatile ("csrwi 0x00D," x)
+#define MIXED_SKIP(x)                                        asm volatile ("csrwi 0x00F," x)
+#define A_ADDRESS(x)                                         asm volatile ("csrw 0x100, %0" :: "r" (x))
+#define W_ADDRESS(x)                                         asm volatile ("csrw 0x101, %0" :: "r" (x))
+#define A_STRIDE(x)                                          asm volatile ("csrw 0x102, %0":: "r" (x))
+#define W_STRIDE(x)                                          asm volatile ("csrw 0x103, %0":: "r" (x))
+#define A_ROLLBACK(x)                                        asm volatile ("csrw 0x104, %0":: "r" (x))
+#define W_ROLLBACK(x)                                        asm volatile ("csrw 0x105, %0":: "r" (x))
+#define A_SKIP(x)                                            asm volatile ("csrwi 0x106," x)
+#define W_SKIP(x)                                            asm volatile ("csrwi 0x107," x)
+
+static uint8_t __attribute__((noinline)) pulp_nn_quant_u2 (
+  int32_t phi,
+  int16_t m,
+  int8_t d
+  ) {
+  int32_t x = (m * phi) >> d;
+  uint8_t res = clip2(x);
+  return res;
+}
+static uint8_t __attribute__((noinline)) pulp_nn_bn_quant_u2 (
+  int32_t phi,
+  int32_t k,
+  int32_t lambda,
+  int8_t d
+  ) {
+  int32_t integer_image_phi = (k * phi) + lambda;
+  int32_t x = (integer_image_phi) >> d;
+  uint8_t res = clip2(x);
+  return res;
+}
+static uint8_t __attribute__((noinline)) pulp_nn_add_quant_u2 (
+  uint8_t pix1,
+  uint8_t pix2,
+  int16_t m1,
+  int16_t m2,
+  int8_t d
+  ) {
+  uint32_t integer_image = pix1*m1 + pix2*m2;
+  uint32_t x = (integer_image) >> d;
+  uint8_t res = clip2(x);
+  return res;
+}
+
+static int8_t __attribute__((noinline)) pulp_nn_quant_i2 (
+  int32_t phi,
+  int16_t m,
+  int8_t d
+  ) {
+  int32_t x = (m * phi) >> d;
+  int8_t res = clips2(x);
+  return res;
+}
+static int8_t __attribute__((noinline)) pulp_nn_bn_quant_i2 (
+  int32_t phi,
+  int32_t k,
+  int32_t lambda,
+  int8_t d
+  ) {
+  int32_t integer_image_phi = (k * phi) + lambda;
+  int32_t x = (integer_image_phi) >> d;
+  int8_t res = clips2(x);
+  return res;
+}
+static int8_t __attribute__((noinline)) pulp_nn_add_quant_i2 (
+  int8_t pix1,
+  int8_t pix2,
+  int16_t m1,
+  int16_t m2,
+  int8_t d
+  ) {
+  int32_t integer_image = pix1*m1 + pix2*m2;
+  int32_t x = (integer_image) >> d;
+  int8_t res = clips2(x);
+  return res;
+}
+
+static uint8_t __attribute__((noinline)) pulp_nn_quant_u4 (
+  int32_t phi,
+  int16_t m,
+  int8_t d
+  ) {
+  int32_t x = (m * phi) >> d;
+  uint8_t res = clip4(x);
+  return res;
+}
+static uint8_t __attribute__((noinline)) pulp_nn_bn_quant_u4 (
+  int32_t phi,
+  int32_t k,
+  int32_t lambda,
+  int8_t d
+  ) {
+  int32_t integer_image_phi = (k * phi) + lambda;
+  int32_t x = (integer_image_phi) >> d;
+  uint8_t res = clip4(x);
+  return res;
+}
+static uint8_t __attribute__((noinline)) pulp_nn_add_quant_u4 (
+  uint8_t pix1,
+  uint8_t pix2,
+  int16_t m1,
+  int16_t m2,
+  int8_t d
+  ) {
+  uint32_t integer_image = pix1*m1 + pix2*m2;
+  uint32_t x = (integer_image) >> d;
+  uint8_t res = clip4(x);
+  return res;
+}
+
+static int8_t __attribute__((noinline)) pulp_nn_quant_i4 (
+  int32_t phi,
+  int16_t m,
+  int8_t d
+  ) {
+  int32_t x = (m * phi) >> d;
+  int8_t res = clips4(x);
+  return res;
+}
+static int8_t __attribute__((noinline)) pulp_nn_bn_quant_i4 (
+  int32_t phi,
+  int32_t k,
+  int32_t lambda,
+  int8_t d
+  ) {
+  int32_t integer_image_phi = (k * phi) + lambda;
+  int32_t x = (integer_image_phi) >> d;
+  int8_t res = clips4(x);
+  return res;
+}
+static int8_t __attribute__((noinline)) pulp_nn_add_quant_i4 (
+  int8_t pix1,
+  int8_t pix2,
+  int16_t m1,
+  int16_t m2,
+  int8_t d
+  ) {
+  int32_t integer_image = pix1*m1 + pix2*m2;
+  int32_t x = (integer_image) >> d;
+  int8_t res = clips4(x);
+  return res;
+}
+
+static uint8_t __attribute__((noinline)) pulp_nn_quant_u8 (
+  int32_t phi,
+  int16_t m,
+  int8_t d
+  ) {
+  int32_t x = (m * phi) >> d;
+  uint8_t res = clip8(x);
+  return res;
+}
+static uint8_t __attribute__((noinline)) pulp_nn_bn_quant_u8 (
+  int32_t phi,
+  int32_t k,
+  int32_t lambda,
+  int8_t d
+  ) {
+  int32_t integer_image_phi = (k * phi) + lambda;
+  int32_t x = (integer_image_phi) >> d;
+  uint8_t res = clip8(x);
+  return res;
+}
+static uint8_t __attribute__((noinline)) pulp_nn_add_quant_u8 (
+  uint8_t pix1,
+  uint8_t pix2,
+  int16_t m1,
+  int16_t m2,
+  int8_t d
+  ) {
+  uint32_t integer_image = pix1*m1 + pix2*m2;
+  uint32_t x = (integer_image) >> d;
+  uint8_t res = clip8(x);
+  return res;
+}
+
+static int8_t __attribute__((noinline)) pulp_nn_quant_i8 (
+  int32_t phi,
+  int16_t m,
+  int8_t d
+  ) {
+  int32_t x = (m * phi) >> d;
+  int8_t res = clips8(x);
+  return res;
+}
+static int8_t __attribute__((noinline)) pulp_nn_bn_quant_i8 (
+  int32_t phi,
+  int32_t k,
+  int32_t lambda,
+  int8_t d
+  ) {
+  int32_t integer_image_phi = (k * phi) + lambda;
+  int32_t x = (integer_image_phi) >> d;
+  int8_t res = clips8(x);
+  return res;
+}
+static int8_t __attribute__((noinline)) pulp_nn_add_quant_i8 (
+  int8_t pix1,
+  int8_t pix2,
+  int16_t m1,
+  int16_t m2,
+  int8_t d
+  ) {
+  int32_t integer_image = pix1*m1 + pix2*m2;
+  int32_t x = (integer_image) >> d;
+  int8_t res = clips8(x);
+  return res;
+}
+
+
+static uint8_t __attribute__((noinline)) pulp_nn_u4_quant(int input, int16_t * pThr)
+{
+  if(input <= pThr[7] )
+  {
+    if( input <= pThr[3])
+    {
+      if( input <= pThr[1])
+      {
+        if( input <= pThr[0])
+          return 0;
+        else
+          return 1;
+      }
+      else
+      {
+        if( input <= pThr[2])
+          return 2;
+        else
+          return 3;
+      }
+    }
+    else
+    {
+      if( input <= pThr[5])
+      {
+        if( input <= pThr[4])
+          return 4;
+        else
+          return 5;
+      }
+      else
+      {
+        if( input <= pThr[6])
+          return 6;
+        else
+          return 7;
+      }
+    }
+  }
+  else
+  {
+    if( input <= pThr[11])
+    {
+      if( input <= pThr[9])
+      {
+        if( input <= pThr[8])
+          return 8;
+        else
+          return 9;
+      }
+      else
+      {
+        if( input <= pThr[10])
+          return 10;
+        else
+          return 11;
+      }
+    }
+    else
+    {
+      if( input <= pThr[13])
+      {
+        if( input <= pThr[12])
+          return 12;
+        else
+          return 13;
+      }
+      else
+      {
+        if( input <= pThr[14])
+          return 14;
+        else
+          return 15;
+      }
+    }
+  }
+}
+
+static uint8_t __attribute__((noinline)) pulp_nn_u2_quant(int input, int16_t * pThr)
+{
+  if( input <= pThr[1])
+  {
+    if( input <= pThr[0])
+        {
+      return 0;
+    }
+        else
+        {
+      return 1;
+    }
+  }
+    else
+    {
+    if( input <= pThr[2])
+        {
+      return 2;
+    }
+        else
+        {
+      return 3;
+    }
+  }
+}
+
+/*
+ * Common
+ */
+
+
+static v4s __attribute__((noinline)) pulp_nn_i4_to_i8_r( int8_t *pSrc)
+{
+  v4s Src = *((v4s*) pSrc);
+  int8_t bext1, bext2, bext3, bext4;
+
+  bext1 = (int8_t) bitext((int) Src, 4, 0);
+  bext2 = (int8_t) bitext((int) Src, 4, 4);
+  bext3 = (int8_t) bitext((int) Src, 4, 8);
+  bext4 = (int8_t) bitext((int) Src, 4, 12);
+  v4s res = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+
+  return res;
+}
+
+static v4u __attribute__((noinline)) pulp_nn_u4_to_u8_r(uint8_t *pSrc)
+{
+  v4u Src = *((v4u*) pSrc);
+  uint8_t bext1, bext2, bext3, bext4;
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 4, 0);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 4, 4);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 4, 8);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 4, 12);
+  v4u res = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+
+  return res;
+}
+
+static v4s __attribute__((noinline)) pulp_nn_i2_to_i8_r( int8_t *pSrc)
+{
+  v4s Src = *((v4s*) pSrc);
+  int8_t bext1, bext2, bext3, bext4;
+
+  bext1 = (int8_t) bitext((int) Src, 2, 0);
+  bext2 = (int8_t) bitext((int) Src, 2, 2);
+  bext3 = (int8_t) bitext((int) Src, 2, 4);
+  bext4 = (int8_t) bitext((int) Src, 2, 6);
+  v4s res = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+
+  return res;
+}
+
+static v4u __attribute__((noinline)) pulp_nn_u2_to_u8_r(uint8_t *pSrc)
+{
+  v4u Src = *((v4u*) pSrc);
+  uint8_t bext1, bext2, bext3, bext4;
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 0);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 2);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 4);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 6);
+  v4u res = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+
+  return res;
+}
+
+static v4s __attribute__((noinline)) pulp_nn_i2_to_i4_r(int8_t *pSrc)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  v4s Src = *((v4s*) pSrc);
+  int8_t bext1, bext2, bext3, bext4;
+  int8_t out1, out2, out3, out4;
+
+  bext1 = (int8_t) bitextu((int) Src, 2, 0);
+  bext2 = (int8_t) bitextu((int) Src, 2, 2);
+  bext3 = (int8_t) bitextu((int) Src, 2, 4);
+  bext4 = (int8_t) bitextu((int) Src, 2, 6);
+
+  out1 = bitins(bext1, n_mask, bext2, mask, off);
+  out2 = bitins(bext3, n_mask, bext4, mask, off);
+
+  bext1 = (int8_t) bitextu((int) Src, 2, 8);
+  bext2 = (int8_t) bitextu((int) Src, 2, 10);
+  bext3 = (int8_t) bitextu((int) Src, 2, 12);
+  bext4 = (int8_t) bitextu((int) Src, 2, 14);
+
+  out3 = bitins(bext1, n_mask, bext2, mask, off);
+  out4 = bitins(bext3, n_mask, bext4, mask, off);
+
+  v4s res = pack((int8_t) out1, (int8_t) out2, (int8_t) out3, (int8_t) out4);
+
+  return res;
+}
+
+static v4u __attribute__((noinline)) pulp_nn_u2_to_u4_r(uint8_t *pSrc)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  v4u Src = *((v4u*) pSrc);
+  uint8_t bext1, bext2, bext3, bext4;
+  uint8_t out1, out2, out3, out4;
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 0);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 2);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 4);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 6);
+
+  out1 = bitins(bext1, n_mask, bext2, mask, off);
+  out2 = bitins(bext3, n_mask, bext4, mask, off);
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 8);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 10);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 12);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 14);
+
+  out3 = bitins(bext1, n_mask, bext2, mask, off);
+  out4 = bitins(bext3, n_mask, bext4, mask, off);
+
+  v4u res = pack((uint8_t) out1, (uint8_t) out2, (uint8_t) out3, (uint8_t) out4);
+
+  return res;
+}
+
+static int8_t *__attribute__((always_inline)) pulp_nn_i4_to_i8( int8_t *pSrc, int8_t *pDst)
+{
+  v4s Src = *((v4s*) pSrc);
+  int8_t bext1, bext2, bext3, bext4;
+
+  pSrc+=4;
+
+  bext1 = (int8_t) bitext((int) Src, 4, 0);
+  bext2 = (int8_t) bitext((int) Src, 4, 4);
+  bext3 = (int8_t) bitext((int) Src, 4, 8);
+  bext4 = (int8_t) bitext((int) Src, 4, 12);
+  *((v4s*)pDst) = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+   MemoryFence();
+  pDst+=4;
+  bext1 = (int8_t) bitext((int) Src, 4, 16);
+  bext2 = (int8_t) bitext((int) Src, 4, 20);
+  bext3 = (int8_t) bitext((int) Src, 4, 24);
+  bext4 = (int8_t) bitext((int) Src, 4, 28);
+  *((v4s*)pDst) = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+
+  return pSrc;
+}
+
+static uint8_t *__attribute__((always_inline)) pulp_nn_u4_to_u8(uint8_t *pSrc, uint8_t *pDst)
+{
+  v4u Src = *((v4u*) pSrc);
+  uint8_t bext1, bext2, bext3, bext4;
+
+  pSrc+=4;
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 4, 0);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 4, 4);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 4, 8);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 4, 12);
+  *((v4u*)pDst) = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 4, 16);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 4, 20);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 4, 24);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 4, 28);
+  *((v4u*)pDst) = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+
+  return pSrc;
+}
+
+static int8_t *__attribute__((always_inline)) pulp_nn_i2_to_i8( int8_t * pSrc, int8_t * pDst)
+{
+  v4s Src = *((v4s*) pSrc);
+  int8_t bext1, bext2, bext3, bext4;
+
+  pSrc+=4;
+
+  bext1 = (int8_t) bitext((int) Src, 2, 0);
+  bext2 = (int8_t) bitext((int) Src, 2, 2);
+  bext3 = (int8_t) bitext((int) Src, 2, 4);
+  bext4 = (int8_t) bitext((int) Src, 2, 6);
+  *((v4s*)pDst) = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (int8_t) bitext((int) Src, 2, 8);
+  bext2 = (int8_t) bitext((int) Src, 2, 10);
+  bext3 = (int8_t) bitext((int) Src, 2, 12);
+  bext4 = (int8_t) bitext((int) Src, 2, 14);
+  *((v4s*)pDst) = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (int8_t) bitext((int) Src, 2, 16);
+  bext2 = (int8_t) bitext((int) Src, 2, 18);
+  bext3 = (int8_t) bitext((int) Src, 2, 20);
+  bext4 = (int8_t) bitext((int) Src, 2, 22);
+  *((v4s*)pDst) = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (int8_t) bitext((int) Src, 2, 24);
+  bext2 = (int8_t) bitext((int) Src, 2, 26);
+  bext3 = (int8_t) bitext((int) Src, 2, 28);
+  bext4 = (int8_t) bitext((int) Src, 2, 30);
+  *((v4s*)pDst) = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+
+  return pSrc;
+}
+
+static uint8_t *__attribute__((always_inline)) pulp_nn_u2_to_u8(uint8_t * pSrc, uint8_t * pDst)
+{
+  v4u Src = *((v4u*) pSrc);
+  uint8_t bext1, bext2, bext3, bext4;
+
+  pSrc+=4;
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 0);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 2);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 4);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 6);
+  *((v4u*)pDst) = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 8);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 10);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 12);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 14);
+  *((v4u*)pDst) = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 16);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 18);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 20);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 22);
+  *((v4u*)pDst) = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 24);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 26);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 28);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 30);
+  *((v4u*)pDst) = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+
+  return pSrc;
+}
+
+static int8_t *__attribute__((always_inline)) pulp_nn_i2_to_i4( int8_t * pSrc, int8_t * pDst)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  v4s Src = *((v4s*) pSrc);
+  int8_t bext1, bext2, bext3, bext4;
+  int8_t out1, out2, out3, out4;
+
+  pSrc+=4;
+
+  bext1 = (int8_t) bitext((int) Src, 2, 0);
+  bext2 = (int8_t) bitext((int) Src, 2, 2);
+  bext3 = (int8_t) bitext((int) Src, 2, 4);
+  bext4 = (int8_t) bitext((int) Src, 2, 6);
+
+  out1 = bitins(bext1, n_mask, bext2, mask, off);
+  out2 = bitins(bext3, n_mask, bext4, mask, off);
+  MemoryFence();
+
+  bext1 = (int8_t) bitext((int) Src, 2, 8);
+  bext2 = (int8_t) bitext((int) Src, 2, 10);
+  bext3 = (int8_t) bitext((int) Src, 2, 12);
+  bext4 = (int8_t) bitext((int) Src, 2, 14);
+
+  out3 = bitins(bext1, n_mask, bext2, mask, off);
+  out4 = bitins(bext3, n_mask, bext4, mask, off);
+
+  *((v4s*)pDst) = pack((int8_t) out1, (int8_t) out2, (int8_t) out3, (int8_t) out4);
+  MemoryFence();
+
+  pDst+=4;
+  bext1 = (int8_t) bitext((int) Src, 2, 16);
+  bext2 = (int8_t) bitext((int) Src, 2, 18);
+  bext3 = (int8_t) bitext((int) Src, 2, 20);
+  bext4 = (int8_t) bitext((int) Src, 2, 22);
+
+  out1 = bitins(bext1, n_mask, bext2, mask, off);
+  out2 = bitins(bext3, n_mask, bext4, mask, off);
+  MemoryFence();
+
+  bext1 = (int8_t) bitext((int) Src, 2, 24);
+  bext2 = (int8_t) bitext((int) Src, 2, 26);
+  bext3 = (int8_t) bitext((int) Src, 2, 28);
+  bext4 = (int8_t) bitext((int) Src, 2, 30);
+
+  out3 = bitins(bext1, n_mask, bext2, mask, off);
+  out4 = bitins(bext3, n_mask, bext4, mask, off);
+
+  *((v4s*)pDst) = pack((int8_t) out1, (int8_t) out2, (int8_t) out3, (int8_t) out4);
+
+  return pSrc;
+}
+
+static uint8_t *__attribute__((always_inline)) pulp_nn_u2_to_u4( uint8_t * pSrc, uint8_t * pDst)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  v4u Src = *((v4u*) pSrc);
+  uint8_t bext1, bext2, bext3, bext4;
+  uint8_t out1, out2, out3, out4;
+
+  pSrc+=4;
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 0);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 2);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 4);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 6);
+
+  out1 = bitins(bext1, n_mask, bext2, mask, off);
+  out2 = bitins(bext3, n_mask, bext4, mask, off);
+  MemoryFence();
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 8);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 10);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 12);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 14);
+
+  out3 = bitins(bext1, n_mask, bext2, mask, off);
+  out4 = bitins(bext3, n_mask, bext4, mask, off);
+
+  *((v4u*)pDst) = pack((uint8_t) out1, (uint8_t) out2, (uint8_t) out3, (uint8_t) out4);
+  MemoryFence();
+
+  pDst+=4;
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 16);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 18);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 20);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 22);
+
+  out1 = bitins(bext1, n_mask, bext2, mask, off);
+  out2 = bitins(bext3, n_mask, bext4, mask, off);
+  MemoryFence();
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 24);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 26);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 28);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 30);
+
+  out3 = bitins(bext1, n_mask, bext2, mask, off);
+  out4 = bitins(bext3, n_mask, bext4, mask, off);
+  
+  *((v4u*)pDst) = pack((uint8_t) out1, (uint8_t) out2, (uint8_t) out3, (uint8_t) out4);
+
+  return pSrc;
+}
+
+/*
+ * XpulpV2
+ */
+
+static void __attribute__((noinline)) pulp_zero_mem(uint8_t * pBuffer, unsigned int size)
+{
+  int lfover = size &0x3;
+  for (int i=0; i<(size>>2); i++)
+  {
+    *((v4u *)pBuffer) = (v4u){0,0,0,0};
+    MemoryFence();
+    pBuffer+=4;
+  }
+  while(lfover)
+  {
+    *pBuffer++=0;
+    lfover--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_im2col_u2_to_u8(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_u2_to_u8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 16;
+  }
+  while (lfover)
+  {
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 0);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 2);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 4);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 6);
+    pOutput++;
+    pInput++;
+    lfover -= 4;
+  }
+}
+static void __attribute__((noinline)) pulp_nn_im2col_i2_to_i8(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_i2_to_i8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 16;
+  }
+  while (lfover)
+  {
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 0);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 2);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 4);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 6);
+    pOutput++;
+    pInput++;
+    lfover -= 4;
+  }
+}
+static void __attribute__((noinline)) pulp_nn_im2col_u4_to_u8(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 3u;
+  int lfover = blockSize & 0x07;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_u4_to_u8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 8;
+  }
+  while (lfover)
+  {
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 4, 0);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 4, 4);
+    pOutput++;
+    pInput++;
+    lfover -= 2;
+  }
+}
+static void __attribute__((noinline)) pulp_nn_im2col_i4_to_i8(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 3u;
+  int lfover = blockSize & 0x07;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_i4_to_i8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 8;
+  }
+  while (lfover)
+  {
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 4, 0);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 4, 4);
+    pOutput++;
+    pInput++;
+    lfover -= 2;
+  }
+}
+static void __attribute__((noinline)) pulp_nn_im2col_u8_to_u8(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 2u;
+  int lfover = blockSize & 0x03;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4u*)pOutput) = *((v4u*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    *((uint8_t*)pOutput) = *((uint8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover--;
+  }
+}
+static void __attribute__((noinline)) pulp_nn_im2col_i8_to_i8(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 2u;
+  int lfover = blockSize & 0x03;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4s*)pOutput) = *((v4s*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    *((int8_t*)pOutput) = *((int8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover--;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_im2col_u2_to_u2(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4u*)pOutput) = *((v4u*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    uint8_t extr;
+    *((uint8_t*)pOutput) = *((uint8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover -= 4;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_i2_to_i2(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4s*)pOutput) = *((v4s*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    int8_t extr;
+    *((int8_t*)pOutput) = *((int8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover -= 4;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_u2_to_u4(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_u2_to_u4(pInput, pOutput);
+    MemoryFence();
+    pOutput += 8;
+  }
+  while (lfover)
+  {
+    uint8_t extr;
+    // TODO: this is potentially dangerous/wrong if the number of channels
+    // is not a multiple of 8!
+    *((v4u *)pOutput) = pulp_nn_u2_to_u4_r(pInput);
+    pInput += 2;
+    pOutput += 4;
+    lfover -= 8;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_i2_to_i4(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_i2_to_i4(pInput, pOutput);
+    MemoryFence();
+    pOutput += 8;
+  }
+  while (lfover)
+  {
+    int8_t extr;
+    // TODO: this is potentially dangerous/wrong if the number of channels
+    // is not a multiple of 8!
+    *((v4s *)pOutput) = pulp_nn_i2_to_i4_r(pInput);
+    pInput += 2;
+    pOutput += 4;
+    lfover -= 8;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_u2_to_u8(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_u2_to_u8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 16;
+  }
+  while (lfover)
+  {
+    uint8_t extr;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 0);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 2);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 4);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 6);
+    pOutput++;
+    pInput++;
+    lfover -= 4;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_i2_to_i8(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_i2_to_i8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 16;
+  }
+  while (lfover)
+  {
+    int8_t extr;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 0);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 2);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 4);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 6);
+    pOutput++;
+    pInput++;
+    lfover -= 4;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_u4_to_u4(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 3u;
+  int lfover = blockSize & 0x07;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4u*)pOutput) = *((v4u*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    uint8_t extr;
+    *((uint8_t*)pOutput) = *((uint8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover -= 2;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_i4_to_i4(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 3u;
+  int lfover = blockSize & 0x07;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4s*)pOutput) = *((v4s*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    int8_t extr;
+    *((int8_t*)pOutput) = *((int8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover -= 2;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_u4_to_u8(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 3u;
+  int lfover = blockSize & 0x07;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_u4_to_u8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 8;
+  }
+  while (lfover)
+  {
+    uint8_t extr;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 4, 0);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 4, 4);
+    pOutput++;
+    pInput++;
+    lfover -= 2;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_i4_to_i8(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 3u;
+  int lfover = blockSize & 0x07;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_i4_to_i8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 8;
+  }
+  while (lfover)
+  {
+    int8_t extr;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 4, 0);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 4, 4);
+    pOutput++;
+    pInput++;
+    lfover -= 2;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_u8_to_u8(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 2u;
+  int lfover = blockSize & 0x03;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4u*)pOutput) = *((v4u*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    uint8_t extr;
+    *((uint8_t*)pOutput) = *((uint8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover -= 1;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_i8_to_i8(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 2u;
+  int lfover = blockSize & 0x03;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4s*)pOutput) = *((v4s*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    int8_t extr;
+    *((int8_t*)pOutput) = *((int8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover -= 1;
+  }
+}
+
+
+
+static void __attribute__((noinline)) pulp_nn_compare_and_replace_if_larger_u8(uint8_t * base,
+                                                uint8_t * target,
+                                                uint16_t length)
+{
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+  v4u inp;
+  v4u com;
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    inp = *((v4u*)pIn);
+    com = *((v4u*)pCom);
+
+    *((v4u*)pIn) = max4(inp, com);
+
+    pCom+=4;
+    pIn+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    if(*pIn<*pCom)
+      *pIn=*pCom;
+    
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_compare_and_replace_if_larger_i8(
+    int8_t * base, int8_t * target, uint16_t length) {
+  int8_t *pIn = base;
+  int8_t *pCom = target;
+  v4s inp;
+  v4s com;
+  int cnt = length >> 2;
+
+  while (cnt > 0u) {
+    inp = *((v4s *)pIn);
+    com = *((v4s *)pCom);
+
+    *((v4s *)pIn) = maxs4(inp, com);
+
+    pCom += 4;
+    pIn += 4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left > 0u) {
+    if (*pIn < *pCom)
+      *pIn = *pCom;
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_avg_and_replace_u8(uint8_t * base,
+                                  uint8_t * target,
+                                  uint16_t length)
+{
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+
+  while (length>0u)
+  {
+    *pIn = ((*pIn + *pCom) >> 1);
+
+    pIn++;
+    pCom++;
+    length--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_compare_and_replace_if_larger_u4(uint8_t * base,
+                                                uint8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+  v4u inp[2];
+  v4u com[2];
+  uint8_t *out;
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    pulp_nn_u4_to_u8(pIn, (uint8_t *)inp);
+    pulp_nn_u4_to_u8(pCom, (uint8_t *)com);
+
+    *((v4u *)out) = max4(inp[0], com[0]);
+
+    *((uint8_t*)pIn) = bitins(*out, n_mask, *(out + 1), mask, off);
+    pIn++;
+    *((uint8_t*)pIn) = bitins(*(out + 2), n_mask, *(out + 3), mask, off);
+    pIn++;
+
+    *((v4u *)out) = max4(inp[1], com[1]);
+
+    *((uint8_t*)pIn) = bitins(*out, n_mask, *(out + 1), mask, off);
+    pIn++;
+    *((uint8_t*)pIn) = bitins(*(out + 2), n_mask, *(out + 3), mask, off);
+    pIn++;
+
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 4, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 4, 4);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 4, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 4, 4);
+
+    if(inA0<inB0)
+      inA0=inB0;
+
+    if(inA1<inB1)
+      inA1=inB1;
+
+    *((uint8_t*)pIn) = bitins(inA0, n_mask, inA1, mask, off);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_compare_and_replace_if_larger_i4(int8_t * base,
+                                                int8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  int8_t *pIn = base;
+  int8_t *pCom = target;
+  v4s inp[2];
+  v4s com[2];
+  int8_t *out;
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    pulp_nn_i4_to_i8(pIn, (int8_t *)inp);
+    pulp_nn_i4_to_i8(pCom, (int8_t *)com);
+
+    *((v4s *)out) = maxs4(inp[0], com[0]);
+
+    *((int8_t*)pIn) = bitins(*out, n_mask, *(out + 1), mask, off);
+    pIn++;
+    *((int8_t*)pIn) = bitins(*(out + 2), n_mask, *(out + 3), mask, off);
+    pIn++;
+
+    *((v4s *)out) = maxs4(inp[1], com[1]);
+
+    *((int8_t*)pIn) = bitins(*out, n_mask, *(out + 1), mask, off);
+    pIn++;
+    *((int8_t*)pIn) = bitins(*(out + 2), n_mask, *(out + 3), mask, off);
+    pIn++;
+
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    int8_t inA0 = (int8_t) bitext((int) *pIn, 4, 0);
+    int8_t inA1 = (int8_t) bitext((int) *pIn, 4, 4);
+    int8_t inB0 = (int8_t) bitext((int) *pCom, 4, 0);
+    int8_t inB1 = (int8_t) bitext((int) *pCom, 4, 4);
+
+    if(inA0<inB0)
+      inA0=inB0;
+
+    if(inA1<inB1)
+      inA1=inB1;
+
+    *((int8_t*)pIn) = bitins(inA0, n_mask, inA1, mask, off);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_avg_and_replace_u4(uint8_t * base,
+                                  uint8_t * target,
+                                  uint16_t length)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+
+  while (length>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 4, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 4, 4);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 4, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 4, 4);
+
+    inA0 = ((inA0 + inB0) >> 1);
+    inA1 = ((inA1 + inB1) >> 1);
+
+    *((uint8_t*)pIn) = bitins(inA0, n_mask, inA1, mask, off);
+
+    pIn++;
+    pCom++;
+    length--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_compare_and_replace_if_larger_u2(uint8_t * base,
+                                                uint8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask2 = 0x0c;
+  int8_t n_mask2 = ~ mask2;
+  int8_t mask4 = 0x30;
+  int8_t n_mask4 = ~ mask4;
+  int8_t mask6 = 0xc0;
+  int8_t n_mask6 = ~ mask6;
+  int8_t off2 = 2;
+  int8_t off4 = 4;
+  int8_t off6 = 6;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+  v4u inp[4];
+  v4u com[4];
+  uint8_t *out;
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    pulp_nn_u2_to_u8(pIn, inp);
+    pulp_nn_u2_to_u8(pCom, com);
+
+    *((v4u*)out) = max4(inp[0], com[0]);
+
+    uint8_t inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    *((v4u*)out) = max4(inp[1], com[1]);
+
+    inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    *((v4u*)out) = max4(inp[2], com[2]);
+
+    inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    *((v4u*)out) = max4(inp[3], com[3]);
+
+    inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 2, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 2, 2);
+    uint8_t inA2 = (uint8_t) bitextu((unsigned int) *pIn, 2, 4);
+    uint8_t inA3 = (uint8_t) bitextu((unsigned int) *pIn, 2, 6);
+    v4u inA4 = pack((uint8_t) inA0, (uint8_t) inA1, (uint8_t) inA2, (uint8_t) inA3);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 2, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 2, 2);
+    uint8_t inB2 = (uint8_t) bitextu((unsigned int) *pCom, 2, 4);
+    uint8_t inB3 = (uint8_t) bitextu((unsigned int) *pCom, 2, 6);
+    v4u inB4 = pack((uint8_t) inB0, (uint8_t) inB1, (uint8_t) inB2, (uint8_t) inB3);
+
+    *((v4u*)out) = max4(inA4, inB4);
+
+    uint8_t inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+
+static void __attribute__((noinline)) pulp_nn_compare_and_replace_if_larger_i2(int8_t * base,
+                                                int8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask2 = 0x0c;
+  int8_t n_mask2 = ~ mask2;
+  int8_t mask4 = 0x30;
+  int8_t n_mask4 = ~ mask4;
+  int8_t mask6 = 0xc0;
+  int8_t n_mask6 = ~ mask6;
+  int8_t off2 = 2;
+  int8_t off4 = 4;
+  int8_t off6 = 6;
+
+  int8_t *pIn = base;
+  int8_t *pCom = target;
+  v4s inp[4];
+  v4s com[4];
+  int8_t *out;
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    pulp_nn_i2_to_i8(pIn, inp);
+    pulp_nn_i2_to_i8(pCom, com);
+
+    *((v4s*)out) = maxs4(inp[0], com[0]);
+
+    int8_t inA = (int8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((int8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    *((v4s*)out) = maxs4(inp[1], com[1]);
+
+    inA = (int8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((int8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    *((v4s*)out) = maxs4(inp[2], com[2]);
+
+    inA = (int8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((int8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    *((v4s*)out) = maxs4(inp[3], com[3]);
+
+    inA = (int8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((int8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    int8_t inA0 = (int8_t) bitext((unsigned int) *pIn, 2, 0);
+    int8_t inA1 = (int8_t) bitext((unsigned int) *pIn, 2, 2);
+    int8_t inA2 = (int8_t) bitext((unsigned int) *pIn, 2, 4);
+    int8_t inA3 = (int8_t) bitext((unsigned int) *pIn, 2, 6);
+    v4s inA4 = pack((int8_t) inA0, (int8_t) inA1, (int8_t) inA2, (int8_t) inA3);
+    int8_t inB0 = (int8_t) bitext((unsigned int) *pCom, 2, 0);
+    int8_t inB1 = (int8_t) bitext((unsigned int) *pCom, 2, 2);
+    int8_t inB2 = (int8_t) bitext((unsigned int) *pCom, 2, 4);
+    int8_t inB3 = (int8_t) bitext((unsigned int) *pCom, 2, 6);
+    v4s inB4 = pack((int8_t) inB0, (int8_t) inB1, (int8_t) inB2, (int8_t) inB3);
+
+    *((v4s*)out) = maxs4(inA4, inB4);
+
+    int8_t inA = (int8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((int8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_avg_and_replace_u2(uint8_t * base,
+                                  uint8_t * target,
+                                  uint16_t length)
+{
+  int8_t mask2 = 0x0c;
+  int8_t n_mask2 = ~ mask2;
+  int8_t mask4 = 0x30;
+  int8_t n_mask4 = ~ mask4;
+  int8_t mask6 = 0xc0;
+  int8_t n_mask6 = ~ mask6;
+  int8_t off2 = 2;
+  int8_t off4 = 4;
+  int8_t off6 = 6;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+
+  while (length>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 2, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 2, 2);
+    uint8_t inA2 = (uint8_t) bitextu((unsigned int) *pIn, 2, 4);
+    uint8_t inA3 = (uint8_t) bitextu((unsigned int) *pIn, 2, 6);
+
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 2, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 2, 2);
+    uint8_t inB2 = (uint8_t) bitextu((unsigned int) *pCom, 2, 4);
+    uint8_t inB3 = (uint8_t) bitextu((unsigned int) *pCom, 2, 6);
+
+    inA0 = ((inA0 + inB0) >> 1);
+    inA1 = ((inA1 + inB1) >> 1);
+    inA2 = ((inA2 + inB2) >> 1);
+    inA3 = ((inA3 + inB3) >> 1);
+
+    uint8_t inA = (uint8_t) bitins(inA0, n_mask2, inA1, mask2, off2);
+    inA = bitins(inA, n_mask4, inA2, mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, inA3, mask6, off6);
+
+    pIn++;
+    pCom++;
+    length--;
+  }
+}
+
+/*
+ * XpulpNN
+ */
+
+static void __attribute__((noinline)) xpulp_nn_zero_mem_u8(uint8_t * pBuffer, unsigned int size)
+{
+  int lfover = size &0x3;
+  for (int i=0; i<(size>>2); i++)
+  {
+    *((v4u *)pBuffer) = (v4u){0,0,0,0};
+    MemoryFence();
+    pBuffer+=4;
+  }
+  while(lfover)
+  {
+    *pBuffer++=0;
+    lfover--;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_zero_mem_u4(uint8_t * pBuffer, unsigned int size)
+{
+  int lfover = size &0x7;
+  for (int i=0; i<(size>>3); i++)
+  {
+    *((v4u *)pBuffer) = (v4u){0,0,0,0};
+    MemoryFence();
+    pBuffer+=4;
+  }
+  while(lfover)
+  {
+    *pBuffer++=0;
+    lfover-=2;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_zero_mem_u2(uint8_t * pBuffer, unsigned int size)
+{
+  int lfover = size &0xf;
+  for (int i=0; i<(size>>4); i++)
+  {
+    *((v4u *)pBuffer) = (v4u){0,0,0,0};
+    MemoryFence();
+    pBuffer+=4;
+  }
+  while(lfover)
+  {
+    *pBuffer++=0;
+    lfover-=4;
+  }
+}
+
+
+static void __attribute__((noinline)) xpulp_nn_compare_and_replace_if_larger_u4(uint8_t * base,
+                                                uint8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    *((uint32_t *)pIn) = max8(*((uint32_t *)pIn), *((uint32_t *)pCom));
+
+    pIn+=4;
+    pCom+=4;
+
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 4, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 4, 4);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 4, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 4, 4);
+
+    if(inA0<inB0)
+      inA0=inB0;
+
+    if(inA1<inB1)
+      inA1=inB1;
+
+    *((uint8_t*)pIn) = bitins(inA0, n_mask, inA1, mask, off);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_compare_and_replace_if_larger_i4(
+    int8_t * base, int8_t * target, uint16_t length) {
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~mask;
+  int8_t off = 0x04;
+
+  int8_t *pIn = base;
+  int8_t *pCom = target;
+
+  int cnt = length >> 2;
+
+  while (cnt > 0u) {
+    *((int32_t *)pIn) = maxs8(*((int32_t *)pIn), *((int32_t *)pCom));
+
+    pIn += 4;
+    pCom += 4;
+
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left > 0u) {
+    int8_t inA0 = (int8_t)bitext((int)*pIn, 4, 0);
+    int8_t inA1 = (int8_t)bitext((int)*pIn, 4, 4);
+    int8_t inB0 = (int8_t)bitext((int)*pCom, 4, 0);
+    int8_t inB1 = (int8_t)bitext((int)*pCom, 4, 4);
+
+    if (inA0 < inB0)
+      inA0 = inB0;
+
+    if (inA1 < inB1)
+      inA1 = inB1;
+
+    *((int8_t *)pIn) = bitins(inA0, n_mask, inA1, mask, off);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_avg_and_replace_u4(uint8_t * base,
+                                  uint8_t * target,
+                                  uint16_t length)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+
+  int cnt = length >> 2;
+
+  while (cnt > 0u)
+  {
+    *((uint32_t *)pIn) = avg8(*((uint32_t *)pIn), *((uint32_t *)pCom));
+
+    pIn+=4;
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+
+  while (left>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 4, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 4, 4);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 4, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 4, 4);
+
+    inA0 = ((inA0 + inB0) >> 1);
+    inA1 = ((inA1 + inB1) >> 1);
+
+    *((uint8_t*)pIn) = bitins(inA0, n_mask, inA1, mask, off);
+
+    pIn++;
+    pCom++;
+    length--;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_compare_and_replace_if_larger_u2(uint8_t * base,
+                                                uint8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask2 = 0x0c;
+  int8_t n_mask2 = ~ mask2;
+  int8_t mask4 = 0x30;
+  int8_t n_mask4 = ~ mask4;
+  int8_t mask6 = 0xc0;
+  int8_t n_mask6 = ~ mask6;
+  int8_t off2 = 2;
+  int8_t off4 = 4;
+  int8_t off6 = 6;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+  uint8_t *out;
+
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    *((uint32_t *)pIn) = max16(*((uint32_t *)pIn), *((uint32_t *)pCom));
+
+    pIn+=4;
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 2, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 2, 2);
+    uint8_t inA2 = (uint8_t) bitextu((unsigned int) *pIn, 2, 4);
+    uint8_t inA3 = (uint8_t) bitextu((unsigned int) *pIn, 2, 6);
+    v4u inA4 = pack((uint8_t) inA0, (uint8_t) inA1, (uint8_t) inA2, (uint8_t) inA3);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 2, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 2, 2);
+    uint8_t inB2 = (uint8_t) bitextu((unsigned int) *pCom, 2, 4);
+    uint8_t inB3 = (uint8_t) bitextu((unsigned int) *pCom, 2, 6);
+    v4u inB4 = pack((uint8_t) inB0, (uint8_t) inB1, (uint8_t) inB2, (uint8_t) inB3);
+
+    *((v4u*)out) = max4(inA4, inB4);
+
+    uint8_t inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_compare_and_replace_if_larger_i2(int8_t * base,
+                                                int8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask2 = 0x0c;
+  int8_t n_mask2 = ~ mask2;
+  int8_t mask4 = 0x30;
+  int8_t n_mask4 = ~ mask4;
+  int8_t mask6 = 0xc0;
+  int8_t n_mask6 = ~ mask6;
+  int8_t off2 = 2;
+  int8_t off4 = 4;
+  int8_t off6 = 6;
+
+  int8_t *pIn = base;
+  int8_t *pCom = target;
+  int8_t *out;
+
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    *((int32_t *)pIn) = maxs16(*((int32_t *)pIn), *((int32_t *)pCom));
+
+    pIn+=4;
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    int8_t inA0 = (int8_t) bitext((int) *pIn, 2, 0);
+    int8_t inA1 = (int8_t) bitext((int) *pIn, 2, 2);
+    int8_t inA2 = (int8_t) bitext((int) *pIn, 2, 4);
+    int8_t inA3 = (int8_t) bitext((int) *pIn, 2, 6);
+    v4s inA4 = pack((int8_t) inA0, (int8_t) inA1, (int8_t) inA2, (int8_t) inA3);
+    int8_t inB0 = (int8_t) bitext((int) *pCom, 2, 0);
+    int8_t inB1 = (int8_t) bitext((int) *pCom, 2, 2);
+    int8_t inB2 = (int8_t) bitext((int) *pCom, 2, 4);
+    int8_t inB3 = (int8_t) bitext((int) *pCom, 2, 6);
+    v4s inB4 = pack((int8_t) inB0, (int8_t) inB1, (int8_t) inB2, (int8_t) inB3);
+
+    *((v4s*)out) = maxs4(inA4, inB4);
+
+    int8_t inA = (int8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((int8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_avg_and_replace_u2(uint8_t * base,
+                                  uint8_t * target,
+                                  uint16_t length)
+{
+  int8_t mask2 = 0x0c;
+  int8_t n_mask2 = ~ mask2;
+  int8_t mask4 = 0x30;
+  int8_t n_mask4 = ~ mask4;
+  int8_t mask6 = 0xc0;
+  int8_t n_mask6 = ~ mask6;
+  int8_t off2 = 2;
+  int8_t off4 = 4;
+  int8_t off6 = 6;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+  uint8_t *out;
+
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    *((uint32_t *)pIn) = avg16(*((uint32_t *)pIn), *((uint32_t *)pCom));
+
+    pIn+=4;
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 2, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 2, 2);
+    uint8_t inA2 = (uint8_t) bitextu((unsigned int) *pIn, 2, 4);
+    uint8_t inA3 = (uint8_t) bitextu((unsigned int) *pIn, 2, 6);
+    v4u inA4 = pack((uint8_t) inA0, (uint8_t) inA1, (uint8_t) inA2, (uint8_t) inA3);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 2, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 2, 2);
+    uint8_t inB2 = (uint8_t) bitextu((unsigned int) *pCom, 2, 4);
+    uint8_t inB3 = (uint8_t) bitextu((unsigned int) *pCom, 2, 6);
+    v4u inB4 = pack((uint8_t) inB0, (uint8_t) inB1, (uint8_t) inB2, (uint8_t) inB3);
+
+    *((v4u*)out) = avg4(inA4, inB4);
+
+    uint8_t inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+
+#endif
diff --git a/rt_nn_tests/xpnn_maxpool_2b/test.c b/rt_nn_tests/xpnn_maxpool_2b/test.c
new file mode 100644
index 0000000..d946a63
--- /dev/null
+++ b/rt_nn_tests/xpnn_maxpool_2b/test.c
@@ -0,0 +1,160 @@
+#include <stdio.h>
+#include <stdint.h>
+#include <stdlib.h>
+
+
+#include "pmsis.h"
+
+#include "data_statstest.h"
+//#include "pulp_nn_kernels.h"
+#include "pulp_nn_mix_kernels.h"
+
+
+#define start_cycle_counter()                                   asm volatile("csrw 0xCC0, 0x01;")
+#define stop_cycle_counter()                                    asm volatile("csrw 0xCC0, 0x00;")
+#define read_cycle_counter(x)                                   asm volatile("csrr %0, 0x780;" : "=r" (x))
+#define reset_cycle_counter()                                   asm volatile("csrw 0x780, 0x0;")
+
+uint8_t im2col[IM2COL_DIM] = {0};
+uint8_t outputs[OUTPUT_DIM] = {0};
+
+int32_t outputs_fp[OUTPUT_DIM_FP] = {0};
+
+#ifndef PROFILE
+int num_cycles;
+#else
+int im2col_cycles;
+int hotloop_prep_cycles;
+int hotloop_cycles;
+int threshold_cycles;
+int requant_cycles;
+int hotloop_leftover_cycles;
+int matmul4x2_leftover_cycles;
+#endif
+
+void call_krnl_0();
+void test_0();
+
+int main(int argc, char *argv[])
+{
+#if KRAKEN_PTEST == 1
+  kraken_padframe_aon_pad_gpioa_cfg_rxe_set(24, 0);
+  kraken_padframe_aon_pad_gpioa_cfg_trie_set(24, 0);
+  kraken_padframe_aon_pad_gpioa_cfg_chip2pad_set(24, 0);
+#endif
+  int n_mismatches = 0;
+  int mismatches_tot = 0;
+
+    
+
+  #ifndef PROFILE
+  num_cycles = 0;
+  #endif
+
+  if (get_core_id() == 0) {
+  printf("===> TEST 0: Running xpulp_nn_maxpool_i2...\n");
+  printf("  dims_in     = [32, 32]\n");
+  printf("  dims_kernel = [2, 2]\n");
+  printf("  ch_in/out   = [32, 32]\n");
+  //printf("  padding_y_top    = [%d]\n", padding_y_top);
+  //printf("  padding_y_bottom = [%d]\n", padding_y_bottom);
+  //printf("  padding_x_left   = [%d]\n", padding_x_left);
+  //printf("  padding_x_right  = [%d]\n", padding_x_right);
+  //printf("  stride_x         = [%d]\n", stride_x);
+  //printf("  stride_y         = [%d]\n", stride_y);
+  }
+  test_0();
+  #ifndef PROFILE
+  stop_cycle_counter();
+  read_cycle_counter(num_cycles);
+  if (get_core_id() == 0) {
+  printf("===> TEST 0: Finished running xpulp_nn_maxpool_i2\n");
+  printf("num_cycles = %d\n", num_cycles);
+  printf("MACs = 1048576\n");
+  printf("MACs/cycle = %.4f\n", 1048576/num_cycles);
+  }
+  #endif
+  if (get_core_id() == 0) {
+  printf("Checking for mismatches..\n");
+  n_mismatches = 0;
+
+  for(int i=0; i < 2048; i++) {
+    if (outputs[i] != exp_outp_0[i]){
+        printf("***Mismatch in test 0 at iteration %d: Expected: %x, got: %x\n", i, exp_outp_0[i], outputs[i]);
+        n_mismatches++;
+    }
+  }
+  }
+  mismatches_tot += n_mismatches;
+
+  if (get_core_id() == 0) {
+  printf("Got %d mismatches in %d tests\n", mismatches_tot, 1);
+  }
+  return mismatches_tot;
+}
+
+
+void call_krnl_0(void) {
+  uint8_t * pInp;
+  uint8_t * pIm2ColBuffer;
+  int8_t * pBias = NULL;
+  uint8_t * pOut;
+  int8_t * pWeight;
+  uint32_t * pThr;
+  int32_t * pKappa, pLambda;
+  pInp = inp_l1;
+  
+  pOut = outp_l1;
+#if KRAKEN_PTEST == 1
+  if (pi_core_id() == 0) {
+    kraken_padframe_aon_pad_gpioa_cfg_chip2pad_set(24, 1);
+  } 
+#endif
+  xpulp_nn_maxpool_i2(
+    pInp,
+    pOut,
+    32,
+    32,
+    32,
+    16,
+    16,
+    2,
+    2,
+    0,
+    0,
+    0,
+    0,
+    2,
+    2);
+
+#if KRAKEN_PTEST == 1
+  if (pi_core_id() == 0) {
+    kraken_padframe_aon_pad_gpioa_cfg_chip2pad_set(24, 0);
+  }
+#endif
+}
+
+void test_0(void) {
+  // DMA transfer inputs from L2 to L1
+  if (pi_core_id() == 0) {
+     plp_dma_memcpy(pIn_0, inp_l1, 8192, 1);
+     plp_dma_barrier();
+  }
+  pi_cl_team_barrier(0);
+  call_krnl_0();
+  // get outputs back with DMA
+  if (pi_core_id() == 0) {
+     plp_dma_memcpy(outputs, outp_l1, 2048, 0);
+     plp_dma_barrier();
+  }
+}
+
+
+
+
+
+
+
+
+
+
diff --git a/rt_nn_tests/xpnn_maxpool_2b/xpulp_nn_maxpool_i2.c b/rt_nn_tests/xpnn_maxpool_2b/xpulp_nn_maxpool_i2.c
new file mode 100644
index 0000000..293bb62
--- /dev/null
+++ b/rt_nn_tests/xpnn_maxpool_2b/xpulp_nn_maxpool_i2.c
@@ -0,0 +1,146 @@
+/*
+ * xpulp_nn_maxpool_i2.c
+ * Nazareno Bruschi <nazareno.bruschi@unibo.it>
+ * Angelo Garofalo <angelo.garofalo@unibo.it>
+ *
+ * Copyright (C) 2018-2020 University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include "pmsis.h"
+#include "pulp_nn_utils.h"
+
+
+
+void __attribute__ ((noinline)) xpulp_nn_maxpool_i2(
+                                                  int8_t * pIn,
+                                                  int8_t * pOut,
+  uint16_t  dim_im_in_x,
+  uint16_t  dim_im_in_y,
+  uint16_t  ch_im_in,
+  uint16_t  dim_im_out_x,
+  uint16_t  dim_im_out_y,
+  uint16_t  dim_kernel_x,
+  uint16_t  dim_kernel_y,
+  uint16_t  padding_t,
+  uint16_t  padding_b,
+  uint16_t  padding_l,
+  uint16_t  padding_r,
+  uint16_t  stride_x,
+  uint16_t  stride_y)
+{
+  int core_id = pi_core_id();
+  int n_cores = NUM_CORES;
+  if (dim_im_in_y < NUM_CORES)
+  {
+    n_cores = dim_im_in_y;
+  }
+  int  Log2Core = log2(n_cores);
+  int ch_im_in_r = ch_im_in >> 2;
+
+  int chunck = (dim_im_in_y >> Log2Core) + ((dim_im_in_y & (NUM_CORES-1))!=0);
+
+  int start = min(chunck * core_id, dim_im_in_y);
+  int stop = min(start + chunck, dim_im_in_y);
+  int   i_x, i_y;
+
+  for (i_y = start; i_y < stop; i_y++)
+  {
+    for (i_x = 0; i_x < dim_im_out_x; i_x++)
+    {
+      /* for each output pixel */
+      int8_t     *target = pIn + (i_y * dim_im_in_x + i_x) * ch_im_in_r;
+      uint8_t     *win_start;
+      uint8_t     *win_stop;
+      if (i_x * stride_x - padding_l < 0)
+      {
+        win_start = target;
+      }
+      else
+      {
+        win_start = pIn + (i_y * dim_im_in_x + i_x * stride_x - padding_l) * ch_im_in_r;
+      }
+
+      if (i_x * stride_x - padding_l + dim_kernel_x >= dim_im_in_x)
+      {
+        win_stop = pIn + (i_y * dim_im_in_x + dim_im_in_x) * ch_im_in_r;
+      }
+      else
+      {
+        win_stop = pIn + (i_y * dim_im_in_x + i_x * stride_x - padding_l + dim_kernel_x) * ch_im_in_r;
+      }
+
+      /* first step is to copy over initial data */
+      for (int i = 0; i< ch_im_in_r; i++) target[i] = win_start[i];
+
+      /* start the max operation from the second part */
+      win_start += ch_im_in_r;
+      for (; win_start < win_stop; win_start += ch_im_in_r)
+      {
+        xpulp_nn_compare_and_replace_if_larger_i2(target, win_start, ch_im_in_r);
+      }
+    }
+  }
+
+  pi_cl_team_barrier();
+  if (dim_im_out_y < NUM_CORES)
+  {
+    n_cores = dim_im_out_y;
+  }
+  Log2Core = log2(n_cores);
+  int chunck2 = (dim_im_out_y >> Log2Core) + ((dim_im_out_y & (NUM_CORES-1))!=0);
+  int start2 = chunck2 * core_id;
+  int stop2 = min(start2 + chunck2, dim_im_out_y);
+
+    /* then does the pooling along y axis */
+  for (i_y = start2; i_y < stop2; i_y++)
+  {
+    /* for each output row */
+    int8_t *target = pOut + i_y * dim_im_out_x * ch_im_in_r;
+    int8_t *row_start;
+    int8_t *row_end;
+    /* setting the starting row */
+    if (i_y * stride_y - padding_t < 0)
+    {
+      row_start = pIn;
+    }
+    else
+    {
+      row_start = pIn + (i_y * stride_y - padding_t) * dim_im_in_x * ch_im_in_r;
+    }
+    /* setting the stopping row */
+    if (i_y * stride_y - padding_t + dim_kernel_y >= dim_im_in_y)
+    {
+      row_end = pIn + dim_im_in_y * dim_im_in_x * ch_im_in_r;
+    }
+    else
+    {
+      row_end = pIn + (i_y * stride_y - padding_t + dim_kernel_y) * dim_im_in_x * ch_im_in_r;
+    }
+
+    /* copy over the first row */
+    for (int i = 0; i< dim_im_out_x * ch_im_in_r; i++)
+    {
+      target[i] = (int8_t) row_start[i];
+    }
+    /* move over to next row */
+    row_start += ch_im_in_r * dim_im_in_x;
+
+    for (; row_start < row_end; row_start += dim_im_in_x * ch_im_in_r)
+    {
+      xpulp_nn_compare_and_replace_if_larger_i2(target, row_start, dim_im_out_x * ch_im_in_r);
+    }
+  }
+  pi_cl_team_barrier();
+}
diff --git a/rt_nn_tests/xpnnv2_conv/Makefile b/rt_nn_tests/xpnnv2_conv/Makefile
new file mode 100644
index 0000000..73b2b32
--- /dev/null
+++ b/rt_nn_tests/xpnnv2_conv/Makefile
@@ -0,0 +1,19 @@
+APP = test
+PULP_APP = test
+PULP_APP_SRCS = test.c
+PULP_APP_SRCS += xpulp_nn_mix_conv_u4_u8_i2.c
+PULP_APP_SRCS += xpulp_nn_mix_matmul_u4_u8_i2.c
+
+CORE=8
+
+
+PULP_CFLAGS += -DNUM_CORES=$(CORE) -I. -O3
+PULP_LDFLAGS += -lc -lm -lgcc -Wl,-print-memory-usage 
+
+PULP_APP_CFLAGS += -DNUM_CORES=$(CORE) -I. -O3
+PULP_APP_LDFLAGS += -lc -lm -lgcc -Wl,-print-memory-usage 
+
+PULP_CFLAGS += -DPULP_RUNTIME -DSINGLE_CORE_DMA -DALWAYS_BLOCK_DMA_TRANSFERS
+
+
+include $(PULP_SDK_HOME)/install/rules/pulp.mk
diff --git a/rt_nn_tests/xpnnv2_conv/data_statstest.h b/rt_nn_tests/xpnnv2_conv/data_statstest.h
new file mode 100644
index 0000000..d679d0f
--- /dev/null
+++ b/rt_nn_tests/xpnnv2_conv/data_statstest.h
@@ -0,0 +1,1894 @@
+// this file is generated automatically by the python script `generate_statstest.py`
+
+#ifndef _DATA_H
+#define _DATA_H
+
+
+
+#define IM2COL_DIM (72 * NUM_CORES * 2) // for NUM_CORES cores
+#define OUTPUT_DIM 512
+#define OUTPUT_DIM_FP (1)
+uint8_t pWeight_0 [] = {
+0x29,
+0xc5,
+0xac,
+0x95,
+0x91,
+0x84,
+0x79,
+0x41,
+0xf7,
+0xd6,
+0x8a,
+0x4e,
+0xa8,
+0x85,
+0x7f,
+0x76,
+0x57,
+0x20,
+0x83,
+0xa5,
+0xfa,
+0x8a,
+0xd4,
+0xf2,
+0x58,
+0xa0,
+0x7d,
+0x60,
+0x0a,
+0x58,
+0x02,
+0x04,
+0x86,
+0xfd,
+0xcc,
+0xfb,
+0x8c,
+0xc0,
+0xc0,
+0x88,
+0xde,
+0x5d,
+0x3b,
+0x3c,
+0x69,
+0x80,
+0xd0,
+0x38,
+0x83,
+0xe5,
+0x35,
+0x27,
+0x14,
+0x5d,
+0x2a,
+0xe8,
+0xd0,
+0xf9,
+0xcc,
+0x81,
+0x16,
+0x31,
+0xd3,
+0xc6,
+0xf1,
+0xfa,
+0xe2,
+0x5b,
+0x08,
+0x6d,
+0xa5,
+0x01,
+0xb2,
+0x28,
+0x54,
+0x06,
+0x78,
+0x97,
+0x40,
+0x73,
+0xd2,
+0x27,
+0x96,
+0xbd,
+0x1f,
+0xc0,
+0xfa,
+0xb8,
+0x49,
+0x68,
+0x75,
+0x2f,
+0xe5,
+0xa4,
+0xcb,
+0xcc,
+0x09,
+0x39,
+0x85,
+0x71,
+0xc3,
+0xfb,
+0x6f,
+0x9b,
+0x42,
+0x0f,
+0x91,
+0x1b,
+0x66,
+0xaf,
+0xfb,
+0xc8,
+0xf8,
+0xba,
+0x9d,
+0x0a,
+0xd2,
+0xa8,
+0x39,
+0x4a,
+0x2d,
+0xcd,
+0xf8,
+0xc4,
+0x8c,
+0x02,
+0xe2,
+0x4f,
+0x68,
+0xff,
+0x96,
+0x7f,
+0x96,
+0xad,
+0xa5,
+0xf2,
+0xb8,
+0xb8,
+0x5d,
+0x02,
+0x4c,
+0x19,
+0xb0,
+0x11,
+0x51,
+0x8d,
+0xb3,
+0xb1,
+0x3f,
+0xcd,
+0xaf,
+0x85,
+0x94,
+0x65,
+0x24,
+0xa1,
+0xbb,
+0x3e,
+0xa1,
+0xe2,
+0xab,
+0xb4,
+0x62,
+0x6c,
+0xf5,
+0x2f,
+0x5a,
+0xbe,
+0x68,
+0xba,
+0x8a,
+0xcf,
+0xa1,
+0x3f,
+0xfa,
+0xb8,
+0xe9,
+0xe8,
+0xfa,
+0x4e,
+0xdd,
+0x57,
+0x45,
+0x1a,
+0x5c,
+0xfd,
+0x58,
+0x77,
+0x52,
+0xd6,
+0x15,
+0x07,
+0x39,
+0xf2,
+0x34,
+0xf2,
+0x66,
+0x90,
+0xea,
+0x42,
+0x42,
+0x43,
+0x36,
+0xd7,
+0xc3,
+0xb8,
+0x2c,
+0x10,
+0xd0,
+0x90,
+0x2d,
+0xf9,
+0x6a,
+0xbc,
+0x57,
+0x19,
+0x69,
+0x60,
+0x65,
+0x7f,
+0xee,
+0x10,
+0x06,
+0x71,
+0xe4,
+0x49,
+0x5b,
+0x5a,
+0x5a,
+0x9c,
+0x6f,
+0x87,
+0x68,
+0x44,
+0xb4,
+0xf8,
+0xe8,
+0xae,
+0x75,
+0xe0,
+0x0f,
+0x43,
+0x51,
+0x71,
+0xde,
+0xce,
+0x56,
+0x13,
+0x3d,
+0xf9,
+0x80,
+0x1f,
+0x0a,
+0x9e,
+0x2b,
+0xa7,
+0x73,
+0x76,
+0xe2,
+0x36,
+0x16,
+0x42,
+0x0c,
+0x73,
+0x59,
+0xf6,
+0x53,
+0xaf,
+0x77,
+0x12,
+0x67,
+0x33,
+0xb7,
+0x90,
+0x41,
+0x55,
+0x90,
+0xbf,
+0x8c,
+0x31,
+0xfc,
+0xdd,
+0x9c,
+0x42,
+0x9a,
+0xfd,
+0x30,
+0xe5,
+0x12,
+0x12,
+0x6c,
+0xc3,
+0xa6,
+0xd8,
+0xb0,
+0x41,
+0x4c,
+0xd0,
+0x27,
+0x97,
+0x7f,
+0x8b,
+0x4c,
+0x44,
+0xb0,
+0x7f,
+0xba,
+0x18,
+0x37,
+0x46,
+0x9e,
+0xdc,
+0xbf,
+0xd8,
+0x97,
+0xce,
+0x3b,
+0x61,
+0x7e,
+0x9b,
+0x36,
+0xe7,
+0x0c,
+0x4e,
+0x57,
+0x21,
+0x36,
+0x18,
+0x4d,
+0x17,
+0x41,
+0x83,
+0xd4,
+0x8b,
+0xf7,
+0x53,
+0x4d,
+0xff,
+0x98,
+0xbb,
+0x79,
+0x38,
+0x2e,
+0x22,
+0x8f,
+0xe5,
+0x5e,
+0x81,
+0xdb,
+0xf0,
+0x2b,
+0x16,
+0xf6,
+0x69,
+0xb1,
+0xe5,
+0xd6,
+0xab,
+0xc6,
+0xb1,
+0xd0,
+0x62,
+0x57,
+0xa4,
+0x6c,
+0xe8,
+0x0e,
+0x65,
+0xb6,
+0x18,
+0x7a,
+0xe7,
+0x99,
+0x92,
+0x68,
+0x83,
+0x7e,
+0xe3,
+0xca,
+0x61,
+0xb0,
+0x5e,
+0x4d,
+0x25,
+0x92,
+0xe5,
+0x02,
+0xbb,
+0x81,
+0xfa,
+0x81,
+0x92,
+0x9d,
+0x80,
+0xd7,
+0xf4,
+0x73,
+0xe3,
+0x63,
+0x64,
+0x5a,
+0x26,
+0x98,
+0x55,
+0x32,
+0x65,
+0xf9,
+0x5c,
+0x5d,
+0xdc,
+0x2e,
+0x59,
+0xf1,
+0x90,
+0xf5,
+0x78,
+0x05,
+0xb1,
+0x5c,
+0xde,
+0xdd,
+0x93,
+0x3f,
+0xd8,
+0xc2,
+0x2a,
+0x94,
+0xfd,
+0x51,
+0x72,
+0x9c,
+0xda,
+0x58,
+0xc6,
+0x94,
+0xe5,
+0xed,
+0x3d,
+0xf3,
+0x7a,
+0xd8,
+0x91,
+0xbf,
+0xeb,
+0x31,
+0xda,
+0xe4,
+0x89,
+0x74,
+0x31,
+0x86,
+0x42,
+0xe2,
+0x1b,
+0x7e,
+0x28,
+0xaf,
+0xf8,
+0xfd,
+0x0b,
+0x9a,
+0xa7,
+0x39,
+0xd8,
+0x9d,
+0x25,
+0x7f,
+0xe5,
+0xc4,
+0x7f,
+0x98,
+0x9f,
+0x79,
+0x5f,
+0x4b,
+0x7a,
+0x05,
+0x07,
+0x6c,
+0xe9,
+0x1f,
+0xd0,
+0x60,
+0xfb,
+0x7f,
+0xf6,
+0x5c,
+0xa2,
+0x3d,
+0x2c,
+0x22,
+0x7b,
+0xbe,
+0x39,
+0x0a,
+0xfc,
+0x8f,
+0x8b,
+0x80,
+0x6f,
+0x6f,
+0x1b,
+0x28,
+0x9b,
+0x79,
+0x0e,
+0xb3,
+0x57,
+0x07,
+0x45,
+0x61,
+0x83,
+0x12,
+0x7a,
+0x53,
+0x3d,
+0x80,
+0xf1,
+0xe1,
+0xd1,
+0x40,
+0x26,
+0xfd,
+0xd8,
+0x81,
+0xab,
+0xbe,
+0x16,
+0xfa,
+0xf2,
+0x25,
+0x15,
+0x8f,
+0xcd,
+0xc7,
+0xbe,
+0xe1,
+0xd9,
+0x95,
+0x3b,
+0xbd,
+0x6b,
+0x81,
+0x59,
+0x26,
+0x27,
+0xe8,
+0x1b,
+0x74,
+0x7b,
+0x95,
+0xb4,
+0xe0,
+0x60,
+0x0b,
+0x44,
+0x67,
+0x7e,
+0xb9,
+0xfb,
+0xc2,
+0xc3,
+0xb2,
+0xd5,
+0x22,
+0x39,
+0xae,
+0xb2,
+0xd9,
+0x2e,
+0x63,
+0x19,
+0x04,
+0xe5,
+0x69,
+0xf5,
+0x5f,
+0x2a,
+0x9f,
+0x8f,
+0xe2,
+0xab,
+0x89,
+0x8d,
+0xd9,
+0x23,
+0x45,
+0x75,
+0xa2,
+0x2e,
+0x1d,
+0xc2,
+0xe4,
+0xf0,
+0xe0,
+0xbb,
+0x26,
+0x8b,
+0x7c,
+0xd4,
+0xaa,
+0x6d,
+0x74,
+0x0e,
+0xa7,
+0x73,
+0x44,
+0xc2,
+0xee,
+0x74,
+0xd4,
+0x07,
+0xf9,
+0xdd,
+0xd6,
+0x57,
+0x84,
+0xdd,
+0xee,
+0x67,
+0x82,
+0xd1,
+0x52,
+0xb3,
+0x35,
+0xb3,
+0xb4,
+0x3e,
+0x19,
+0xd7,
+0x55,
+0x22,
+0xf0,
+0xdd,
+0xe1,
+0x80,
+0x15,
+0x71,
+0xff,
+0x58,
+0x87,
+0xe2,
+0xf2,
+0xd4,
+0x05,
+0x68,
+0x46,
+0xf2,
+0x22,
+0xd0,
+0x46,
+0xa8,
+0xa4,
+0x16,
+0x81,
+0xc5,
+0x7d,
+0x1b,
+0x50,
+0x25,
+0x91,
+0xcd,
+0x29,
+0xa6,
+0x35,
+0x32,
+0xf5,
+0xff,
+0x53,
+0xd7,
+0x00,
+0xf6,
+0xab,
+0x39,
+0xac,
+0x85,
+0xe9,
+0x1a,
+0x80,
+0x1f,
+0x89,
+0x30,
+0x77,
+0x9a,
+0x5e,
+0xb2,
+0xdb,
+0x74,
+0x53,
+0x48,
+0x6a,
+0x91,
+0x11,
+0xf0,
+0xd8,
+0x94,
+0x63,
+0xbc,
+0x32,
+0x84,
+0x55,
+0x04,
+0x84,
+0x8a,
+0x29,
+0x38,
+0x9c,
+0x2c,
+0x14,
+0x23,
+0xfe,
+0xb9,
+0x88,
+0x6e,
+0xac,
+0xf4,
+0xcf,
+0xd6,
+0x4a,
+0xe5,
+0x19,
+0xad,
+0x80,
+0x4b,
+0x34,
+0x0f,
+0xbc,
+0x93,
+0x86,
+0x1c,
+0x77,
+0x1c,
+0x7e,
+0xda,
+0x39,
+0xf3,
+0xb0,
+0x7b,
+0xae,
+0x0a,
+0x4a,
+0x49,
+0xba,
+0x69,
+0xf5,
+0xa5,
+0x05,
+0x3b,
+0xab,
+0xfd,
+0xa1,
+0xc6,
+0xc7,
+0x8d,
+0x7e,
+0x1a,
+0x1d,
+0xa4,
+0x6d,
+0x54,
+0xe2,
+0x1b,
+0xb9,
+0x46,
+0x42,
+0xbe,
+0xc9,
+0x6a,
+0xc7,
+0x14,
+0xa1,
+0xd4,
+0x27,
+0x39,
+0x65,
+0xf8,
+0xb3,
+0xf4,
+0x1a,
+0x31,
+0x9a,
+0xc5,
+0x74,
+0x63,
+0xb3,
+0xa9,
+0x10,
+0xe2,
+0x22,
+0x1d,
+0xc6,
+0x20,
+0x00,
+0x58,
+0xa7,
+0xaa,
+0x82,
+0x25,
+0x71,
+0xf6,
+0xb0,
+0x5d,
+0xd8,
+0xeb,
+0xf5,
+0x17,
+0xd0,
+0x04,
+0x07,
+0xac,
+0xf5,
+0x63,
+0x7a,
+0x26,
+0x0a,
+0x68,
+0xe3,
+0x74,
+0xc5,
+0x44,
+0x44,
+0xd8,
+0xf3,
+0xa5,
+0x99,
+0x77,
+0xc4,
+0x48,
+0x47,
+0xc0,
+0x94,
+0xd6,
+0x27,
+0x79,
+0x3d,
+0x0a,
+0x83,
+0x46,
+0x0b,
+0x8f,
+0xd7,
+0x03,
+0x7c,
+0xa7,
+0x35,
+0xd5,
+0x70,
+0xb2,
+0x19,
+0x29,
+0x14,
+0xb6,
+0xbe,
+0x61,
+0xfb,
+0x5c,
+0x04,
+0xf4,
+0x7f,
+0xb4,
+0xbd,
+0xcb,
+0x4c,
+0xae,
+0xd0,
+0x20,
+0x3e,
+0xd9,
+0x25,
+0xcc,
+0xa9,
+0xd2,
+0x13,
+0xf7,
+0x66,
+0x4e,
+0xbe,
+0xeb,
+0xed,
+0x6b,
+0xb0,
+0x9c,
+0xd3,
+0x94,
+0x21,
+0x33,
+0x70,
+0x16,
+0x53,
+0xfe,
+0x21,
+0x08,
+0xf5,
+0xbc,
+0x1b,
+0xb1,
+0x92,
+0x3a,
+0xf4,
+0xd2,
+0xc7,
+0x40,
+0x96,
+0x01,
+0xbc,
+0xd5,
+0x96,
+0x35,
+0x17,
+0xa0,
+0x0c,
+0x1b,
+0x85,
+0x2e,
+0xb3,
+0xab,
+0x55,
+0xcb,
+0xc0,
+0xe0,
+0x75,
+0x61,
+0xf8,
+0xa6,
+0x96,
+0xa2,
+0x3f,
+0x57,
+0xd6,
+0xb7,
+0xfc,
+0x8a,
+0x6a,
+0x6c,
+0xcc,
+0x38,
+0x7f,
+0x8c,
+0x99,
+0x6e,
+0x5e,
+0x19,
+0x23,
+0x21,
+0xb0,
+0xaa,
+0x77,
+0x43,
+0x05,
+0x9b,
+0x70,
+0xd4,
+0xde,
+0xf8,
+0x7d,
+0xbb,
+0xad,
+0x28,
+0x4e,
+0xc3,
+0x5f,
+0x97,
+0x1d,
+0x79,
+0xd9,
+0x55,
+0xce,
+0x28,
+0x13,
+0x91,
+0xc0,
+0x94,
+0x5b,
+0x04,
+0x70,
+0xcc,
+0xd8,
+0x38,
+0xa9,
+0x72,
+0x0e,
+0xd1,
+0x93,
+0xe4,
+0x81,
+0x47,
+0x29,
+0xec,
+0x0f,
+0x4c,
+0x5d,
+0x50,
+0x6e,
+0xd7,
+0xa2,
+0xe3,
+0x4e,
+0x58,
+0x29,
+0x05,
+0xa2,
+0x37,
+0x27,
+0xfe,
+0x84,
+0x51,
+0xac,
+0xc2,
+0xc8,
+0x03,
+0xe6,
+0x5f,
+0xea,
+0x15,
+0x25,
+0x7a,
+0x39,
+0x60,
+0xd1,
+0xc8,
+0x94,
+0xb3,
+0xf9,
+0x8b,
+0x20,
+0x45,
+0xec,
+0x75,
+0x54,
+0xd1,
+0xdd,
+0x6a,
+0x39,
+0x31,
+0xc9,
+0x62,
+0x67,
+0x84,
+0xb3,
+0x47,
+0xbb,
+0x10,
+0xb1,
+0x0f,
+0x5e,
+0x11,
+0x00,
+0x1e,
+0xc5,
+0x79,
+0xf9,
+0xd9,
+0xb7,
+0x18,
+0x0a,
+0x55,
+0x1d,
+0x99,
+0xd5,
+0x46,
+0x2a,
+0x22,
+0xd9,
+0x2e,
+0xc3,
+0x2d,
+0xd1,
+0x9e,
+0xb9,
+0x3a,
+0x32,
+0x1a,
+0x13,
+0xdb,
+0xe5,
+0x45,
+0x1a,
+0x83,
+0x38,
+0xec,
+0x45,
+0x7b,
+0x87,
+0x01,
+0x3c,
+0x32,
+0xd3,
+0x60,
+0x09,
+0xdd,
+0xce,
+0x87,
+0xa3,
+0x03,
+0x02,
+0x91,
+0xfa,
+0xc5,
+0x4d,
+0x4d,
+0x08,
+0x6d,
+0xca,
+0x9e,
+0xfa,
+0x7e,
+0x8c,
+0xdf,
+0x3e,
+0xb3,
+0x10,
+0xfb,
+0xa3,
+0x2f,
+0xef,
+0x7e,
+0x1e,
+0x9f,
+0xd3,
+0x78,
+0xfb,
+0x5f,
+0x50,
+0x4c,
+0xaa,
+0x16,
+0x0c,
+0x35,
+0x46,
+0x10,
+0x3c,
+0x67,
+0xaa,
+0x95,
+0x8a,
+0x38,
+0xf3,
+0xb5,
+0x07,
+0xdb,
+0x6c,
+0x52,
+0xde,
+0xc6,
+0x77,
+0x0e,
+0xd0,
+0xf7,
+0xea,
+0x1f,
+0x96,
+0xae
+};
+
+uint8_t pIn_0 [] = {
+0x09,
+0xe8,
+0x1c,
+0x6f,
+0x55,
+0xe1,
+0xa3,
+0x4d,
+0xad,
+0x94,
+0x4a,
+0x9d,
+0x90,
+0xfc,
+0x76,
+0xc9,
+0x84,
+0xce,
+0xa8,
+0x18,
+0xe2,
+0x30,
+0x85,
+0x41,
+0x6f,
+0x11,
+0x09,
+0xff,
+0x07,
+0xb0,
+0x5b,
+0xe4,
+0x40,
+0xf3,
+0x8f,
+0xcd,
+0xa3,
+0x2f,
+0x85,
+0x67,
+0x31,
+0x96,
+0xc2,
+0x86,
+0x1d,
+0xe5,
+0xc1,
+0xee,
+0x0b,
+0x4d,
+0x50,
+0xa6,
+0x00,
+0x69,
+0xff,
+0x8c,
+0x4c,
+0xb5,
+0x46,
+0xd5,
+0x14,
+0xd0,
+0xdb,
+0x47,
+0xf9,
+0x2d,
+0x77,
+0x9a,
+0x92,
+0xa4,
+0xe8,
+0xab,
+0xbd,
+0xd3,
+0x88,
+0xb6,
+0x00,
+0xbc,
+0xb0,
+0xe7,
+0x60,
+0xd9,
+0x8b,
+0x1b,
+0x9e,
+0x06,
+0x95,
+0x99,
+0x8d,
+0x6e,
+0x21,
+0x20,
+0xa3,
+0x5e,
+0x99,
+0x27,
+0xb1,
+0x7d,
+0xce,
+0xdc,
+0xcf,
+0xd6,
+0xe8,
+0x16,
+0xaa,
+0xec,
+0x67,
+0x8c,
+0x23,
+0xd0,
+0x15,
+0xa0,
+0xe8,
+0x86,
+0x2b,
+0x90,
+0xe1,
+0xb2,
+0x5b,
+0x13,
+0x19,
+0xbe,
+0x8c,
+0x9e,
+0x12,
+0x13,
+0xeb,
+0x69
+};
+
+int32_t pLambda_0 [] = {
+0x00247d20,
+0x001b6100,
+0x00124840,
+0x00247d20,
+0x00247d20,
+0x00124840,
+0x001b6100,
+0x00247d20,
+0x00092c20,
+0x00247d20,
+0x001b6100,
+0x00124840,
+0x00124840,
+0x00124840,
+0x00124840,
+0x001b6100,
+0x00124840,
+0x00124840,
+0x00092c20,
+0x00124840,
+0x001b6100,
+0x00092c20,
+0x00124840,
+0x001b6100,
+0x001b6100,
+0x00247d20,
+0x00247d20,
+0x00092c20,
+0x001b6100,
+0x00092c20,
+0x00247d20,
+0x001b6100
+};
+
+int32_t pKappa_0 [] = {
+0x00000acb,
+0x00000818,
+0x00000566,
+0x00000acb,
+0x00000acb,
+0x00000566,
+0x00000818,
+0x00000acb,
+0x000002b3,
+0x00000acb,
+0x00000818,
+0x00000566,
+0x00000566,
+0x00000566,
+0x00000566,
+0x00000818,
+0x00000566,
+0x00000566,
+0x000002b3,
+0x00000566,
+0x00000818,
+0x000002b3,
+0x00000566,
+0x00000818,
+0x00000818,
+0x00000acb,
+0x00000acb,
+0x000002b3,
+0x00000818,
+0x000002b3,
+0x00000acb,
+0x00000818
+};
+
+
+/* -----------EXPECTED OUTPUTS----------- */
+uint8_t exp_outp_0 [] = {
+0xbb,
+0x94,
+0x64,
+0xc4,
+0x8c,
+0x6e,
+0x90,
+0xff,
+0x34,
+0xbe,
+0x8a,
+0x76,
+0x5f,
+0x58,
+0x58,
+0x88,
+0x6a,
+0x6b,
+0x34,
+0x5f,
+0x77,
+0x30,
+0x66,
+0xa7,
+0xa9,
+0xbe,
+0xe5,
+0x33,
+0xc3,
+0x2d,
+0xbc,
+0x80,
+0x91,
+0x97,
+0x53,
+0xb6,
+0x25,
+0x55,
+0xa3,
+0xc9,
+0x29,
+0x7f,
+0x79,
+0x66,
+0x58,
+0x3e,
+0x3c,
+0x79,
+0x46,
+0x68,
+0x2d,
+0x51,
+0x5d,
+0x2a,
+0x65,
+0x92,
+0x5b,
+0x6c,
+0xbd,
+0x33,
+0x97,
+0x26,
+0x79,
+0x7e,
+0xb6,
+0x98,
+0x45,
+0xa0,
+0x37,
+0x54,
+0x9a,
+0xd5,
+0x20,
+0x87,
+0x7d,
+0x69,
+0x5f,
+0x43,
+0x4b,
+0x7e,
+0x49,
+0x77,
+0x26,
+0x5f,
+0x54,
+0x2b,
+0x77,
+0x91,
+0x71,
+0x94,
+0xec,
+0x27,
+0x85,
+0x27,
+0x78,
+0x6a,
+0xe0,
+0xb9,
+0x76,
+0xb4,
+0x9e,
+0x6b,
+0xc0,
+0xf0,
+0x34,
+0xd6,
+0x7d,
+0x7a,
+0x5a,
+0x55,
+0x69,
+0x86,
+0x52,
+0x7e,
+0x34,
+0x6b,
+0x80,
+0x39,
+0x80,
+0xc3,
+0x90,
+0xc0,
+0xf0,
+0x3d,
+0xb4,
+0x38,
+0xc6,
+0xb3,
+0xa9,
+0x7c,
+0x4e,
+0x5d,
+0x68,
+0x74,
+0xa2,
+0xc5,
+0x26,
+0xac,
+0x7d,
+0x5e,
+0x44,
+0x48,
+0x55,
+0x77,
+0x52,
+0x6e,
+0x30,
+0x60,
+0x5d,
+0x1f,
+0x57,
+0x84,
+0x8f,
+0x99,
+0xa4,
+0x29,
+0x97,
+0x2e,
+0x80,
+0x6f,
+0x69,
+0x63,
+0x42,
+0x35,
+0x00,
+0x61,
+0x68,
+0x9c,
+0x21,
+0x82,
+0x36,
+0x38,
+0x10,
+0x24,
+0x48,
+0x39,
+0x16,
+0x55,
+0x14,
+0x2c,
+0x28,
+0x22,
+0x43,
+0x58,
+0x41,
+0x4b,
+0x8c,
+0x21,
+0x53,
+0x22,
+0x58,
+0x49,
+0x72,
+0x62,
+0x27,
+0x13,
+0x0b,
+0x52,
+0x89,
+0x7b,
+0x19,
+0x93,
+0x6b,
+0x4c,
+0x1b,
+0x22,
+0x41,
+0x36,
+0x1a,
+0x61,
+0x1e,
+0x29,
+0x4b,
+0x12,
+0x54,
+0x5d,
+0x49,
+0x68,
+0x93,
+0x1d,
+0x72,
+0x26,
+0x5a,
+0x5b,
+0xe8,
+0x9d,
+0x64,
+0x8b,
+0x71,
+0x65,
+0xb3,
+0xb4,
+0x2e,
+0xd2,
+0x8e,
+0x69,
+0x5b,
+0x48,
+0x6c,
+0x6e,
+0x33,
+0x70,
+0x2c,
+0x57,
+0x67,
+0x39,
+0x75,
+0xb2,
+0x80,
+0xba,
+0xdf,
+0x38,
+0x9c,
+0x33,
+0xc1,
+0x94,
+0x92,
+0x67,
+0x49,
+0x90,
+0x5e,
+0x75,
+0x78,
+0xd7,
+0x2b,
+0xaa,
+0x65,
+0x72,
+0x43,
+0x41,
+0x56,
+0x5e,
+0x4c,
+0x55,
+0x2d,
+0x5a,
+0x6a,
+0x25,
+0x54,
+0x97,
+0x89,
+0xcb,
+0xab,
+0x24,
+0x9f,
+0x27,
+0x84,
+0x6b,
+0x87,
+0x73,
+0x3b,
+0x25,
+0x00,
+0x62,
+0x63,
+0x82,
+0x16,
+0x7b,
+0x38,
+0x47,
+0x22,
+0x1c,
+0x49,
+0x57,
+0x24,
+0x58,
+0x26,
+0x33,
+0x25,
+0x1c,
+0x54,
+0x81,
+0x31,
+0x5f,
+0x7f,
+0x28,
+0x6f,
+0x23,
+0x52,
+0x5b,
+0x8c,
+0x65,
+0x39,
+0x34,
+0x05,
+0x5d,
+0x7f,
+0x7f,
+0x1c,
+0x7c,
+0x4d,
+0x73,
+0x11,
+0x21,
+0x3e,
+0x5d,
+0x26,
+0x55,
+0x1a,
+0x58,
+0x1d,
+0x22,
+0x43,
+0x65,
+0x42,
+0x56,
+0xac,
+0x26,
+0x5c,
+0x2c,
+0x4c,
+0x5d,
+0xf7,
+0x87,
+0x58,
+0x65,
+0x59,
+0x62,
+0x95,
+0xa8,
+0x27,
+0xad,
+0x80,
+0x5b,
+0x51,
+0x54,
+0x70,
+0x6a,
+0x34,
+0x77,
+0x27,
+0x75,
+0x75,
+0x30,
+0x6c,
+0x95,
+0x88,
+0xa9,
+0xca,
+0x2c,
+0x7c,
+0x37,
+0x8f,
+0x84,
+0xe0,
+0xa4,
+0x62,
+0x7d,
+0xaa,
+0x77,
+0xa7,
+0xfd,
+0x2f,
+0xc5,
+0x8a,
+0x79,
+0x58,
+0x68,
+0x6b,
+0x80,
+0x55,
+0x7c,
+0x43,
+0x7b,
+0x84,
+0x2f,
+0x69,
+0xa7,
+0xa8,
+0xf3,
+0xbb,
+0x35,
+0xbd,
+0x30,
+0xd8,
+0x82,
+0x9a,
+0xa9,
+0x5b,
+0x60,
+0x7f,
+0x87,
+0x82,
+0xc9,
+0x2c,
+0xcf,
+0x7f,
+0x71,
+0x51,
+0x63,
+0x61,
+0x7d,
+0x35,
+0x6a,
+0x31,
+0x5f,
+0x63,
+0x38,
+0x50,
+0xaf,
+0x6d,
+0xd0,
+0x8d,
+0x31,
+0x88,
+0x31,
+0xc9,
+0x8b,
+0xaa,
+0x8e,
+0x6a,
+0x6c,
+0x82,
+0x78,
+0x83,
+0xc4,
+0x31,
+0xdc,
+0x77,
+0x77,
+0x4c,
+0x57,
+0x6c,
+0x8b,
+0x47,
+0x6b,
+0x3b,
+0x69,
+0x7d,
+0x36,
+0x52,
+0xab,
+0x7b,
+0xda,
+0xbc,
+0x33,
+0xa2,
+0x2e,
+0xbe,
+0x75,
+0xf6,
+0xab,
+0x6a,
+0xb6,
+0xad,
+0x7b,
+0xa4,
+0xc6,
+0x33,
+0xe6,
+0x8e,
+0x7b,
+0x77,
+0x62,
+0x79,
+0x84,
+0x4a,
+0x79,
+0x34,
+0x7d,
+0x88,
+0x36,
+0x6a,
+0xbf,
+0x88,
+0xf9,
+0xdf,
+0x3d,
+0x98,
+0x3a,
+0xfd,
+0xaa
+};
+
+
+
+L1_DATA uint32_t threshs_l1 [32] = {0};
+L1_DATA int32_t kappa_l1 [32] = {0};
+L1_DATA int32_t lambda_l1 [32] = {0};
+L1_DATA uint8_t inp_l1 [128] = {0};
+L1_DATA uint8_t outp_l1 [512] = {0};
+L1_DATA int8_t wt_l1 [1152] = {0};
+L1_DATA uint8_t im2col_l1 [IM2COL_DIM] = {0};
+
+#endif
diff --git a/rt_nn_tests/xpnnv2_conv/pulp_nn_kernels.h b/rt_nn_tests/xpnnv2_conv/pulp_nn_kernels.h
new file mode 100644
index 0000000..f4ec977
--- /dev/null
+++ b/rt_nn_tests/xpnnv2_conv/pulp_nn_kernels.h
@@ -0,0 +1,26528 @@
+/*
+ * pulp_nn_kernels.h
+ * Nazareno Bruschi <nazareno.bruschi@unibo.it>
+ *
+ * Copyright (C) 2019-2020 University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef __XPULPNN_KERNELS__
+#define __XPULPNN_KERNELS__
+
+void xpulp_nn_conv_u8_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+#ifndef PROFILE
+                        uint8_t flag_batch_norm);
+#else
+  uint8_t flag_batch_norm,
+    uint32_t *im2col_cycles,
+    uint32_t *hotloop_prep_cycles,
+    uint32_t *hotloop_cycles,
+    uint32_t *requant_cycles,
+    uint32_t *hotloop_leftover_cycles,
+    uint32_t *matmul4x2_leftover_cycles);
+#endif
+                        
+
+void xpulp_nn_conv_i2_i2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_i8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_i8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_i8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_i4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_i4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_i4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_i2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_i2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_i2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_i8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_i8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_i8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_i4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_i4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_i4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_i2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_i2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_i2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_i8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_i8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_i8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_i4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_i4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_i4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_i2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_i2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_i2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u8_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i8_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u8_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i8_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u8_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i8_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u4_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i4_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u4_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i4_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u4_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i4_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u2_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i2_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u2_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i2_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u2_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i2_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u8_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i8_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u8_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i8_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u8_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i8_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u4_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i4_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u4_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i4_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u4_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i4_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u2_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i2_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u2_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i2_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u2_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i2_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u8_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i8_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u8_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i8_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u8_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i8_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u4_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i4_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u4_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i4_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u4_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i4_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u2_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i2_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u2_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i2_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+#ifndef PROFILE
+                        uint8_t flag_batch_norm);
+#else
+  uint8_t flag_batch_norm,
+    uint32_t *hotloop_prep_cycles,
+    uint32_t *hotloop_cycles,
+    uint32_t *requant_cycles,
+    uint32_t *hotloop_leftover_cycles);
+#endif
+
+uint8_t *xpulp_nn_matmul_u2_u2_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i2_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_i8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_i8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_i8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_i4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_i4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_i4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_i2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_i2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_i2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_i8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_i8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_i8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_i4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_i4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_i4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_i2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_i2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_i2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_i8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_i8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_i8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_i4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_i4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_i4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_i2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_i2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_i2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_i32_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_i8_i32_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_u8_i32_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_i8_i32_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_u8_i32_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_i8_i32_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_u4_i32_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_i4_i32_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_u4_i32_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_i4_i32_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_u4_i32_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_i4_i32_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_u2_i32_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_i2_i32_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_u2_i32_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_i2_i32_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_u2_i32_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_i2_i32_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_u8_u8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_i8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_u8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_i8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_u8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_i8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_u4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_i4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_u4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_i4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_u4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_i4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_u2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_i2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_u2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_i2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_u2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_i2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_u8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_i8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_u8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_i8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_u8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_i8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_u4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_i4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_u4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_i4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_u4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_i4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_u2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_i2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_u2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_i2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_u2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_i2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_u8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_i8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_u8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_i8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_u8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_i8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_u4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_i4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_u4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_i4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_u4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_i4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_u2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_i2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_u2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_i2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_u2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_i2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_maxpool_u8(
+                        uint8_t * pIn,
+                        uint8_t * pOut,
+                        uint16_t  dim_im_in_x,
+                        uint16_t  dim_im_in_y,
+                        uint16_t  ch_im_in,
+                        uint16_t  dim_im_out_x,
+                        uint16_t  dim_im_out_y,
+                        uint16_t  dim_kernel_x,
+                        uint16_t  dim_kernel_y,
+                        uint16_t  padding_t,
+                        uint16_t  padding_b,
+                        uint16_t  padding_l,
+                        uint16_t  padding_r,
+                        uint16_t  stride_x,
+                        uint16_t  stride_y);
+
+void xpulp_nn_maxpool_i8(
+                        int8_t * pIn,
+                        int8_t * pOut,
+                        uint16_t  dim_im_in_x,
+                        uint16_t  dim_im_in_y,
+                        uint16_t  ch_im_in,
+                        uint16_t  dim_im_out_x,
+                        uint16_t  dim_im_out_y,
+                        uint16_t  dim_kernel_x,
+                        uint16_t  dim_kernel_y,
+                        uint16_t  padding_t,
+                        uint16_t  padding_b,
+                        uint16_t  padding_l,
+                        uint16_t  padding_r,
+                        uint16_t  stride_x,
+                        uint16_t  stride_y);
+
+void xpulp_nn_maxpool_u4(
+                        uint8_t * pIn,
+                        uint8_t * pOut,
+                        uint16_t  dim_im_in_x,
+                        uint16_t  dim_im_in_y,
+                        uint16_t  ch_im_in,
+                        uint16_t  dim_im_out_x,
+                        uint16_t  dim_im_out_y,
+                        uint16_t  dim_kernel_x,
+                        uint16_t  dim_kernel_y,
+                        uint16_t  padding_t,
+                        uint16_t  padding_b,
+                        uint16_t  padding_l,
+                        uint16_t  padding_r,
+                        uint16_t  stride_x,
+                        uint16_t  stride_y);
+
+void xpulp_nn_maxpool_i4(
+                        int8_t * pIn,
+                        int8_t * pOut,
+                        uint16_t  dim_im_in_x,
+                        uint16_t  dim_im_in_y,
+                        uint16_t  ch_im_in,
+                        uint16_t  dim_im_out_x,
+                        uint16_t  dim_im_out_y,
+                        uint16_t  dim_kernel_x,
+                        uint16_t  dim_kernel_y,
+                        uint16_t  padding_t,
+                        uint16_t  padding_b,
+                        uint16_t  padding_l,
+                        uint16_t  padding_r,
+                        uint16_t  stride_x,
+                        uint16_t  stride_y);
+
+void xpulp_nn_maxpool_u2(
+                        uint8_t * pIn,
+                        uint8_t * pOut,
+                        uint16_t  dim_im_in_x,
+                        uint16_t  dim_im_in_y,
+                        uint16_t  ch_im_in,
+                        uint16_t  dim_im_out_x,
+                        uint16_t  dim_im_out_y,
+                        uint16_t  dim_kernel_x,
+                        uint16_t  dim_kernel_y,
+                        uint16_t  padding_t,
+                        uint16_t  padding_b,
+                        uint16_t  padding_l,
+                        uint16_t  padding_r,
+                        uint16_t  stride_x,
+                        uint16_t  stride_y);
+
+void xpulp_nn_maxpool_i2(
+                        int8_t * pIn,
+                        int8_t * pOut,
+                        uint16_t  dim_im_in_x,
+                        uint16_t  dim_im_in_y,
+                        uint16_t  ch_im_in,
+                        uint16_t  dim_im_out_x,
+                        uint16_t  dim_im_out_y,
+                        uint16_t  dim_kernel_x,
+                        uint16_t  dim_kernel_y,
+                        uint16_t  padding_t,
+                        uint16_t  padding_b,
+                        uint16_t  padding_l,
+                        uint16_t  padding_r,
+                        uint16_t  stride_x,
+                        uint16_t  stride_y);
+
+void xpulp_nn_avgpool_u8_u8(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u8_i8(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i8_u8(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i8_i8(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u8_u4(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u8_i4(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i8_u4(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i8_i4(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u8_u2(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u8_i2(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i8_u2(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i8_i2(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u4_u8(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u4_i8(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i4_u8(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i4_i8(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u4_u4(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u4_i4(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i4_u4(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i4_i4(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u4_u2(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u4_i2(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i4_u2(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i4_i2(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u2_u8(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u2_i8(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i2_u8(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i2_i8(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u2_u4(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u2_i4(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i2_u4(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i2_i4(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u2_u2(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u2_i2(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i2_u2(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i2_i2(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_add_u8_u8(
+                        uint8_t * pIn1,
+                        uint8_t * pIn2,
+                        uint8_t * pOut,
+                        uint16_t out_mult1,
+                        uint16_t out_mult2,
+                        uint16_t out_shift,
+                        uint16_t dim_im_in_x,
+                        uint16_t dim_im_in_y,
+                        uint16_t ch_im_in);
+
+void xpulp_nn_add_u8_u8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u4(
+                        uint8_t * pIn1,
+                        uint8_t * pIn2,
+                        uint8_t * pOut,
+                        uint16_t out_mult1,
+                        uint16_t out_mult2,
+                        uint16_t out_shift,
+                        uint16_t dim_im_in_x,
+                        uint16_t dim_im_in_y,
+                        uint16_t ch_im_in);
+
+void xpulp_nn_add_u8_u4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u2(
+                        uint8_t * pIn1,
+                        uint8_t * pIn2,
+                        uint8_t * pOut,
+                        uint16_t out_mult1,
+                        uint16_t out_mult2,
+                        uint16_t out_shift,
+                        uint16_t dim_im_in_x,
+                        uint16_t dim_im_in_y,
+                        uint16_t ch_im_in);
+
+void xpulp_nn_add_u8_u2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u4(
+                        uint8_t * pIn1,
+                        uint8_t * pIn2,
+                        uint8_t * pOut,
+                        uint16_t out_mult1,
+                        uint16_t out_mult2,
+                        uint16_t out_shift,
+                        uint16_t dim_im_in_x,
+                        uint16_t dim_im_in_y,
+                        uint16_t ch_im_in);
+
+void xpulp_nn_add_u4_u4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u2(
+                        uint8_t * pIn1,
+                        uint8_t * pIn2,
+                        uint8_t * pOut,
+                        uint16_t out_mult1,
+                        uint16_t out_mult2,
+                        uint16_t out_shift,
+                        uint16_t dim_im_in_x,
+                        uint16_t dim_im_in_y,
+                        uint16_t ch_im_in);
+
+void xpulp_nn_add_u4_u2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u2(
+                        uint8_t * pIn1,
+                        uint8_t * pIn2,
+                        uint8_t * pOut,
+                        uint16_t out_mult1,
+                        uint16_t out_mult2,
+                        uint16_t out_shift,
+                        uint16_t dim_im_in_x,
+                        uint16_t dim_im_in_y,
+                        uint16_t ch_im_in);
+
+void xpulp_nn_add_u2_u2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+
+
+#endif
diff --git a/rt_nn_tests/xpnnv2_conv/pulp_nn_mix_kernels.h b/rt_nn_tests/xpnnv2_conv/pulp_nn_mix_kernels.h
new file mode 100644
index 0000000..8b2a3c6
--- /dev/null
+++ b/rt_nn_tests/xpnnv2_conv/pulp_nn_mix_kernels.h
@@ -0,0 +1,7093 @@
+/*
+ * pulp_nn_kernels.h
+ * Nazareno Bruschi <nazareno.bruschi@unibo.it>
+ *
+ * Copyright (C) 2019-2020 University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef __XPULPNN-MIXED_KERNELS__
+#define __XPULPNN-MIXED_KERNELS__
+
+void xpulp_nn_mix_conv_u8_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u8_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u8_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u8_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u4_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u4_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u4_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u2_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u2_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u2_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u8_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u8_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u8_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u4_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u4_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u4_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u2_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u2_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u2_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u8_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u8_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u8_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u4_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u4_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u4_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u2_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u2_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u2_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_avgpool_u8_u8(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u8_i8(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i8_u8(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i8_i8(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u8_u4(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u8_i4(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i8_u4(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i8_i4(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u8_u2(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u8_i2(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i8_u2(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i8_i2(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u4_u8(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u4_i8(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i4_u8(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i4_i8(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u4_u4(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u4_i4(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i4_u4(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i4_i4(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u4_u2(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u4_i2(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i4_u2(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i4_i2(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u2_u8(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u2_i8(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i2_u8(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i2_i8(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u2_u4(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u2_i4(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i2_u4(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i2_i4(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u2_u2(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u2_i2(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i2_u2(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i2_i2(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_add_u8_u8(
+                        uint8_t * pIn1,
+                        uint8_t * pIn2,
+                        uint8_t * pOut,
+                        uint16_t out_mult1,
+                        uint16_t out_mult2,
+                        uint16_t out_shift,
+                        uint16_t dim_im_in_x,
+                        uint16_t dim_im_in_y,
+                        uint16_t ch_im_in);
+
+void xpulp_nn_add_u8_u8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u4(
+                        uint8_t * pIn1,
+                        uint8_t * pIn2,
+                        uint8_t * pOut,
+                        uint16_t out_mult1,
+                        uint16_t out_mult2,
+                        uint16_t out_shift,
+                        uint16_t dim_im_in_x,
+                        uint16_t dim_im_in_y,
+                        uint16_t ch_im_in);
+
+void xpulp_nn_add_u8_u4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u2(
+                        uint8_t * pIn1,
+                        uint8_t * pIn2,
+                        uint8_t * pOut,
+                        uint16_t out_mult1,
+                        uint16_t out_mult2,
+                        uint16_t out_shift,
+                        uint16_t dim_im_in_x,
+                        uint16_t dim_im_in_y,
+                        uint16_t ch_im_in);
+
+void xpulp_nn_add_u8_u2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u4(
+                        uint8_t * pIn1,
+                        uint8_t * pIn2,
+                        uint8_t * pOut,
+                        uint16_t out_mult1,
+                        uint16_t out_mult2,
+                        uint16_t out_shift,
+                        uint16_t dim_im_in_x,
+                        uint16_t dim_im_in_y,
+                        uint16_t ch_im_in);
+
+void xpulp_nn_add_u4_u4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u2(
+                        uint8_t * pIn1,
+                        uint8_t * pIn2,
+                        uint8_t * pOut,
+                        uint16_t out_mult1,
+                        uint16_t out_mult2,
+                        uint16_t out_shift,
+                        uint16_t dim_im_in_x,
+                        uint16_t dim_im_in_y,
+                        uint16_t ch_im_in);
+
+void xpulp_nn_add_u4_u2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u2(
+                        uint8_t * pIn1,
+                        uint8_t * pIn2,
+                        uint8_t * pOut,
+                        uint16_t out_mult1,
+                        uint16_t out_mult2,
+                        uint16_t out_shift,
+                        uint16_t dim_im_in_x,
+                        uint16_t dim_im_in_y,
+                        uint16_t ch_im_in);
+
+void xpulp_nn_add_u2_u2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+
+
+#endif
\ No newline at end of file
diff --git a/rt_nn_tests/xpnnv2_conv/pulp_nn_utils.h b/rt_nn_tests/xpnnv2_conv/pulp_nn_utils.h
new file mode 100644
index 0000000..44d2251
--- /dev/null
+++ b/rt_nn_tests/xpnnv2_conv/pulp_nn_utils.h
@@ -0,0 +1,2079 @@
+/*
+ * pulp_nn_utils.h
+ * Nazareno   Bruschi  <nazareno.bruschi@unibo.it>
+ * Alessandro Nadalini <alessandro.nadalini3@unibo.it>
+ * Georg Rutishauser   <georgr@iis.ee.ethz.ch>
+ *
+ * Copyright (C) 2019-2020 ETH Zurich & University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef __PULPNN_UTILS__
+#define __PULPNN_UTILS__
+
+#include <stdint.h>
+
+typedef   signed short v2s __attribute__((vector_size (4)));
+
+
+
+#define min(a,b)                                             ((a)<(b)?(a):(b))
+#define log2(x)                                              __builtin_pulp_fl1(x)
+#define PACK_INT2_SIZE(x)                                    ((x) >> 2)
+
+#define MemoryFence()                                        asm volatile("":::"memory")
+
+#define CHANS_DECOMPR(x)                                     (5*x >> 2) // equivalent to division by 0.8
+
+/* Functions for Compressed MAC */
+#define CompressedMAC(sum, ptr, config) asm volatile(                \
+    "pv.smlsdotsp.t %[shum], %[phtr], %[chonfig];"                    \
+    : [shum] "+r" (sum), [phtr] "+r" (ptr): [chonfig] "I" (config))
+
+#define CompressedMACUnsigned(sum, ptr, config) asm volatile(                   \
+    "pv.smlsdotsup.t %[shum], %[phtr], %[chonfig];" \
+    : [shum] "+r" (sum), [phtr] "+r" (ptr): [chonfig] "I" (config))
+
+#define InitNNRF(ptr, config) asm volatile(        \
+    "pv.smlsdotsp.t x0, %[phtr], %[chonfig];"       \
+    : [phtr] "+r" (ptr) : [chonfig] "I" (config))
+
+#define ThresholdCompress(res, val, thrs) asm volatile(                                                \
+    "pv.thrc %[rhes], %[vhal], %[thhrs];" : [rhes] "+r" (res) : [vhal] "r" (val), [thhrs] "r" (thrs))
+
+#define GetConfig(a_update, b_update, a_reg, b_reg) a_update << 4 | b_update << 3 | a_reg << 1 | b_reg
+
+/* Functions for threshold&compress */
+#define check_store(res, pOut)            \
+  if ((res & 0xe0000000) == 0x00000000) { \
+    *pOut = res & 0xff;                   \
+    pOut++;                               \
+    incr_val=ch_out_r; }
+
+#define check_store_4x1(res, pOut)        \
+  if ((res & 0xe0000000) == 0x00000000) { \
+    *pOut = res & 0xff;                   \
+    pOut++; }
+
+#define reset_currThr()                           \
+  if ((uint32_t *) currThr == (uint32_t *) (pThr + ch_out)) { \
+    currThr = (v2s *) pThr;                                    \
+  }
+
+#define MacLoads20(a_update, b_update, a_reg, b_reg, ptr, sum)  __builtin_pulp_mlsdotsp20_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+
+#define MacLoad20(a_update, b_update, a_reg, b_reg, ptr, sum)  __builtin_pulp_mlsdotsup20_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+
+/* Functions for compressed min/max */
+#define CompressedMax(res, in1, in2) asm volatile( \
+                                                   "pv.max.t %[res_], %[in1_], %[in2_];" : [res_] "=r" (res) : [in1_] "r" (in1), [in2_] "r" (in2))
+
+#define CompressedMin(res, in1, in2) asm volatile(                      \
+                                                   "pv.min.t %[res_], %[in1_], %[in2_];" : [res_] "=r" (res) : [in1_] "r" (in1), [in2_] "r" (in2))
+
+#define Max16(res, in1, in2) asm volatile(                      \
+                                                   "pv.max.c %[res_], %[in1_], %[in2_];" : [res_] "=r" (res) : [in1_] "r" (in1), [in2_] "r" (in2))
+
+#define Min16(res, in1, in2) asm volatile(                      \
+                                                                        "pv.min.c %[res_], %[in1_], %[in2_];" : [res_] "=r" (res) : [in1_] "r" (in1), [in2_] "r" (in2))
+
+#define thr_cmp(state, val, threshs)  __builtin_pulp_thresh_compr(state, val, threshs)
+
+typedef unsigned char  v4u __attribute__((vector_size (4)));
+typedef   signed char  v4s __attribute__((vector_size (4)));
+
+#define bitext(x,size,off)                                      __builtin_pulp_bextract(x,size,off)
+#define bitextu(x,size,off)                                     __builtin_pulp_bextractu(x,size,off)
+#ifdef __clang__
+#define bitins(dst,not_mask_imm,src,mask_imm,off)               __builtin_binsert(dst,not_mask_imm,src,mask_imm,off)
+#else
+#define bitins(dst,not_mask_imm,src,mask_imm,off)               __builtin_pulp_binsert(dst,not_mask_imm,src,mask_imm,off)
+#endif
+#define pack(x,y,z,t)                                           __builtin_pulp_pack4(x,y,z,t)
+#define max4(a,b)                                               __builtin_pulp_maxu4(a,b)
+#define maxs4(a, b)                                             __builtin_pulp_max4(a, b)
+#define max8(a, b)                                              __builtin_pulp_maxu8(a, b)
+#define maxs8(a, b)                                             __builtin_pulp_max8(a, b)
+#define max16(a, b)                                             __builtin_pulp_maxu16(a, b)
+#define maxs16(a, b)                                            __builtin_pulp_max16(a, b)
+#define maxs20(a, b)                                            __builtin_pulp_max20(a, b)
+#define max32(a,b)                                              __builtin_pulp_maxusi(a,b)
+#define maxs32(a,b)                                             __builtin_pulp_maxsi(a,b)
+#define min32(a,b)                                              __builtin_pulp_minusi(a,b)
+#define mins32(a,b)                                             __builtin_pulp_minsi(a,b)
+#define min4(a, b)                                              __builtin_pulp_minu4(a, b)
+#define mins4(a, b)                                             __builtin_pulp_min4(a, b)
+#define min8(a, b)                                              __builtin_pulp_minu8(a, b)
+#define mins8(a, b)                                             __builtin_pulp_min8(a, b)
+#define min16(a, b)                                             __builtin_pulp_minu16(a, b)
+#define mins16(a, b)                                            __builtin_pulp_min16(a, b)
+#define mins20(a, b)                                            __builtin_pulp_min20(a, b)
+#define avg4(a,b)                                               __builtin_pulp_avgu4(a,b)
+#define avg8(a,b)                                               __builtin_pulp_avgu8(a,b)
+#define avg16(a,b)                                              __builtin_pulp_avgu16(a,b)
+#define log2(x)                                                 __builtin_pulp_fl1(x)
+#define min(a,b)                                                ((a)<(b)?(a):(b))
+#define SumDotp4(a, b, c)                                       __builtin_pulp_sdotusp4(a, b, c)
+#define SumDotp8(a, b, c)                                       __builtin_pulp_sdotusp8(a, b, c)
+#define SumDotp16(a, b, c)                                      __builtin_pulp_sdotusp16(a, b, c)
+#define SumDotps4(a, b, c)                                      __builtin_pulp_sdotsp4(a, b, c)
+#define SumDotps8(a, b, c)                                      __builtin_pulp_sdotsp8(a, b, c)
+#define SumDotps16(a, b, c)                                     __builtin_pulp_sdotsp16(a, b, c)
+#define clip4(x)                                                __builtin_pulp_clipu_r(x, 15)
+#define clip2(x)                                                __builtin_pulp_clipu_r(x, 3)
+#define clip8(x)                                                __builtin_pulp_clipu_r(x, 255)
+
+#define clips4(x)                                               __builtin_pulp_clip_r(x, 7)
+#define clips2(x)                                               __builtin_pulp_clip_r(x, 1)
+#define clips8(x)                                               __builtin_pulp_clip_r(x, 127)
+#define MacLoadInit(a_update, b_update, a_reg, b_reg, ptr)      __builtin_pulp_mlinitspr_v3(a_update, b_update, a_reg, b_reg, ptr)
+#define MacLoadUpdate(ptr)                                      __builtin_pulp_mlupdatespr_v3(ptr)
+#define MacLoadAssign(ptr)                                      __builtin_pulp_mlassignspr_v3(ptr)
+#define MacLoad4(a_update, b_update, a_reg, b_reg, ptr, sum)    __builtin_pulp_mlsdotsup4_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+#define MacLoad8(a_update, b_update, a_reg, b_reg, ptr, sum)    __builtin_pulp_mlsdotsup8_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+#define MacLoad16(a_update, b_update, a_reg, b_reg, ptr, sum)   __builtin_pulp_mlsdotsup16_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+#define MacLoads4(a_update, b_update, a_reg, b_reg, ptr, sum)   __builtin_pulp_mlsdotsp4_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+#define MacLoads8(a_update, b_update, a_reg, b_reg, ptr, sum)   __builtin_pulp_mlsdotsp8_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+#define MacLoads16(a_update, b_update, a_reg, b_reg, ptr, sum)  __builtin_pulp_mlsdotsp16_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+
+#define PACK_INT8_SIZE(x)                                    (x)
+#define PACK_INT4_SIZE(x)                                    ((x) >> 1)
+#define PACK_INT2_SIZE(x)                                    ((x) >> 2)
+
+#define MemoryFence()                                        asm volatile("":::"memory")
+
+#define LEGACY_MODE(x)                                       asm volatile ("csrwi 0x010," x)
+#define IVEC_FMT(x)                                          asm volatile ("csrwi 0x00D," x)
+#define MIXED_SKIP(x)                                        asm volatile ("csrwi 0x00F," x)
+#define A_ADDRESS(x)                                         asm volatile ("csrw 0x100, %0" :: "r" (x))
+#define W_ADDRESS(x)                                         asm volatile ("csrw 0x101, %0" :: "r" (x))
+#define A_STRIDE(x)                                          asm volatile ("csrw 0x102, %0":: "r" (x))
+#define W_STRIDE(x)                                          asm volatile ("csrw 0x103, %0":: "r" (x))
+#define A_ROLLBACK(x)                                        asm volatile ("csrw 0x104, %0":: "r" (x))
+#define W_ROLLBACK(x)                                        asm volatile ("csrw 0x105, %0":: "r" (x))
+#define A_SKIP(x)                                            asm volatile ("csrwi 0x106," x)
+#define W_SKIP(x)                                            asm volatile ("csrwi 0x107," x)
+
+static uint8_t __attribute__((noinline)) pulp_nn_quant_u2 (
+  int32_t phi,
+  int16_t m,
+  int8_t d
+  ) {
+  int32_t x = (m * phi) >> d;
+  uint8_t res = clip2(x);
+  return res;
+}
+static uint8_t __attribute__((noinline)) pulp_nn_bn_quant_u2 (
+  int32_t phi,
+  int32_t k,
+  int32_t lambda,
+  int8_t d
+  ) {
+  int32_t integer_image_phi = (k * phi) + lambda;
+  int32_t x = (integer_image_phi) >> d;
+  uint8_t res = clip2(x);
+  return res;
+}
+static uint8_t __attribute__((noinline)) pulp_nn_add_quant_u2 (
+  uint8_t pix1,
+  uint8_t pix2,
+  int16_t m1,
+  int16_t m2,
+  int8_t d
+  ) {
+  uint32_t integer_image = pix1*m1 + pix2*m2;
+  uint32_t x = (integer_image) >> d;
+  uint8_t res = clip2(x);
+  return res;
+}
+
+static int8_t __attribute__((noinline)) pulp_nn_quant_i2 (
+  int32_t phi,
+  int16_t m,
+  int8_t d
+  ) {
+  int32_t x = (m * phi) >> d;
+  int8_t res = clips2(x);
+  return res;
+}
+static int8_t __attribute__((noinline)) pulp_nn_bn_quant_i2 (
+  int32_t phi,
+  int32_t k,
+  int32_t lambda,
+  int8_t d
+  ) {
+  int32_t integer_image_phi = (k * phi) + lambda;
+  int32_t x = (integer_image_phi) >> d;
+  int8_t res = clips2(x);
+  return res;
+}
+static int8_t __attribute__((noinline)) pulp_nn_add_quant_i2 (
+  int8_t pix1,
+  int8_t pix2,
+  int16_t m1,
+  int16_t m2,
+  int8_t d
+  ) {
+  int32_t integer_image = pix1*m1 + pix2*m2;
+  int32_t x = (integer_image) >> d;
+  int8_t res = clips2(x);
+  return res;
+}
+
+static uint8_t __attribute__((noinline)) pulp_nn_quant_u4 (
+  int32_t phi,
+  int16_t m,
+  int8_t d
+  ) {
+  int32_t x = (m * phi) >> d;
+  uint8_t res = clip4(x);
+  return res;
+}
+static uint8_t __attribute__((noinline)) pulp_nn_bn_quant_u4 (
+  int32_t phi,
+  int32_t k,
+  int32_t lambda,
+  int8_t d
+  ) {
+  int32_t integer_image_phi = (k * phi) + lambda;
+  int32_t x = (integer_image_phi) >> d;
+  uint8_t res = clip4(x);
+  return res;
+}
+static uint8_t __attribute__((noinline)) pulp_nn_add_quant_u4 (
+  uint8_t pix1,
+  uint8_t pix2,
+  int16_t m1,
+  int16_t m2,
+  int8_t d
+  ) {
+  uint32_t integer_image = pix1*m1 + pix2*m2;
+  uint32_t x = (integer_image) >> d;
+  uint8_t res = clip4(x);
+  return res;
+}
+
+static int8_t __attribute__((noinline)) pulp_nn_quant_i4 (
+  int32_t phi,
+  int16_t m,
+  int8_t d
+  ) {
+  int32_t x = (m * phi) >> d;
+  int8_t res = clips4(x);
+  return res;
+}
+static int8_t __attribute__((noinline)) pulp_nn_bn_quant_i4 (
+  int32_t phi,
+  int32_t k,
+  int32_t lambda,
+  int8_t d
+  ) {
+  int32_t integer_image_phi = (k * phi) + lambda;
+  int32_t x = (integer_image_phi) >> d;
+  int8_t res = clips4(x);
+  return res;
+}
+static int8_t __attribute__((noinline)) pulp_nn_add_quant_i4 (
+  int8_t pix1,
+  int8_t pix2,
+  int16_t m1,
+  int16_t m2,
+  int8_t d
+  ) {
+  int32_t integer_image = pix1*m1 + pix2*m2;
+  int32_t x = (integer_image) >> d;
+  int8_t res = clips4(x);
+  return res;
+}
+
+static uint8_t __attribute__((noinline)) pulp_nn_quant_u8 (
+  int32_t phi,
+  int16_t m,
+  int8_t d
+  ) {
+  int32_t x = (m * phi) >> d;
+  uint8_t res = clip8(x);
+  return res;
+}
+static uint8_t __attribute__((noinline)) pulp_nn_bn_quant_u8 (
+  int32_t phi,
+  int32_t k,
+  int32_t lambda,
+  int8_t d
+  ) {
+  int32_t integer_image_phi = (k * phi) + lambda;
+  int32_t x = (integer_image_phi) >> d;
+  uint8_t res = clip8(x);
+  return res;
+}
+static uint8_t __attribute__((noinline)) pulp_nn_add_quant_u8 (
+  uint8_t pix1,
+  uint8_t pix2,
+  int16_t m1,
+  int16_t m2,
+  int8_t d
+  ) {
+  uint32_t integer_image = pix1*m1 + pix2*m2;
+  uint32_t x = (integer_image) >> d;
+  uint8_t res = clip8(x);
+  return res;
+}
+
+static int8_t __attribute__((noinline)) pulp_nn_quant_i8 (
+  int32_t phi,
+  int16_t m,
+  int8_t d
+  ) {
+  int32_t x = (m * phi) >> d;
+  int8_t res = clips8(x);
+  return res;
+}
+static int8_t __attribute__((noinline)) pulp_nn_bn_quant_i8 (
+  int32_t phi,
+  int32_t k,
+  int32_t lambda,
+  int8_t d
+  ) {
+  int32_t integer_image_phi = (k * phi) + lambda;
+  int32_t x = (integer_image_phi) >> d;
+  int8_t res = clips8(x);
+  return res;
+}
+static int8_t __attribute__((noinline)) pulp_nn_add_quant_i8 (
+  int8_t pix1,
+  int8_t pix2,
+  int16_t m1,
+  int16_t m2,
+  int8_t d
+  ) {
+  int32_t integer_image = pix1*m1 + pix2*m2;
+  int32_t x = (integer_image) >> d;
+  int8_t res = clips8(x);
+  return res;
+}
+
+
+static uint8_t __attribute__((noinline)) pulp_nn_u4_quant(int input, int16_t * pThr)
+{
+  if(input <= pThr[7] )
+  {
+    if( input <= pThr[3])
+    {
+      if( input <= pThr[1])
+      {
+        if( input <= pThr[0])
+          return 0;
+        else
+          return 1;
+      }
+      else
+      {
+        if( input <= pThr[2])
+          return 2;
+        else
+          return 3;
+      }
+    }
+    else
+    {
+      if( input <= pThr[5])
+      {
+        if( input <= pThr[4])
+          return 4;
+        else
+          return 5;
+      }
+      else
+      {
+        if( input <= pThr[6])
+          return 6;
+        else
+          return 7;
+      }
+    }
+  }
+  else
+  {
+    if( input <= pThr[11])
+    {
+      if( input <= pThr[9])
+      {
+        if( input <= pThr[8])
+          return 8;
+        else
+          return 9;
+      }
+      else
+      {
+        if( input <= pThr[10])
+          return 10;
+        else
+          return 11;
+      }
+    }
+    else
+    {
+      if( input <= pThr[13])
+      {
+        if( input <= pThr[12])
+          return 12;
+        else
+          return 13;
+      }
+      else
+      {
+        if( input <= pThr[14])
+          return 14;
+        else
+          return 15;
+      }
+    }
+  }
+}
+
+static uint8_t __attribute__((noinline)) pulp_nn_u2_quant(int input, int16_t * pThr)
+{
+  if( input <= pThr[1])
+  {
+    if( input <= pThr[0])
+        {
+      return 0;
+    }
+        else
+        {
+      return 1;
+    }
+  }
+    else
+    {
+    if( input <= pThr[2])
+        {
+      return 2;
+    }
+        else
+        {
+      return 3;
+    }
+  }
+}
+
+/*
+ * Common
+ */
+
+
+static v4s __attribute__((noinline)) pulp_nn_i4_to_i8_r( int8_t *pSrc)
+{
+  v4s Src = *((v4s*) pSrc);
+  int8_t bext1, bext2, bext3, bext4;
+
+  bext1 = (int8_t) bitext((int) Src, 4, 0);
+  bext2 = (int8_t) bitext((int) Src, 4, 4);
+  bext3 = (int8_t) bitext((int) Src, 4, 8);
+  bext4 = (int8_t) bitext((int) Src, 4, 12);
+  v4s res = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+
+  return res;
+}
+
+static v4u __attribute__((noinline)) pulp_nn_u4_to_u8_r(uint8_t *pSrc)
+{
+  v4u Src = *((v4u*) pSrc);
+  uint8_t bext1, bext2, bext3, bext4;
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 4, 0);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 4, 4);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 4, 8);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 4, 12);
+  v4u res = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+
+  return res;
+}
+
+static v4s __attribute__((noinline)) pulp_nn_i2_to_i8_r( int8_t *pSrc)
+{
+  v4s Src = *((v4s*) pSrc);
+  int8_t bext1, bext2, bext3, bext4;
+
+  bext1 = (int8_t) bitext((int) Src, 2, 0);
+  bext2 = (int8_t) bitext((int) Src, 2, 2);
+  bext3 = (int8_t) bitext((int) Src, 2, 4);
+  bext4 = (int8_t) bitext((int) Src, 2, 6);
+  v4s res = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+
+  return res;
+}
+
+static v4u __attribute__((noinline)) pulp_nn_u2_to_u8_r(uint8_t *pSrc)
+{
+  v4u Src = *((v4u*) pSrc);
+  uint8_t bext1, bext2, bext3, bext4;
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 0);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 2);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 4);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 6);
+  v4u res = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+
+  return res;
+}
+
+static v4s __attribute__((noinline)) pulp_nn_i2_to_i4_r(int8_t *pSrc)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  v4s Src = *((v4s*) pSrc);
+  int8_t bext1, bext2, bext3, bext4;
+  int8_t out1, out2, out3, out4;
+
+  bext1 = (int8_t) bitextu((int) Src, 2, 0);
+  bext2 = (int8_t) bitextu((int) Src, 2, 2);
+  bext3 = (int8_t) bitextu((int) Src, 2, 4);
+  bext4 = (int8_t) bitextu((int) Src, 2, 6);
+
+  out1 = bitins(bext1, n_mask, bext2, mask, off);
+  out2 = bitins(bext3, n_mask, bext4, mask, off);
+
+  bext1 = (int8_t) bitextu((int) Src, 2, 8);
+  bext2 = (int8_t) bitextu((int) Src, 2, 10);
+  bext3 = (int8_t) bitextu((int) Src, 2, 12);
+  bext4 = (int8_t) bitextu((int) Src, 2, 14);
+
+  out3 = bitins(bext1, n_mask, bext2, mask, off);
+  out4 = bitins(bext3, n_mask, bext4, mask, off);
+
+  v4s res = pack((int8_t) out1, (int8_t) out2, (int8_t) out3, (int8_t) out4);
+
+  return res;
+}
+
+static v4u __attribute__((noinline)) pulp_nn_u2_to_u4_r(uint8_t *pSrc)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  v4u Src = *((v4u*) pSrc);
+  uint8_t bext1, bext2, bext3, bext4;
+  uint8_t out1, out2, out3, out4;
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 0);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 2);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 4);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 6);
+
+  out1 = bitins(bext1, n_mask, bext2, mask, off);
+  out2 = bitins(bext3, n_mask, bext4, mask, off);
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 8);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 10);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 12);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 14);
+
+  out3 = bitins(bext1, n_mask, bext2, mask, off);
+  out4 = bitins(bext3, n_mask, bext4, mask, off);
+
+  v4u res = pack((uint8_t) out1, (uint8_t) out2, (uint8_t) out3, (uint8_t) out4);
+
+  return res;
+}
+
+static int8_t *__attribute__((always_inline)) pulp_nn_i4_to_i8( int8_t *pSrc, int8_t *pDst)
+{
+  v4s Src = *((v4s*) pSrc);
+  int8_t bext1, bext2, bext3, bext4;
+
+  pSrc+=4;
+
+  bext1 = (int8_t) bitext((int) Src, 4, 0);
+  bext2 = (int8_t) bitext((int) Src, 4, 4);
+  bext3 = (int8_t) bitext((int) Src, 4, 8);
+  bext4 = (int8_t) bitext((int) Src, 4, 12);
+  *((v4s*)pDst) = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+   MemoryFence();
+  pDst+=4;
+  bext1 = (int8_t) bitext((int) Src, 4, 16);
+  bext2 = (int8_t) bitext((int) Src, 4, 20);
+  bext3 = (int8_t) bitext((int) Src, 4, 24);
+  bext4 = (int8_t) bitext((int) Src, 4, 28);
+  *((v4s*)pDst) = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+
+  return pSrc;
+}
+
+static uint8_t *__attribute__((always_inline)) pulp_nn_u4_to_u8(uint8_t *pSrc, uint8_t *pDst)
+{
+  v4u Src = *((v4u*) pSrc);
+  uint8_t bext1, bext2, bext3, bext4;
+
+  pSrc+=4;
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 4, 0);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 4, 4);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 4, 8);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 4, 12);
+  *((v4u*)pDst) = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 4, 16);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 4, 20);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 4, 24);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 4, 28);
+  *((v4u*)pDst) = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+
+  return pSrc;
+}
+
+static int8_t *__attribute__((always_inline)) pulp_nn_i2_to_i8( int8_t * pSrc, int8_t * pDst)
+{
+  v4s Src = *((v4s*) pSrc);
+  int8_t bext1, bext2, bext3, bext4;
+
+  pSrc+=4;
+
+  bext1 = (int8_t) bitext((int) Src, 2, 0);
+  bext2 = (int8_t) bitext((int) Src, 2, 2);
+  bext3 = (int8_t) bitext((int) Src, 2, 4);
+  bext4 = (int8_t) bitext((int) Src, 2, 6);
+  *((v4s*)pDst) = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (int8_t) bitext((int) Src, 2, 8);
+  bext2 = (int8_t) bitext((int) Src, 2, 10);
+  bext3 = (int8_t) bitext((int) Src, 2, 12);
+  bext4 = (int8_t) bitext((int) Src, 2, 14);
+  *((v4s*)pDst) = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (int8_t) bitext((int) Src, 2, 16);
+  bext2 = (int8_t) bitext((int) Src, 2, 18);
+  bext3 = (int8_t) bitext((int) Src, 2, 20);
+  bext4 = (int8_t) bitext((int) Src, 2, 22);
+  *((v4s*)pDst) = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (int8_t) bitext((int) Src, 2, 24);
+  bext2 = (int8_t) bitext((int) Src, 2, 26);
+  bext3 = (int8_t) bitext((int) Src, 2, 28);
+  bext4 = (int8_t) bitext((int) Src, 2, 30);
+  *((v4s*)pDst) = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+
+  return pSrc;
+}
+
+static uint8_t *__attribute__((always_inline)) pulp_nn_u2_to_u8(uint8_t * pSrc, uint8_t * pDst)
+{
+  v4u Src = *((v4u*) pSrc);
+  uint8_t bext1, bext2, bext3, bext4;
+
+  pSrc+=4;
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 0);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 2);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 4);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 6);
+  *((v4u*)pDst) = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 8);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 10);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 12);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 14);
+  *((v4u*)pDst) = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 16);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 18);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 20);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 22);
+  *((v4u*)pDst) = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 24);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 26);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 28);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 30);
+  *((v4u*)pDst) = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+
+  return pSrc;
+}
+
+static int8_t *__attribute__((always_inline)) pulp_nn_i2_to_i4( int8_t * pSrc, int8_t * pDst)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  v4s Src = *((v4s*) pSrc);
+  int8_t bext1, bext2, bext3, bext4;
+  int8_t out1, out2, out3, out4;
+
+  pSrc+=4;
+
+  bext1 = (int8_t) bitext((int) Src, 2, 0);
+  bext2 = (int8_t) bitext((int) Src, 2, 2);
+  bext3 = (int8_t) bitext((int) Src, 2, 4);
+  bext4 = (int8_t) bitext((int) Src, 2, 6);
+
+  out1 = bitins(bext1, n_mask, bext2, mask, off);
+  out2 = bitins(bext3, n_mask, bext4, mask, off);
+  MemoryFence();
+
+  bext1 = (int8_t) bitext((int) Src, 2, 8);
+  bext2 = (int8_t) bitext((int) Src, 2, 10);
+  bext3 = (int8_t) bitext((int) Src, 2, 12);
+  bext4 = (int8_t) bitext((int) Src, 2, 14);
+
+  out3 = bitins(bext1, n_mask, bext2, mask, off);
+  out4 = bitins(bext3, n_mask, bext4, mask, off);
+
+  *((v4s*)pDst) = pack((int8_t) out1, (int8_t) out2, (int8_t) out3, (int8_t) out4);
+  MemoryFence();
+
+  pDst+=4;
+  bext1 = (int8_t) bitext((int) Src, 2, 16);
+  bext2 = (int8_t) bitext((int) Src, 2, 18);
+  bext3 = (int8_t) bitext((int) Src, 2, 20);
+  bext4 = (int8_t) bitext((int) Src, 2, 22);
+
+  out1 = bitins(bext1, n_mask, bext2, mask, off);
+  out2 = bitins(bext3, n_mask, bext4, mask, off);
+  MemoryFence();
+
+  bext1 = (int8_t) bitext((int) Src, 2, 24);
+  bext2 = (int8_t) bitext((int) Src, 2, 26);
+  bext3 = (int8_t) bitext((int) Src, 2, 28);
+  bext4 = (int8_t) bitext((int) Src, 2, 30);
+
+  out3 = bitins(bext1, n_mask, bext2, mask, off);
+  out4 = bitins(bext3, n_mask, bext4, mask, off);
+
+  *((v4s*)pDst) = pack((int8_t) out1, (int8_t) out2, (int8_t) out3, (int8_t) out4);
+
+  return pSrc;
+}
+
+static uint8_t *__attribute__((always_inline)) pulp_nn_u2_to_u4( uint8_t * pSrc, uint8_t * pDst)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  v4u Src = *((v4u*) pSrc);
+  uint8_t bext1, bext2, bext3, bext4;
+  uint8_t out1, out2, out3, out4;
+
+  pSrc+=4;
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 0);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 2);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 4);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 6);
+
+  out1 = bitins(bext1, n_mask, bext2, mask, off);
+  out2 = bitins(bext3, n_mask, bext4, mask, off);
+  MemoryFence();
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 8);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 10);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 12);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 14);
+
+  out3 = bitins(bext1, n_mask, bext2, mask, off);
+  out4 = bitins(bext3, n_mask, bext4, mask, off);
+
+  *((v4u*)pDst) = pack((uint8_t) out1, (uint8_t) out2, (uint8_t) out3, (uint8_t) out4);
+  MemoryFence();
+
+  pDst+=4;
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 16);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 18);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 20);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 22);
+
+  out1 = bitins(bext1, n_mask, bext2, mask, off);
+  out2 = bitins(bext3, n_mask, bext4, mask, off);
+  MemoryFence();
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 24);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 26);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 28);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 30);
+
+  out3 = bitins(bext1, n_mask, bext2, mask, off);
+  out4 = bitins(bext3, n_mask, bext4, mask, off);
+  
+  *((v4u*)pDst) = pack((uint8_t) out1, (uint8_t) out2, (uint8_t) out3, (uint8_t) out4);
+
+  return pSrc;
+}
+
+/*
+ * XpulpV2
+ */
+
+static void __attribute__((noinline)) pulp_zero_mem(uint8_t * pBuffer, unsigned int size)
+{
+  int lfover = size &0x3;
+  for (int i=0; i<(size>>2); i++)
+  {
+    *((v4u *)pBuffer) = (v4u){0,0,0,0};
+    MemoryFence();
+    pBuffer+=4;
+  }
+  while(lfover)
+  {
+    *pBuffer++=0;
+    lfover--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_im2col_u2_to_u8(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_u2_to_u8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 16;
+  }
+  while (lfover)
+  {
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 0);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 2);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 4);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 6);
+    pOutput++;
+    pInput++;
+    lfover -= 4;
+  }
+}
+static void __attribute__((noinline)) pulp_nn_im2col_i2_to_i8(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_i2_to_i8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 16;
+  }
+  while (lfover)
+  {
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 0);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 2);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 4);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 6);
+    pOutput++;
+    pInput++;
+    lfover -= 4;
+  }
+}
+static void __attribute__((noinline)) pulp_nn_im2col_u4_to_u8(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 3u;
+  int lfover = blockSize & 0x07;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_u4_to_u8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 8;
+  }
+  while (lfover)
+  {
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 4, 0);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 4, 4);
+    pOutput++;
+    pInput++;
+    lfover -= 2;
+  }
+}
+static void __attribute__((noinline)) pulp_nn_im2col_i4_to_i8(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 3u;
+  int lfover = blockSize & 0x07;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_i4_to_i8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 8;
+  }
+  while (lfover)
+  {
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 4, 0);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 4, 4);
+    pOutput++;
+    pInput++;
+    lfover -= 2;
+  }
+}
+static void __attribute__((noinline)) pulp_nn_im2col_u8_to_u8(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 2u;
+  int lfover = blockSize & 0x03;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4u*)pOutput) = *((v4u*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    *((uint8_t*)pOutput) = *((uint8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover--;
+  }
+}
+static void __attribute__((noinline)) pulp_nn_im2col_i8_to_i8(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 2u;
+  int lfover = blockSize & 0x03;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4s*)pOutput) = *((v4s*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    *((int8_t*)pOutput) = *((int8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover--;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_im2col_u2_to_u2(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4u*)pOutput) = *((v4u*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    uint8_t extr;
+    *((uint8_t*)pOutput) = *((uint8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover -= 4;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_i2_to_i2(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4s*)pOutput) = *((v4s*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    int8_t extr;
+    *((int8_t*)pOutput) = *((int8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover -= 4;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_u2_to_u4(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_u2_to_u4(pInput, pOutput);
+    MemoryFence();
+    pOutput += 8;
+  }
+  while (lfover)
+  {
+    uint8_t extr;
+    // TODO: this is potentially dangerous/wrong if the number of channels
+    // is not a multiple of 8!
+    *((v4u *)pOutput) = pulp_nn_u2_to_u4_r(pInput);
+    pInput += 2;
+    pOutput += 4;
+    lfover -= 8;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_i2_to_i4(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_i2_to_i4(pInput, pOutput);
+    MemoryFence();
+    pOutput += 8;
+  }
+  while (lfover)
+  {
+    int8_t extr;
+    // TODO: this is potentially dangerous/wrong if the number of channels
+    // is not a multiple of 8!
+    *((v4s *)pOutput) = pulp_nn_i2_to_i4_r(pInput);
+    pInput += 2;
+    pOutput += 4;
+    lfover -= 8;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_u2_to_u8(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_u2_to_u8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 16;
+  }
+  while (lfover)
+  {
+    uint8_t extr;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 0);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 2);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 4);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 6);
+    pOutput++;
+    pInput++;
+    lfover -= 4;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_i2_to_i8(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_i2_to_i8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 16;
+  }
+  while (lfover)
+  {
+    int8_t extr;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 0);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 2);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 4);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 6);
+    pOutput++;
+    pInput++;
+    lfover -= 4;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_u4_to_u4(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 3u;
+  int lfover = blockSize & 0x07;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4u*)pOutput) = *((v4u*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    uint8_t extr;
+    *((uint8_t*)pOutput) = *((uint8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover -= 2;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_i4_to_i4(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 3u;
+  int lfover = blockSize & 0x07;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4s*)pOutput) = *((v4s*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    int8_t extr;
+    *((int8_t*)pOutput) = *((int8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover -= 2;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_u4_to_u8(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 3u;
+  int lfover = blockSize & 0x07;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_u4_to_u8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 8;
+  }
+  while (lfover)
+  {
+    uint8_t extr;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 4, 0);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 4, 4);
+    pOutput++;
+    pInput++;
+    lfover -= 2;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_i4_to_i8(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 3u;
+  int lfover = blockSize & 0x07;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_i4_to_i8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 8;
+  }
+  while (lfover)
+  {
+    int8_t extr;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 4, 0);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 4, 4);
+    pOutput++;
+    pInput++;
+    lfover -= 2;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_u8_to_u8(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 2u;
+  int lfover = blockSize & 0x03;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4u*)pOutput) = *((v4u*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    uint8_t extr;
+    *((uint8_t*)pOutput) = *((uint8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover -= 1;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_i8_to_i8(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 2u;
+  int lfover = blockSize & 0x03;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4s*)pOutput) = *((v4s*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    int8_t extr;
+    *((int8_t*)pOutput) = *((int8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover -= 1;
+  }
+}
+
+
+
+static void __attribute__((noinline)) pulp_nn_compare_and_replace_if_larger_u8(uint8_t * base,
+                                                uint8_t * target,
+                                                uint16_t length)
+{
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+  v4u inp;
+  v4u com;
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    inp = *((v4u*)pIn);
+    com = *((v4u*)pCom);
+
+    *((v4u*)pIn) = max4(inp, com);
+
+    pCom+=4;
+    pIn+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    if(*pIn<*pCom)
+      *pIn=*pCom;
+    
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_compare_and_replace_if_larger_i8(
+    int8_t * base, int8_t * target, uint16_t length) {
+  int8_t *pIn = base;
+  int8_t *pCom = target;
+  v4s inp;
+  v4s com;
+  int cnt = length >> 2;
+
+  while (cnt > 0u) {
+    inp = *((v4s *)pIn);
+    com = *((v4s *)pCom);
+
+    *((v4s *)pIn) = maxs4(inp, com);
+
+    pCom += 4;
+    pIn += 4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left > 0u) {
+    if (*pIn < *pCom)
+      *pIn = *pCom;
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_avg_and_replace_u8(uint8_t * base,
+                                  uint8_t * target,
+                                  uint16_t length)
+{
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+
+  while (length>0u)
+  {
+    *pIn = ((*pIn + *pCom) >> 1);
+
+    pIn++;
+    pCom++;
+    length--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_compare_and_replace_if_larger_u4(uint8_t * base,
+                                                uint8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+  v4u inp[2];
+  v4u com[2];
+  uint8_t *out;
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    pulp_nn_u4_to_u8(pIn, (uint8_t *)inp);
+    pulp_nn_u4_to_u8(pCom, (uint8_t *)com);
+
+    *((v4u *)out) = max4(inp[0], com[0]);
+
+    *((uint8_t*)pIn) = bitins(*out, n_mask, *(out + 1), mask, off);
+    pIn++;
+    *((uint8_t*)pIn) = bitins(*(out + 2), n_mask, *(out + 3), mask, off);
+    pIn++;
+
+    *((v4u *)out) = max4(inp[1], com[1]);
+
+    *((uint8_t*)pIn) = bitins(*out, n_mask, *(out + 1), mask, off);
+    pIn++;
+    *((uint8_t*)pIn) = bitins(*(out + 2), n_mask, *(out + 3), mask, off);
+    pIn++;
+
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 4, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 4, 4);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 4, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 4, 4);
+
+    if(inA0<inB0)
+      inA0=inB0;
+
+    if(inA1<inB1)
+      inA1=inB1;
+
+    *((uint8_t*)pIn) = bitins(inA0, n_mask, inA1, mask, off);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_compare_and_replace_if_larger_i4(int8_t * base,
+                                                int8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  int8_t *pIn = base;
+  int8_t *pCom = target;
+  v4s inp[2];
+  v4s com[2];
+  int8_t *out;
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    pulp_nn_i4_to_i8(pIn, (int8_t *)inp);
+    pulp_nn_i4_to_i8(pCom, (int8_t *)com);
+
+    *((v4s *)out) = maxs4(inp[0], com[0]);
+
+    *((int8_t*)pIn) = bitins(*out, n_mask, *(out + 1), mask, off);
+    pIn++;
+    *((int8_t*)pIn) = bitins(*(out + 2), n_mask, *(out + 3), mask, off);
+    pIn++;
+
+    *((v4s *)out) = maxs4(inp[1], com[1]);
+
+    *((int8_t*)pIn) = bitins(*out, n_mask, *(out + 1), mask, off);
+    pIn++;
+    *((int8_t*)pIn) = bitins(*(out + 2), n_mask, *(out + 3), mask, off);
+    pIn++;
+
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    int8_t inA0 = (int8_t) bitext((int) *pIn, 4, 0);
+    int8_t inA1 = (int8_t) bitext((int) *pIn, 4, 4);
+    int8_t inB0 = (int8_t) bitext((int) *pCom, 4, 0);
+    int8_t inB1 = (int8_t) bitext((int) *pCom, 4, 4);
+
+    if(inA0<inB0)
+      inA0=inB0;
+
+    if(inA1<inB1)
+      inA1=inB1;
+
+    *((int8_t*)pIn) = bitins(inA0, n_mask, inA1, mask, off);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_avg_and_replace_u4(uint8_t * base,
+                                  uint8_t * target,
+                                  uint16_t length)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+
+  while (length>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 4, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 4, 4);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 4, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 4, 4);
+
+    inA0 = ((inA0 + inB0) >> 1);
+    inA1 = ((inA1 + inB1) >> 1);
+
+    *((uint8_t*)pIn) = bitins(inA0, n_mask, inA1, mask, off);
+
+    pIn++;
+    pCom++;
+    length--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_compare_and_replace_if_larger_u2(uint8_t * base,
+                                                uint8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask2 = 0x0c;
+  int8_t n_mask2 = ~ mask2;
+  int8_t mask4 = 0x30;
+  int8_t n_mask4 = ~ mask4;
+  int8_t mask6 = 0xc0;
+  int8_t n_mask6 = ~ mask6;
+  int8_t off2 = 2;
+  int8_t off4 = 4;
+  int8_t off6 = 6;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+  v4u inp[4];
+  v4u com[4];
+  uint8_t *out;
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    pulp_nn_u2_to_u8(pIn, inp);
+    pulp_nn_u2_to_u8(pCom, com);
+
+    *((v4u*)out) = max4(inp[0], com[0]);
+
+    uint8_t inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    *((v4u*)out) = max4(inp[1], com[1]);
+
+    inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    *((v4u*)out) = max4(inp[2], com[2]);
+
+    inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    *((v4u*)out) = max4(inp[3], com[3]);
+
+    inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 2, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 2, 2);
+    uint8_t inA2 = (uint8_t) bitextu((unsigned int) *pIn, 2, 4);
+    uint8_t inA3 = (uint8_t) bitextu((unsigned int) *pIn, 2, 6);
+    v4u inA4 = pack((uint8_t) inA0, (uint8_t) inA1, (uint8_t) inA2, (uint8_t) inA3);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 2, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 2, 2);
+    uint8_t inB2 = (uint8_t) bitextu((unsigned int) *pCom, 2, 4);
+    uint8_t inB3 = (uint8_t) bitextu((unsigned int) *pCom, 2, 6);
+    v4u inB4 = pack((uint8_t) inB0, (uint8_t) inB1, (uint8_t) inB2, (uint8_t) inB3);
+
+    *((v4u*)out) = max4(inA4, inB4);
+
+    uint8_t inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+
+static void __attribute__((noinline)) pulp_nn_compare_and_replace_if_larger_i2(int8_t * base,
+                                                int8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask2 = 0x0c;
+  int8_t n_mask2 = ~ mask2;
+  int8_t mask4 = 0x30;
+  int8_t n_mask4 = ~ mask4;
+  int8_t mask6 = 0xc0;
+  int8_t n_mask6 = ~ mask6;
+  int8_t off2 = 2;
+  int8_t off4 = 4;
+  int8_t off6 = 6;
+
+  int8_t *pIn = base;
+  int8_t *pCom = target;
+  v4s inp[4];
+  v4s com[4];
+  int8_t *out;
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    pulp_nn_i2_to_i8(pIn, inp);
+    pulp_nn_i2_to_i8(pCom, com);
+
+    *((v4s*)out) = maxs4(inp[0], com[0]);
+
+    int8_t inA = (int8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((int8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    *((v4s*)out) = maxs4(inp[1], com[1]);
+
+    inA = (int8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((int8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    *((v4s*)out) = maxs4(inp[2], com[2]);
+
+    inA = (int8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((int8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    *((v4s*)out) = maxs4(inp[3], com[3]);
+
+    inA = (int8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((int8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    int8_t inA0 = (int8_t) bitext((unsigned int) *pIn, 2, 0);
+    int8_t inA1 = (int8_t) bitext((unsigned int) *pIn, 2, 2);
+    int8_t inA2 = (int8_t) bitext((unsigned int) *pIn, 2, 4);
+    int8_t inA3 = (int8_t) bitext((unsigned int) *pIn, 2, 6);
+    v4s inA4 = pack((int8_t) inA0, (int8_t) inA1, (int8_t) inA2, (int8_t) inA3);
+    int8_t inB0 = (int8_t) bitext((unsigned int) *pCom, 2, 0);
+    int8_t inB1 = (int8_t) bitext((unsigned int) *pCom, 2, 2);
+    int8_t inB2 = (int8_t) bitext((unsigned int) *pCom, 2, 4);
+    int8_t inB3 = (int8_t) bitext((unsigned int) *pCom, 2, 6);
+    v4s inB4 = pack((int8_t) inB0, (int8_t) inB1, (int8_t) inB2, (int8_t) inB3);
+
+    *((v4s*)out) = maxs4(inA4, inB4);
+
+    int8_t inA = (int8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((int8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_avg_and_replace_u2(uint8_t * base,
+                                  uint8_t * target,
+                                  uint16_t length)
+{
+  int8_t mask2 = 0x0c;
+  int8_t n_mask2 = ~ mask2;
+  int8_t mask4 = 0x30;
+  int8_t n_mask4 = ~ mask4;
+  int8_t mask6 = 0xc0;
+  int8_t n_mask6 = ~ mask6;
+  int8_t off2 = 2;
+  int8_t off4 = 4;
+  int8_t off6 = 6;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+
+  while (length>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 2, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 2, 2);
+    uint8_t inA2 = (uint8_t) bitextu((unsigned int) *pIn, 2, 4);
+    uint8_t inA3 = (uint8_t) bitextu((unsigned int) *pIn, 2, 6);
+
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 2, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 2, 2);
+    uint8_t inB2 = (uint8_t) bitextu((unsigned int) *pCom, 2, 4);
+    uint8_t inB3 = (uint8_t) bitextu((unsigned int) *pCom, 2, 6);
+
+    inA0 = ((inA0 + inB0) >> 1);
+    inA1 = ((inA1 + inB1) >> 1);
+    inA2 = ((inA2 + inB2) >> 1);
+    inA3 = ((inA3 + inB3) >> 1);
+
+    uint8_t inA = (uint8_t) bitins(inA0, n_mask2, inA1, mask2, off2);
+    inA = bitins(inA, n_mask4, inA2, mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, inA3, mask6, off6);
+
+    pIn++;
+    pCom++;
+    length--;
+  }
+}
+
+/*
+ * XpulpNN
+ */
+
+static void __attribute__((noinline)) xpulp_nn_zero_mem_u8(uint8_t * pBuffer, unsigned int size)
+{
+  int lfover = size &0x3;
+  for (int i=0; i<(size>>2); i++)
+  {
+    *((v4u *)pBuffer) = (v4u){0,0,0,0};
+    MemoryFence();
+    pBuffer+=4;
+  }
+  while(lfover)
+  {
+    *pBuffer++=0;
+    lfover--;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_zero_mem_u4(uint8_t * pBuffer, unsigned int size)
+{
+  int lfover = size &0x7;
+  for (int i=0; i<(size>>3); i++)
+  {
+    *((v4u *)pBuffer) = (v4u){0,0,0,0};
+    MemoryFence();
+    pBuffer+=4;
+  }
+  while(lfover)
+  {
+    *pBuffer++=0;
+    lfover-=2;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_zero_mem_u2(uint8_t * pBuffer, unsigned int size)
+{
+  int lfover = size &0xf;
+  for (int i=0; i<(size>>4); i++)
+  {
+    *((v4u *)pBuffer) = (v4u){0,0,0,0};
+    MemoryFence();
+    pBuffer+=4;
+  }
+  while(lfover)
+  {
+    *pBuffer++=0;
+    lfover-=4;
+  }
+}
+
+
+static void __attribute__((noinline)) xpulp_tnn_zero_mem_ternary(uint8_t * pBuffer, unsigned int size, unsigned int uns)
+{
+  uint8_t pad_val = 0xd9;
+  uint32_t pad_vec = 0xd9d9d9d9;
+  if (uns) {
+    // if we are using an unsigned kernel, we need to pad with -1 because the hardware will add a +1 to ALL values!
+    pad_val = 0xff;
+    pad_vec = 0xffffffff;
+  }
+  int lfover = size &0xf;
+  for (int i=0; i<(size>>4); i++)
+    {
+      *((v4u *)pBuffer) = (v4u)pad_vec;
+      MemoryFence();
+      pBuffer+=4;
+    }
+  while(lfover)
+    {
+      *pBuffer++=pad_val;
+      lfover-=4;
+    }
+}
+
+static void __attribute__((noinline)) xpulp_nn_compare_and_replace_if_larger_u4(uint8_t * base,
+                                                uint8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    *((uint32_t *)pIn) = max8(*((uint32_t *)pIn), *((uint32_t *)pCom));
+
+    pIn+=4;
+    pCom+=4;
+
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 4, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 4, 4);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 4, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 4, 4);
+
+    if(inA0<inB0)
+      inA0=inB0;
+
+    if(inA1<inB1)
+      inA1=inB1;
+
+    *((uint8_t*)pIn) = bitins(inA0, n_mask, inA1, mask, off);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_compare_and_replace_if_larger_i4(
+    int8_t * base, int8_t * target, uint16_t length) {
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~mask;
+  int8_t off = 0x04;
+
+  int8_t *pIn = base;
+  int8_t *pCom = target;
+
+  int cnt = length >> 2;
+
+  while (cnt > 0u) {
+    *((int32_t *)pIn) = maxs8(*((int32_t *)pIn), *((int32_t *)pCom));
+
+    pIn += 4;
+    pCom += 4;
+
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left > 0u) {
+    int8_t inA0 = (int8_t)bitext((int)*pIn, 4, 0);
+    int8_t inA1 = (int8_t)bitext((int)*pIn, 4, 4);
+    int8_t inB0 = (int8_t)bitext((int)*pCom, 4, 0);
+    int8_t inB1 = (int8_t)bitext((int)*pCom, 4, 4);
+
+    if (inA0 < inB0)
+      inA0 = inB0;
+
+    if (inA1 < inB1)
+      inA1 = inB1;
+
+    *((int8_t *)pIn) = bitins(inA0, n_mask, inA1, mask, off);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_avg_and_replace_u4(uint8_t * base,
+                                  uint8_t * target,
+                                  uint16_t length)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+
+  int cnt = length >> 2;
+
+  while (cnt > 0u)
+  {
+    *((uint32_t *)pIn) = avg8(*((uint32_t *)pIn), *((uint32_t *)pCom));
+
+    pIn+=4;
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+
+  while (left>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 4, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 4, 4);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 4, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 4, 4);
+
+    inA0 = ((inA0 + inB0) >> 1);
+    inA1 = ((inA1 + inB1) >> 1);
+
+    *((uint8_t*)pIn) = bitins(inA0, n_mask, inA1, mask, off);
+
+    pIn++;
+    pCom++;
+    length--;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_compare_and_replace_if_larger_u2(uint8_t * base,
+                                                uint8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask2 = 0x0c;
+  int8_t n_mask2 = ~ mask2;
+  int8_t mask4 = 0x30;
+  int8_t n_mask4 = ~ mask4;
+  int8_t mask6 = 0xc0;
+  int8_t n_mask6 = ~ mask6;
+  int8_t off2 = 2;
+  int8_t off4 = 4;
+  int8_t off6 = 6;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+  uint8_t *out;
+
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    *((uint32_t *)pIn) = max16(*((uint32_t *)pIn), *((uint32_t *)pCom));
+
+    pIn+=4;
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 2, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 2, 2);
+    uint8_t inA2 = (uint8_t) bitextu((unsigned int) *pIn, 2, 4);
+    uint8_t inA3 = (uint8_t) bitextu((unsigned int) *pIn, 2, 6);
+    v4u inA4 = pack((uint8_t) inA0, (uint8_t) inA1, (uint8_t) inA2, (uint8_t) inA3);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 2, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 2, 2);
+    uint8_t inB2 = (uint8_t) bitextu((unsigned int) *pCom, 2, 4);
+    uint8_t inB3 = (uint8_t) bitextu((unsigned int) *pCom, 2, 6);
+    v4u inB4 = pack((uint8_t) inB0, (uint8_t) inB1, (uint8_t) inB2, (uint8_t) inB3);
+
+    *((v4u*)out) = max4(inA4, inB4);
+
+    uint8_t inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_compare_and_replace_if_larger_i2(int8_t * base,
+                                                int8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask2 = 0x0c;
+  int8_t n_mask2 = ~ mask2;
+  int8_t mask4 = 0x30;
+  int8_t n_mask4 = ~ mask4;
+  int8_t mask6 = 0xc0;
+  int8_t n_mask6 = ~ mask6;
+  int8_t off2 = 2;
+  int8_t off4 = 4;
+  int8_t off6 = 6;
+
+  int8_t *pIn = base;
+  int8_t *pCom = target;
+  int8_t *out;
+
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    *((int32_t *)pIn) = maxs16(*((int32_t *)pIn), *((int32_t *)pCom));
+
+    pIn+=4;
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    int8_t inA0 = (int8_t) bitext((int) *pIn, 2, 0);
+    int8_t inA1 = (int8_t) bitext((int) *pIn, 2, 2);
+    int8_t inA2 = (int8_t) bitext((int) *pIn, 2, 4);
+    int8_t inA3 = (int8_t) bitext((int) *pIn, 2, 6);
+    v4s inA4 = pack((int8_t) inA0, (int8_t) inA1, (int8_t) inA2, (int8_t) inA3);
+    int8_t inB0 = (int8_t) bitext((int) *pCom, 2, 0);
+    int8_t inB1 = (int8_t) bitext((int) *pCom, 2, 2);
+    int8_t inB2 = (int8_t) bitext((int) *pCom, 2, 4);
+    int8_t inB3 = (int8_t) bitext((int) *pCom, 2, 6);
+    v4s inB4 = pack((int8_t) inB0, (int8_t) inB1, (int8_t) inB2, (int8_t) inB3);
+
+    *((v4s*)out) = maxs4(inA4, inB4);
+
+    int8_t inA = (int8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((int8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_tnn_compare_and_replace_if_larger_ternary(int8_t * base,
+                                                int8_t * target,
+                                                uint16_t length)
+{
+  uint8_t mask2 = 0x0c;
+  uint8_t n_mask2 = ~ mask2;
+  uint8_t mask4 = 0x30;
+  uint8_t n_mask4 = ~ mask4;
+  uint8_t mask6 = 0xc0;
+  uint8_t n_mask6 = ~ mask6;
+  uint8_t off2 = 2;
+  uint8_t off4 = 4;
+  uint8_t off6 = 6;
+
+  uint8_t *pIn = (uint8_t *) base;
+  uint8_t *pCom = (uint8_t *) target;
+  uint8_t *out;
+
+  int cnt = length >> 2;
+  uint32_t result;
+
+  while(cnt > 0u)
+  {
+    uint32_t in1 = *((uint32_t *)pIn);
+    uint32_t in2 = *((int32_t *)pCom);
+    result = maxs20(in1, in2);
+    *((uint32_t *)pIn) = result;
+
+    pIn+=4;
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  if (left>0u)
+  {
+    // do the vector max on the whole word - we won't use the leftover bytes
+    uint32_t in1 = *((uint32_t *)pIn);
+    uint32_t in2 = *((int32_t *)pCom);
+    result = maxs20(in1, in2);
+
+    // ...and copy back the relevant bytes of the result to pIn
+    for (int i=0; i<left; i++)
+      *((uint8_t*)(pIn + i)) = (uint8_t) (result >> (8*i));
+
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_avg_and_replace_u2(uint8_t * base,
+                                  uint8_t * target,
+                                  uint16_t length)
+{
+  int8_t mask2 = 0x0c;
+  int8_t n_mask2 = ~ mask2;
+  int8_t mask4 = 0x30;
+  int8_t n_mask4 = ~ mask4;
+  int8_t mask6 = 0xc0;
+  int8_t n_mask6 = ~ mask6;
+  int8_t off2 = 2;
+  int8_t off4 = 4;
+  int8_t off6 = 6;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+  uint8_t *out;
+
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    *((uint32_t *)pIn) = avg16(*((uint32_t *)pIn), *((uint32_t *)pCom));
+
+    pIn+=4;
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 2, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 2, 2);
+    uint8_t inA2 = (uint8_t) bitextu((unsigned int) *pIn, 2, 4);
+    uint8_t inA3 = (uint8_t) bitextu((unsigned int) *pIn, 2, 6);
+    v4u inA4 = pack((uint8_t) inA0, (uint8_t) inA1, (uint8_t) inA2, (uint8_t) inA3);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 2, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 2, 2);
+    uint8_t inB2 = (uint8_t) bitextu((unsigned int) *pCom, 2, 4);
+    uint8_t inB3 = (uint8_t) bitextu((unsigned int) *pCom, 2, 6);
+    v4u inB4 = pack((uint8_t) inB0, (uint8_t) inB1, (uint8_t) inB2, (uint8_t) inB3);
+
+    *((v4u*)out) = avg4(inA4, inB4);
+
+    uint8_t inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+#endif
diff --git a/rt_nn_tests/xpnnv2_conv/pulp_nn_utils_xpnn.h b/rt_nn_tests/xpnnv2_conv/pulp_nn_utils_xpnn.h
new file mode 100644
index 0000000..0c783ae
--- /dev/null
+++ b/rt_nn_tests/xpnnv2_conv/pulp_nn_utils_xpnn.h
@@ -0,0 +1,1937 @@
+/*
+ * pulp_nn_utils.h
+ * Nazareno   Bruschi  <nazareno.bruschi@unibo.it>
+ * Alessandro Nadalini <alessandro.nadalini3@unibo.it>
+ * Georg Rutishauser   <georgr@iis.ee.ethz.ch>
+ *
+ * Copyright (C) 2019-2020 ETH Zurich & University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef __PULPNN_UTILS__
+#define __PULPNN_UTILS__
+
+#include "pmsis.h"
+#ifdef GAP_SDK
+#include "pulp.h"
+#endif
+
+#define bitext(x,size,off)                                      __builtin_pulp_bextract(x,size,off)
+#define bitextu(x,size,off)                                     __builtin_pulp_bextractu(x,size,off)
+#define bitins(dst,not_mask_imm,src,mask_imm,off)               __builtin_pulp_binsert(dst,not_mask_imm,src,mask_imm,off)
+#define pack(x,y,z,t)                                           __builtin_pulp_pack4(x,y,z,t)
+#define max4(a,b)                                               __builtin_pulp_maxu4(a,b)
+#define maxs4(a, b)                                             __builtin_pulp_max4(a, b)
+#define max8(a, b)                                              __builtin_pulp_maxu8(a, b)
+#define maxs8(a, b)                                             __builtin_pulp_max8(a, b)
+#define max16(a, b)                                             __builtin_pulp_maxu16(a, b)
+#define maxs16(a, b)                                            __builtin_pulp_max16(a, b)
+#define max32(a,b)                                              __builtin_pulp_maxusi(a,b)
+#define maxs32(a,b)                                             __builtin_pulp_maxsi(a,b)
+#define min32(a,b)                                              __builtin_pulp_minusi(a,b)
+#define mins32(a,b)                                             __builtin_pulp_minsi(a,b)
+#define min4(a, b)                                              __builtin_pulp_minu4(a, b)
+#define mins4(a, b)                                             __builtin_pulp_min4(a, b)
+#define min8(a, b)                                              __builtin_pulp_minu8(a, b)
+#define mins8(a, b)                                             __builtin_pulp_min8(a, b)
+#define min16(a, b)                                             __builtin_pulp_minu16(a, b)
+#define mins16(a, b)                                            __builtin_pulp_min16(a, b)
+#define avg4(a,b)                                               __builtin_pulp_avgu4(a,b)
+#define avg8(a,b)                                               __builtin_pulp_avgu8(a,b)
+#define avg16(a,b)                                              __builtin_pulp_avgu16(a,b)
+#define log2(x)                                                 __builtin_pulp_fl1(x)
+#define min(a,b)                                                ((a)<(b)?(a):(b))
+#define SumDotp4(a, b, c)                                       __builtin_pulp_sdotusp4(a, b, c)
+#define SumDotp8(a, b, c)                                       __builtin_pulp_sdotusp8(a, b, c)
+#define SumDotp16(a, b, c)                                      __builtin_pulp_sdotusp16(a, b, c)
+#define SumDotps4(a, b, c)                                      __builtin_pulp_sdotsp4(a, b, c)
+#define SumDotps8(a, b, c)                                      __builtin_pulp_sdotsp8(a, b, c)
+#define SumDotps16(a, b, c)                                     __builtin_pulp_sdotsp16(a, b, c)
+#define clip4(x)                                                __builtin_pulp_clipu_r(x, 15)
+#define clip2(x)                                                __builtin_pulp_clipu_r(x, 3)
+#define clip8(x)                                                __builtin_pulp_clipu_r(x, 255)
+
+#define clips4(x)                                               __builtin_pulp_clip_r(x, 7)
+#define clips2(x)                                               __builtin_pulp_clip_r(x, 1)
+#define clips8(x)                                               __builtin_pulp_clip_r(x, 127)
+#define MacLoadInit(a_update, b_update, a_reg, b_reg, ptr)      __builtin_pulp_mlinitspr_v3(a_update, b_update, a_reg, b_reg, ptr)
+#define MacLoadUpdate(ptr)                                      __builtin_pulp_mlupdatespr_v3(ptr)
+#define MacLoadAssign(ptr)                                      __builtin_pulp_mlassignspr_v3(ptr)
+#define MacLoad4(a_update, b_update, a_reg, b_reg, ptr, sum)    __builtin_pulp_mlsdotsup4_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+#define MacLoad8(a_update, b_update, a_reg, b_reg, ptr, sum)    __builtin_pulp_mlsdotsup8_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+#define MacLoad16(a_update, b_update, a_reg, b_reg, ptr, sum)   __builtin_pulp_mlsdotsup16_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+#define MacLoads4(a_update, b_update, a_reg, b_reg, ptr, sum)   __builtin_pulp_mlsdotsp4_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+#define MacLoads8(a_update, b_update, a_reg, b_reg, ptr, sum)   __builtin_pulp_mlsdotsp8_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+#define MacLoads16(a_update, b_update, a_reg, b_reg, ptr, sum)  __builtin_pulp_mlsdotsp16_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+
+#define PACK_INT8_SIZE(x)                                    (x)
+#define PACK_INT4_SIZE(x)                                    ((x) >> 1)
+#define PACK_INT2_SIZE(x)                                    ((x) >> 2)
+
+#define MemoryFence()                                        asm volatile("":::"memory")
+
+#define LEGACY_MODE(x)                                       asm volatile ("csrwi 0x010," x)
+#define IVEC_FMT(x)                                          asm volatile ("csrwi 0x00D," x)
+#define MIXED_SKIP(x)                                        asm volatile ("csrwi 0x00F," x)
+#define A_ADDRESS(x)                                         asm volatile ("csrw 0x100, %0" :: "r" (x))
+#define W_ADDRESS(x)                                         asm volatile ("csrw 0x101, %0" :: "r" (x))
+#define A_STRIDE(x)                                          asm volatile ("csrw 0x102, %0":: "r" (x))
+#define W_STRIDE(x)                                          asm volatile ("csrw 0x103, %0":: "r" (x))
+#define A_ROLLBACK(x)                                        asm volatile ("csrw 0x104, %0":: "r" (x))
+#define W_ROLLBACK(x)                                        asm volatile ("csrw 0x105, %0":: "r" (x))
+#define A_SKIP(x)                                            asm volatile ("csrwi 0x106," x)
+#define W_SKIP(x)                                            asm volatile ("csrwi 0x107," x)
+
+static uint8_t __attribute__((noinline)) pulp_nn_quant_u2 (
+  int32_t phi,
+  int16_t m,
+  int8_t d
+  ) {
+  int32_t x = (m * phi) >> d;
+  uint8_t res = clip2(x);
+  return res;
+}
+static uint8_t __attribute__((noinline)) pulp_nn_bn_quant_u2 (
+  int32_t phi,
+  int32_t k,
+  int32_t lambda,
+  int8_t d
+  ) {
+  int32_t integer_image_phi = (k * phi) + lambda;
+  int32_t x = (integer_image_phi) >> d;
+  uint8_t res = clip2(x);
+  return res;
+}
+static uint8_t __attribute__((noinline)) pulp_nn_add_quant_u2 (
+  uint8_t pix1,
+  uint8_t pix2,
+  int16_t m1,
+  int16_t m2,
+  int8_t d
+  ) {
+  uint32_t integer_image = pix1*m1 + pix2*m2;
+  uint32_t x = (integer_image) >> d;
+  uint8_t res = clip2(x);
+  return res;
+}
+
+static int8_t __attribute__((noinline)) pulp_nn_quant_i2 (
+  int32_t phi,
+  int16_t m,
+  int8_t d
+  ) {
+  int32_t x = (m * phi) >> d;
+  int8_t res = clips2(x);
+  return res;
+}
+static int8_t __attribute__((noinline)) pulp_nn_bn_quant_i2 (
+  int32_t phi,
+  int32_t k,
+  int32_t lambda,
+  int8_t d
+  ) {
+  int32_t integer_image_phi = (k * phi) + lambda;
+  int32_t x = (integer_image_phi) >> d;
+  int8_t res = clips2(x);
+  return res;
+}
+static int8_t __attribute__((noinline)) pulp_nn_add_quant_i2 (
+  int8_t pix1,
+  int8_t pix2,
+  int16_t m1,
+  int16_t m2,
+  int8_t d
+  ) {
+  int32_t integer_image = pix1*m1 + pix2*m2;
+  int32_t x = (integer_image) >> d;
+  int8_t res = clips2(x);
+  return res;
+}
+
+static uint8_t __attribute__((noinline)) pulp_nn_quant_u4 (
+  int32_t phi,
+  int16_t m,
+  int8_t d
+  ) {
+  int32_t x = (m * phi) >> d;
+  uint8_t res = clip4(x);
+  return res;
+}
+static uint8_t __attribute__((noinline)) pulp_nn_bn_quant_u4 (
+  int32_t phi,
+  int32_t k,
+  int32_t lambda,
+  int8_t d
+  ) {
+  int32_t integer_image_phi = (k * phi) + lambda;
+  int32_t x = (integer_image_phi) >> d;
+  uint8_t res = clip4(x);
+  return res;
+}
+static uint8_t __attribute__((noinline)) pulp_nn_add_quant_u4 (
+  uint8_t pix1,
+  uint8_t pix2,
+  int16_t m1,
+  int16_t m2,
+  int8_t d
+  ) {
+  uint32_t integer_image = pix1*m1 + pix2*m2;
+  uint32_t x = (integer_image) >> d;
+  uint8_t res = clip4(x);
+  return res;
+}
+
+static int8_t __attribute__((noinline)) pulp_nn_quant_i4 (
+  int32_t phi,
+  int16_t m,
+  int8_t d
+  ) {
+  int32_t x = (m * phi) >> d;
+  int8_t res = clips4(x);
+  return res;
+}
+static int8_t __attribute__((noinline)) pulp_nn_bn_quant_i4 (
+  int32_t phi,
+  int32_t k,
+  int32_t lambda,
+  int8_t d
+  ) {
+  int32_t integer_image_phi = (k * phi) + lambda;
+  int32_t x = (integer_image_phi) >> d;
+  int8_t res = clips4(x);
+  return res;
+}
+static int8_t __attribute__((noinline)) pulp_nn_add_quant_i4 (
+  int8_t pix1,
+  int8_t pix2,
+  int16_t m1,
+  int16_t m2,
+  int8_t d
+  ) {
+  int32_t integer_image = pix1*m1 + pix2*m2;
+  int32_t x = (integer_image) >> d;
+  int8_t res = clips4(x);
+  return res;
+}
+
+static uint8_t __attribute__((noinline)) pulp_nn_quant_u8 (
+  int32_t phi,
+  int16_t m,
+  int8_t d
+  ) {
+  int32_t x = (m * phi) >> d;
+  uint8_t res = clip8(x);
+  return res;
+}
+static uint8_t __attribute__((noinline)) pulp_nn_bn_quant_u8 (
+  int32_t phi,
+  int32_t k,
+  int32_t lambda,
+  int8_t d
+  ) {
+  int32_t integer_image_phi = (k * phi) + lambda;
+  int32_t x = (integer_image_phi) >> d;
+  uint8_t res = clip8(x);
+  return res;
+}
+static uint8_t __attribute__((noinline)) pulp_nn_add_quant_u8 (
+  uint8_t pix1,
+  uint8_t pix2,
+  int16_t m1,
+  int16_t m2,
+  int8_t d
+  ) {
+  uint32_t integer_image = pix1*m1 + pix2*m2;
+  uint32_t x = (integer_image) >> d;
+  uint8_t res = clip8(x);
+  return res;
+}
+
+static int8_t __attribute__((noinline)) pulp_nn_quant_i8 (
+  int32_t phi,
+  int16_t m,
+  int8_t d
+  ) {
+  int32_t x = (m * phi) >> d;
+  int8_t res = clips8(x);
+  return res;
+}
+static int8_t __attribute__((noinline)) pulp_nn_bn_quant_i8 (
+  int32_t phi,
+  int32_t k,
+  int32_t lambda,
+  int8_t d
+  ) {
+  int32_t integer_image_phi = (k * phi) + lambda;
+  int32_t x = (integer_image_phi) >> d;
+  int8_t res = clips8(x);
+  return res;
+}
+static int8_t __attribute__((noinline)) pulp_nn_add_quant_i8 (
+  int8_t pix1,
+  int8_t pix2,
+  int16_t m1,
+  int16_t m2,
+  int8_t d
+  ) {
+  int32_t integer_image = pix1*m1 + pix2*m2;
+  int32_t x = (integer_image) >> d;
+  int8_t res = clips8(x);
+  return res;
+}
+
+
+static uint8_t __attribute__((noinline)) pulp_nn_u4_quant(int input, int16_t * pThr)
+{
+  if(input <= pThr[7] )
+  {
+    if( input <= pThr[3])
+    {
+      if( input <= pThr[1])
+      {
+        if( input <= pThr[0])
+          return 0;
+        else
+          return 1;
+      }
+      else
+      {
+        if( input <= pThr[2])
+          return 2;
+        else
+          return 3;
+      }
+    }
+    else
+    {
+      if( input <= pThr[5])
+      {
+        if( input <= pThr[4])
+          return 4;
+        else
+          return 5;
+      }
+      else
+      {
+        if( input <= pThr[6])
+          return 6;
+        else
+          return 7;
+      }
+    }
+  }
+  else
+  {
+    if( input <= pThr[11])
+    {
+      if( input <= pThr[9])
+      {
+        if( input <= pThr[8])
+          return 8;
+        else
+          return 9;
+      }
+      else
+      {
+        if( input <= pThr[10])
+          return 10;
+        else
+          return 11;
+      }
+    }
+    else
+    {
+      if( input <= pThr[13])
+      {
+        if( input <= pThr[12])
+          return 12;
+        else
+          return 13;
+      }
+      else
+      {
+        if( input <= pThr[14])
+          return 14;
+        else
+          return 15;
+      }
+    }
+  }
+}
+
+static uint8_t __attribute__((noinline)) pulp_nn_u2_quant(int input, int16_t * pThr)
+{
+  if( input <= pThr[1])
+  {
+    if( input <= pThr[0])
+        {
+      return 0;
+    }
+        else
+        {
+      return 1;
+    }
+  }
+    else
+    {
+    if( input <= pThr[2])
+        {
+      return 2;
+    }
+        else
+        {
+      return 3;
+    }
+  }
+}
+
+/*
+ * Common
+ */
+
+
+static v4s __attribute__((noinline)) pulp_nn_i4_to_i8_r( int8_t *pSrc)
+{
+  v4s Src = *((v4s*) pSrc);
+  int8_t bext1, bext2, bext3, bext4;
+
+  bext1 = (int8_t) bitext((int) Src, 4, 0);
+  bext2 = (int8_t) bitext((int) Src, 4, 4);
+  bext3 = (int8_t) bitext((int) Src, 4, 8);
+  bext4 = (int8_t) bitext((int) Src, 4, 12);
+  v4s res = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+
+  return res;
+}
+
+static v4u __attribute__((noinline)) pulp_nn_u4_to_u8_r(uint8_t *pSrc)
+{
+  v4u Src = *((v4u*) pSrc);
+  uint8_t bext1, bext2, bext3, bext4;
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 4, 0);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 4, 4);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 4, 8);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 4, 12);
+  v4u res = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+
+  return res;
+}
+
+static v4s __attribute__((noinline)) pulp_nn_i2_to_i8_r( int8_t *pSrc)
+{
+  v4s Src = *((v4s*) pSrc);
+  int8_t bext1, bext2, bext3, bext4;
+
+  bext1 = (int8_t) bitext((int) Src, 2, 0);
+  bext2 = (int8_t) bitext((int) Src, 2, 2);
+  bext3 = (int8_t) bitext((int) Src, 2, 4);
+  bext4 = (int8_t) bitext((int) Src, 2, 6);
+  v4s res = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+
+  return res;
+}
+
+static v4u __attribute__((noinline)) pulp_nn_u2_to_u8_r(uint8_t *pSrc)
+{
+  v4u Src = *((v4u*) pSrc);
+  uint8_t bext1, bext2, bext3, bext4;
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 0);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 2);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 4);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 6);
+  v4u res = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+
+  return res;
+}
+
+static v4s __attribute__((noinline)) pulp_nn_i2_to_i4_r(int8_t *pSrc)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  v4s Src = *((v4s*) pSrc);
+  int8_t bext1, bext2, bext3, bext4;
+  int8_t out1, out2, out3, out4;
+
+  bext1 = (int8_t) bitextu((int) Src, 2, 0);
+  bext2 = (int8_t) bitextu((int) Src, 2, 2);
+  bext3 = (int8_t) bitextu((int) Src, 2, 4);
+  bext4 = (int8_t) bitextu((int) Src, 2, 6);
+
+  out1 = bitins(bext1, n_mask, bext2, mask, off);
+  out2 = bitins(bext3, n_mask, bext4, mask, off);
+
+  bext1 = (int8_t) bitextu((int) Src, 2, 8);
+  bext2 = (int8_t) bitextu((int) Src, 2, 10);
+  bext3 = (int8_t) bitextu((int) Src, 2, 12);
+  bext4 = (int8_t) bitextu((int) Src, 2, 14);
+
+  out3 = bitins(bext1, n_mask, bext2, mask, off);
+  out4 = bitins(bext3, n_mask, bext4, mask, off);
+
+  v4s res = pack((int8_t) out1, (int8_t) out2, (int8_t) out3, (int8_t) out4);
+
+  return res;
+}
+
+static v4u __attribute__((noinline)) pulp_nn_u2_to_u4_r(uint8_t *pSrc)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  v4u Src = *((v4u*) pSrc);
+  uint8_t bext1, bext2, bext3, bext4;
+  uint8_t out1, out2, out3, out4;
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 0);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 2);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 4);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 6);
+
+  out1 = bitins(bext1, n_mask, bext2, mask, off);
+  out2 = bitins(bext3, n_mask, bext4, mask, off);
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 8);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 10);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 12);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 14);
+
+  out3 = bitins(bext1, n_mask, bext2, mask, off);
+  out4 = bitins(bext3, n_mask, bext4, mask, off);
+
+  v4u res = pack((uint8_t) out1, (uint8_t) out2, (uint8_t) out3, (uint8_t) out4);
+
+  return res;
+}
+
+static int8_t *__attribute__((always_inline)) pulp_nn_i4_to_i8( int8_t *pSrc, int8_t *pDst)
+{
+  v4s Src = *((v4s*) pSrc);
+  int8_t bext1, bext2, bext3, bext4;
+
+  pSrc+=4;
+
+  bext1 = (int8_t) bitext((int) Src, 4, 0);
+  bext2 = (int8_t) bitext((int) Src, 4, 4);
+  bext3 = (int8_t) bitext((int) Src, 4, 8);
+  bext4 = (int8_t) bitext((int) Src, 4, 12);
+  *((v4s*)pDst) = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+   MemoryFence();
+  pDst+=4;
+  bext1 = (int8_t) bitext((int) Src, 4, 16);
+  bext2 = (int8_t) bitext((int) Src, 4, 20);
+  bext3 = (int8_t) bitext((int) Src, 4, 24);
+  bext4 = (int8_t) bitext((int) Src, 4, 28);
+  *((v4s*)pDst) = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+
+  return pSrc;
+}
+
+static uint8_t *__attribute__((always_inline)) pulp_nn_u4_to_u8(uint8_t *pSrc, uint8_t *pDst)
+{
+  v4u Src = *((v4u*) pSrc);
+  uint8_t bext1, bext2, bext3, bext4;
+
+  pSrc+=4;
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 4, 0);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 4, 4);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 4, 8);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 4, 12);
+  *((v4u*)pDst) = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 4, 16);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 4, 20);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 4, 24);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 4, 28);
+  *((v4u*)pDst) = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+
+  return pSrc;
+}
+
+static int8_t *__attribute__((always_inline)) pulp_nn_i2_to_i8( int8_t * pSrc, int8_t * pDst)
+{
+  v4s Src = *((v4s*) pSrc);
+  int8_t bext1, bext2, bext3, bext4;
+
+  pSrc+=4;
+
+  bext1 = (int8_t) bitext((int) Src, 2, 0);
+  bext2 = (int8_t) bitext((int) Src, 2, 2);
+  bext3 = (int8_t) bitext((int) Src, 2, 4);
+  bext4 = (int8_t) bitext((int) Src, 2, 6);
+  *((v4s*)pDst) = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (int8_t) bitext((int) Src, 2, 8);
+  bext2 = (int8_t) bitext((int) Src, 2, 10);
+  bext3 = (int8_t) bitext((int) Src, 2, 12);
+  bext4 = (int8_t) bitext((int) Src, 2, 14);
+  *((v4s*)pDst) = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (int8_t) bitext((int) Src, 2, 16);
+  bext2 = (int8_t) bitext((int) Src, 2, 18);
+  bext3 = (int8_t) bitext((int) Src, 2, 20);
+  bext4 = (int8_t) bitext((int) Src, 2, 22);
+  *((v4s*)pDst) = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (int8_t) bitext((int) Src, 2, 24);
+  bext2 = (int8_t) bitext((int) Src, 2, 26);
+  bext3 = (int8_t) bitext((int) Src, 2, 28);
+  bext4 = (int8_t) bitext((int) Src, 2, 30);
+  *((v4s*)pDst) = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+
+  return pSrc;
+}
+
+static uint8_t *__attribute__((always_inline)) pulp_nn_u2_to_u8(uint8_t * pSrc, uint8_t * pDst)
+{
+  v4u Src = *((v4u*) pSrc);
+  uint8_t bext1, bext2, bext3, bext4;
+
+  pSrc+=4;
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 0);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 2);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 4);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 6);
+  *((v4u*)pDst) = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 8);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 10);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 12);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 14);
+  *((v4u*)pDst) = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 16);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 18);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 20);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 22);
+  *((v4u*)pDst) = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 24);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 26);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 28);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 30);
+  *((v4u*)pDst) = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+
+  return pSrc;
+}
+
+static int8_t *__attribute__((always_inline)) pulp_nn_i2_to_i4( int8_t * pSrc, int8_t * pDst)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  v4s Src = *((v4s*) pSrc);
+  int8_t bext1, bext2, bext3, bext4;
+  int8_t out1, out2, out3, out4;
+
+  pSrc+=4;
+
+  bext1 = (int8_t) bitext((int) Src, 2, 0);
+  bext2 = (int8_t) bitext((int) Src, 2, 2);
+  bext3 = (int8_t) bitext((int) Src, 2, 4);
+  bext4 = (int8_t) bitext((int) Src, 2, 6);
+
+  out1 = bitins(bext1, n_mask, bext2, mask, off);
+  out2 = bitins(bext3, n_mask, bext4, mask, off);
+  MemoryFence();
+
+  bext1 = (int8_t) bitext((int) Src, 2, 8);
+  bext2 = (int8_t) bitext((int) Src, 2, 10);
+  bext3 = (int8_t) bitext((int) Src, 2, 12);
+  bext4 = (int8_t) bitext((int) Src, 2, 14);
+
+  out3 = bitins(bext1, n_mask, bext2, mask, off);
+  out4 = bitins(bext3, n_mask, bext4, mask, off);
+
+  *((v4s*)pDst) = pack((int8_t) out1, (int8_t) out2, (int8_t) out3, (int8_t) out4);
+  MemoryFence();
+
+  pDst+=4;
+  bext1 = (int8_t) bitext((int) Src, 2, 16);
+  bext2 = (int8_t) bitext((int) Src, 2, 18);
+  bext3 = (int8_t) bitext((int) Src, 2, 20);
+  bext4 = (int8_t) bitext((int) Src, 2, 22);
+
+  out1 = bitins(bext1, n_mask, bext2, mask, off);
+  out2 = bitins(bext3, n_mask, bext4, mask, off);
+  MemoryFence();
+
+  bext1 = (int8_t) bitext((int) Src, 2, 24);
+  bext2 = (int8_t) bitext((int) Src, 2, 26);
+  bext3 = (int8_t) bitext((int) Src, 2, 28);
+  bext4 = (int8_t) bitext((int) Src, 2, 30);
+
+  out3 = bitins(bext1, n_mask, bext2, mask, off);
+  out4 = bitins(bext3, n_mask, bext4, mask, off);
+
+  *((v4s*)pDst) = pack((int8_t) out1, (int8_t) out2, (int8_t) out3, (int8_t) out4);
+
+  return pSrc;
+}
+
+static uint8_t *__attribute__((always_inline)) pulp_nn_u2_to_u4( uint8_t * pSrc, uint8_t * pDst)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  v4u Src = *((v4u*) pSrc);
+  uint8_t bext1, bext2, bext3, bext4;
+  uint8_t out1, out2, out3, out4;
+
+  pSrc+=4;
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 0);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 2);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 4);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 6);
+
+  out1 = bitins(bext1, n_mask, bext2, mask, off);
+  out2 = bitins(bext3, n_mask, bext4, mask, off);
+  MemoryFence();
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 8);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 10);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 12);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 14);
+
+  out3 = bitins(bext1, n_mask, bext2, mask, off);
+  out4 = bitins(bext3, n_mask, bext4, mask, off);
+
+  *((v4u*)pDst) = pack((uint8_t) out1, (uint8_t) out2, (uint8_t) out3, (uint8_t) out4);
+  MemoryFence();
+
+  pDst+=4;
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 16);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 18);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 20);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 22);
+
+  out1 = bitins(bext1, n_mask, bext2, mask, off);
+  out2 = bitins(bext3, n_mask, bext4, mask, off);
+  MemoryFence();
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 24);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 26);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 28);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 30);
+
+  out3 = bitins(bext1, n_mask, bext2, mask, off);
+  out4 = bitins(bext3, n_mask, bext4, mask, off);
+  
+  *((v4u*)pDst) = pack((uint8_t) out1, (uint8_t) out2, (uint8_t) out3, (uint8_t) out4);
+
+  return pSrc;
+}
+
+/*
+ * XpulpV2
+ */
+
+static void __attribute__((noinline)) pulp_zero_mem(uint8_t * pBuffer, unsigned int size)
+{
+  int lfover = size &0x3;
+  for (int i=0; i<(size>>2); i++)
+  {
+    *((v4u *)pBuffer) = (v4u){0,0,0,0};
+    MemoryFence();
+    pBuffer+=4;
+  }
+  while(lfover)
+  {
+    *pBuffer++=0;
+    lfover--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_im2col_u2_to_u8(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_u2_to_u8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 16;
+  }
+  while (lfover)
+  {
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 0);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 2);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 4);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 6);
+    pOutput++;
+    pInput++;
+    lfover -= 4;
+  }
+}
+static void __attribute__((noinline)) pulp_nn_im2col_i2_to_i8(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_i2_to_i8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 16;
+  }
+  while (lfover)
+  {
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 0);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 2);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 4);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 6);
+    pOutput++;
+    pInput++;
+    lfover -= 4;
+  }
+}
+static void __attribute__((noinline)) pulp_nn_im2col_u4_to_u8(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 3u;
+  int lfover = blockSize & 0x07;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_u4_to_u8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 8;
+  }
+  while (lfover)
+  {
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 4, 0);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 4, 4);
+    pOutput++;
+    pInput++;
+    lfover -= 2;
+  }
+}
+static void __attribute__((noinline)) pulp_nn_im2col_i4_to_i8(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 3u;
+  int lfover = blockSize & 0x07;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_i4_to_i8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 8;
+  }
+  while (lfover)
+  {
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 4, 0);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 4, 4);
+    pOutput++;
+    pInput++;
+    lfover -= 2;
+  }
+}
+static void __attribute__((noinline)) pulp_nn_im2col_u8_to_u8(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 2u;
+  int lfover = blockSize & 0x03;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4u*)pOutput) = *((v4u*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    *((uint8_t*)pOutput) = *((uint8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover--;
+  }
+}
+static void __attribute__((noinline)) pulp_nn_im2col_i8_to_i8(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 2u;
+  int lfover = blockSize & 0x03;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4s*)pOutput) = *((v4s*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    *((int8_t*)pOutput) = *((int8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover--;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_im2col_u2_to_u2(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4u*)pOutput) = *((v4u*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    uint8_t extr;
+    *((uint8_t*)pOutput) = *((uint8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover -= 4;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_i2_to_i2(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4s*)pOutput) = *((v4s*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    int8_t extr;
+    *((int8_t*)pOutput) = *((int8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover -= 4;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_u2_to_u4(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_u2_to_u4(pInput, pOutput);
+    MemoryFence();
+    pOutput += 8;
+  }
+  while (lfover)
+  {
+    uint8_t extr;
+    // TODO: this is potentially dangerous/wrong if the number of channels
+    // is not a multiple of 8!
+    *((v4u *)pOutput) = pulp_nn_u2_to_u4_r(pInput);
+    pInput += 2;
+    pOutput += 4;
+    lfover -= 8;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_i2_to_i4(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_i2_to_i4(pInput, pOutput);
+    MemoryFence();
+    pOutput += 8;
+  }
+  while (lfover)
+  {
+    int8_t extr;
+    // TODO: this is potentially dangerous/wrong if the number of channels
+    // is not a multiple of 8!
+    *((v4s *)pOutput) = pulp_nn_i2_to_i4_r(pInput);
+    pInput += 2;
+    pOutput += 4;
+    lfover -= 8;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_u2_to_u8(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_u2_to_u8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 16;
+  }
+  while (lfover)
+  {
+    uint8_t extr;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 0);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 2);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 4);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 6);
+    pOutput++;
+    pInput++;
+    lfover -= 4;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_i2_to_i8(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_i2_to_i8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 16;
+  }
+  while (lfover)
+  {
+    int8_t extr;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 0);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 2);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 4);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 6);
+    pOutput++;
+    pInput++;
+    lfover -= 4;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_u4_to_u4(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 3u;
+  int lfover = blockSize & 0x07;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4u*)pOutput) = *((v4u*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    uint8_t extr;
+    *((uint8_t*)pOutput) = *((uint8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover -= 2;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_i4_to_i4(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 3u;
+  int lfover = blockSize & 0x07;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4s*)pOutput) = *((v4s*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    int8_t extr;
+    *((int8_t*)pOutput) = *((int8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover -= 2;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_u4_to_u8(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 3u;
+  int lfover = blockSize & 0x07;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_u4_to_u8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 8;
+  }
+  while (lfover)
+  {
+    uint8_t extr;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 4, 0);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 4, 4);
+    pOutput++;
+    pInput++;
+    lfover -= 2;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_i4_to_i8(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 3u;
+  int lfover = blockSize & 0x07;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_i4_to_i8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 8;
+  }
+  while (lfover)
+  {
+    int8_t extr;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 4, 0);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 4, 4);
+    pOutput++;
+    pInput++;
+    lfover -= 2;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_u8_to_u8(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 2u;
+  int lfover = blockSize & 0x03;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4u*)pOutput) = *((v4u*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    uint8_t extr;
+    *((uint8_t*)pOutput) = *((uint8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover -= 1;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_i8_to_i8(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 2u;
+  int lfover = blockSize & 0x03;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4s*)pOutput) = *((v4s*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    int8_t extr;
+    *((int8_t*)pOutput) = *((int8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover -= 1;
+  }
+}
+
+
+
+static void __attribute__((noinline)) pulp_nn_compare_and_replace_if_larger_u8(uint8_t * base,
+                                                uint8_t * target,
+                                                uint16_t length)
+{
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+  v4u inp;
+  v4u com;
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    inp = *((v4u*)pIn);
+    com = *((v4u*)pCom);
+
+    *((v4u*)pIn) = max4(inp, com);
+
+    pCom+=4;
+    pIn+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    if(*pIn<*pCom)
+      *pIn=*pCom;
+    
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_compare_and_replace_if_larger_i8(
+    int8_t * base, int8_t * target, uint16_t length) {
+  int8_t *pIn = base;
+  int8_t *pCom = target;
+  v4s inp;
+  v4s com;
+  int cnt = length >> 2;
+
+  while (cnt > 0u) {
+    inp = *((v4s *)pIn);
+    com = *((v4s *)pCom);
+
+    *((v4s *)pIn) = maxs4(inp, com);
+
+    pCom += 4;
+    pIn += 4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left > 0u) {
+    if (*pIn < *pCom)
+      *pIn = *pCom;
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_avg_and_replace_u8(uint8_t * base,
+                                  uint8_t * target,
+                                  uint16_t length)
+{
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+
+  while (length>0u)
+  {
+    *pIn = ((*pIn + *pCom) >> 1);
+
+    pIn++;
+    pCom++;
+    length--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_compare_and_replace_if_larger_u4(uint8_t * base,
+                                                uint8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+  v4u inp[2];
+  v4u com[2];
+  uint8_t *out;
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    pulp_nn_u4_to_u8(pIn, (uint8_t *)inp);
+    pulp_nn_u4_to_u8(pCom, (uint8_t *)com);
+
+    *((v4u *)out) = max4(inp[0], com[0]);
+
+    *((uint8_t*)pIn) = bitins(*out, n_mask, *(out + 1), mask, off);
+    pIn++;
+    *((uint8_t*)pIn) = bitins(*(out + 2), n_mask, *(out + 3), mask, off);
+    pIn++;
+
+    *((v4u *)out) = max4(inp[1], com[1]);
+
+    *((uint8_t*)pIn) = bitins(*out, n_mask, *(out + 1), mask, off);
+    pIn++;
+    *((uint8_t*)pIn) = bitins(*(out + 2), n_mask, *(out + 3), mask, off);
+    pIn++;
+
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 4, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 4, 4);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 4, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 4, 4);
+
+    if(inA0<inB0)
+      inA0=inB0;
+
+    if(inA1<inB1)
+      inA1=inB1;
+
+    *((uint8_t*)pIn) = bitins(inA0, n_mask, inA1, mask, off);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_compare_and_replace_if_larger_i4(int8_t * base,
+                                                int8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  int8_t *pIn = base;
+  int8_t *pCom = target;
+  v4s inp[2];
+  v4s com[2];
+  int8_t *out;
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    pulp_nn_i4_to_i8(pIn, (int8_t *)inp);
+    pulp_nn_i4_to_i8(pCom, (int8_t *)com);
+
+    *((v4s *)out) = maxs4(inp[0], com[0]);
+
+    *((int8_t*)pIn) = bitins(*out, n_mask, *(out + 1), mask, off);
+    pIn++;
+    *((int8_t*)pIn) = bitins(*(out + 2), n_mask, *(out + 3), mask, off);
+    pIn++;
+
+    *((v4s *)out) = maxs4(inp[1], com[1]);
+
+    *((int8_t*)pIn) = bitins(*out, n_mask, *(out + 1), mask, off);
+    pIn++;
+    *((int8_t*)pIn) = bitins(*(out + 2), n_mask, *(out + 3), mask, off);
+    pIn++;
+
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    int8_t inA0 = (int8_t) bitext((int) *pIn, 4, 0);
+    int8_t inA1 = (int8_t) bitext((int) *pIn, 4, 4);
+    int8_t inB0 = (int8_t) bitext((int) *pCom, 4, 0);
+    int8_t inB1 = (int8_t) bitext((int) *pCom, 4, 4);
+
+    if(inA0<inB0)
+      inA0=inB0;
+
+    if(inA1<inB1)
+      inA1=inB1;
+
+    *((int8_t*)pIn) = bitins(inA0, n_mask, inA1, mask, off);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_avg_and_replace_u4(uint8_t * base,
+                                  uint8_t * target,
+                                  uint16_t length)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+
+  while (length>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 4, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 4, 4);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 4, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 4, 4);
+
+    inA0 = ((inA0 + inB0) >> 1);
+    inA1 = ((inA1 + inB1) >> 1);
+
+    *((uint8_t*)pIn) = bitins(inA0, n_mask, inA1, mask, off);
+
+    pIn++;
+    pCom++;
+    length--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_compare_and_replace_if_larger_u2(uint8_t * base,
+                                                uint8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask2 = 0x0c;
+  int8_t n_mask2 = ~ mask2;
+  int8_t mask4 = 0x30;
+  int8_t n_mask4 = ~ mask4;
+  int8_t mask6 = 0xc0;
+  int8_t n_mask6 = ~ mask6;
+  int8_t off2 = 2;
+  int8_t off4 = 4;
+  int8_t off6 = 6;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+  v4u inp[4];
+  v4u com[4];
+  uint8_t *out;
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    pulp_nn_u2_to_u8(pIn, inp);
+    pulp_nn_u2_to_u8(pCom, com);
+
+    *((v4u*)out) = max4(inp[0], com[0]);
+
+    uint8_t inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    *((v4u*)out) = max4(inp[1], com[1]);
+
+    inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    *((v4u*)out) = max4(inp[2], com[2]);
+
+    inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    *((v4u*)out) = max4(inp[3], com[3]);
+
+    inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 2, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 2, 2);
+    uint8_t inA2 = (uint8_t) bitextu((unsigned int) *pIn, 2, 4);
+    uint8_t inA3 = (uint8_t) bitextu((unsigned int) *pIn, 2, 6);
+    v4u inA4 = pack((uint8_t) inA0, (uint8_t) inA1, (uint8_t) inA2, (uint8_t) inA3);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 2, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 2, 2);
+    uint8_t inB2 = (uint8_t) bitextu((unsigned int) *pCom, 2, 4);
+    uint8_t inB3 = (uint8_t) bitextu((unsigned int) *pCom, 2, 6);
+    v4u inB4 = pack((uint8_t) inB0, (uint8_t) inB1, (uint8_t) inB2, (uint8_t) inB3);
+
+    *((v4u*)out) = max4(inA4, inB4);
+
+    uint8_t inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+
+static void __attribute__((noinline)) pulp_nn_compare_and_replace_if_larger_i2(int8_t * base,
+                                                int8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask2 = 0x0c;
+  int8_t n_mask2 = ~ mask2;
+  int8_t mask4 = 0x30;
+  int8_t n_mask4 = ~ mask4;
+  int8_t mask6 = 0xc0;
+  int8_t n_mask6 = ~ mask6;
+  int8_t off2 = 2;
+  int8_t off4 = 4;
+  int8_t off6 = 6;
+
+  int8_t *pIn = base;
+  int8_t *pCom = target;
+  v4s inp[4];
+  v4s com[4];
+  int8_t *out;
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    pulp_nn_i2_to_i8(pIn, inp);
+    pulp_nn_i2_to_i8(pCom, com);
+
+    *((v4s*)out) = maxs4(inp[0], com[0]);
+
+    int8_t inA = (int8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((int8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    *((v4s*)out) = maxs4(inp[1], com[1]);
+
+    inA = (int8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((int8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    *((v4s*)out) = maxs4(inp[2], com[2]);
+
+    inA = (int8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((int8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    *((v4s*)out) = maxs4(inp[3], com[3]);
+
+    inA = (int8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((int8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    int8_t inA0 = (int8_t) bitext((unsigned int) *pIn, 2, 0);
+    int8_t inA1 = (int8_t) bitext((unsigned int) *pIn, 2, 2);
+    int8_t inA2 = (int8_t) bitext((unsigned int) *pIn, 2, 4);
+    int8_t inA3 = (int8_t) bitext((unsigned int) *pIn, 2, 6);
+    v4s inA4 = pack((int8_t) inA0, (int8_t) inA1, (int8_t) inA2, (int8_t) inA3);
+    int8_t inB0 = (int8_t) bitext((unsigned int) *pCom, 2, 0);
+    int8_t inB1 = (int8_t) bitext((unsigned int) *pCom, 2, 2);
+    int8_t inB2 = (int8_t) bitext((unsigned int) *pCom, 2, 4);
+    int8_t inB3 = (int8_t) bitext((unsigned int) *pCom, 2, 6);
+    v4s inB4 = pack((int8_t) inB0, (int8_t) inB1, (int8_t) inB2, (int8_t) inB3);
+
+    *((v4s*)out) = maxs4(inA4, inB4);
+
+    int8_t inA = (int8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((int8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_avg_and_replace_u2(uint8_t * base,
+                                  uint8_t * target,
+                                  uint16_t length)
+{
+  int8_t mask2 = 0x0c;
+  int8_t n_mask2 = ~ mask2;
+  int8_t mask4 = 0x30;
+  int8_t n_mask4 = ~ mask4;
+  int8_t mask6 = 0xc0;
+  int8_t n_mask6 = ~ mask6;
+  int8_t off2 = 2;
+  int8_t off4 = 4;
+  int8_t off6 = 6;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+
+  while (length>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 2, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 2, 2);
+    uint8_t inA2 = (uint8_t) bitextu((unsigned int) *pIn, 2, 4);
+    uint8_t inA3 = (uint8_t) bitextu((unsigned int) *pIn, 2, 6);
+
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 2, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 2, 2);
+    uint8_t inB2 = (uint8_t) bitextu((unsigned int) *pCom, 2, 4);
+    uint8_t inB3 = (uint8_t) bitextu((unsigned int) *pCom, 2, 6);
+
+    inA0 = ((inA0 + inB0) >> 1);
+    inA1 = ((inA1 + inB1) >> 1);
+    inA2 = ((inA2 + inB2) >> 1);
+    inA3 = ((inA3 + inB3) >> 1);
+
+    uint8_t inA = (uint8_t) bitins(inA0, n_mask2, inA1, mask2, off2);
+    inA = bitins(inA, n_mask4, inA2, mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, inA3, mask6, off6);
+
+    pIn++;
+    pCom++;
+    length--;
+  }
+}
+
+/*
+ * XpulpNN
+ */
+
+static void __attribute__((noinline)) xpulp_nn_zero_mem_u8(uint8_t * pBuffer, unsigned int size)
+{
+  int lfover = size &0x3;
+  for (int i=0; i<(size>>2); i++)
+  {
+    *((v4u *)pBuffer) = (v4u){0,0,0,0};
+    MemoryFence();
+    pBuffer+=4;
+  }
+  while(lfover)
+  {
+    *pBuffer++=0;
+    lfover--;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_zero_mem_u4(uint8_t * pBuffer, unsigned int size)
+{
+  int lfover = size &0x7;
+  for (int i=0; i<(size>>3); i++)
+  {
+    *((v4u *)pBuffer) = (v4u){0,0,0,0};
+    MemoryFence();
+    pBuffer+=4;
+  }
+  while(lfover)
+  {
+    *pBuffer++=0;
+    lfover-=2;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_zero_mem_u2(uint8_t * pBuffer, unsigned int size)
+{
+  int lfover = size &0xf;
+  for (int i=0; i<(size>>4); i++)
+  {
+    *((v4u *)pBuffer) = (v4u){0,0,0,0};
+    MemoryFence();
+    pBuffer+=4;
+  }
+  while(lfover)
+  {
+    *pBuffer++=0;
+    lfover-=4;
+  }
+}
+
+
+static void __attribute__((noinline)) xpulp_nn_compare_and_replace_if_larger_u4(uint8_t * base,
+                                                uint8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    *((uint32_t *)pIn) = max8(*((uint32_t *)pIn), *((uint32_t *)pCom));
+
+    pIn+=4;
+    pCom+=4;
+
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 4, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 4, 4);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 4, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 4, 4);
+
+    if(inA0<inB0)
+      inA0=inB0;
+
+    if(inA1<inB1)
+      inA1=inB1;
+
+    *((uint8_t*)pIn) = bitins(inA0, n_mask, inA1, mask, off);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_compare_and_replace_if_larger_i4(
+    int8_t * base, int8_t * target, uint16_t length) {
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~mask;
+  int8_t off = 0x04;
+
+  int8_t *pIn = base;
+  int8_t *pCom = target;
+
+  int cnt = length >> 2;
+
+  while (cnt > 0u) {
+    *((int32_t *)pIn) = maxs8(*((int32_t *)pIn), *((int32_t *)pCom));
+
+    pIn += 4;
+    pCom += 4;
+
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left > 0u) {
+    int8_t inA0 = (int8_t)bitext((int)*pIn, 4, 0);
+    int8_t inA1 = (int8_t)bitext((int)*pIn, 4, 4);
+    int8_t inB0 = (int8_t)bitext((int)*pCom, 4, 0);
+    int8_t inB1 = (int8_t)bitext((int)*pCom, 4, 4);
+
+    if (inA0 < inB0)
+      inA0 = inB0;
+
+    if (inA1 < inB1)
+      inA1 = inB1;
+
+    *((int8_t *)pIn) = bitins(inA0, n_mask, inA1, mask, off);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_avg_and_replace_u4(uint8_t * base,
+                                  uint8_t * target,
+                                  uint16_t length)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+
+  int cnt = length >> 2;
+
+  while (cnt > 0u)
+  {
+    *((uint32_t *)pIn) = avg8(*((uint32_t *)pIn), *((uint32_t *)pCom));
+
+    pIn+=4;
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+
+  while (left>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 4, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 4, 4);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 4, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 4, 4);
+
+    inA0 = ((inA0 + inB0) >> 1);
+    inA1 = ((inA1 + inB1) >> 1);
+
+    *((uint8_t*)pIn) = bitins(inA0, n_mask, inA1, mask, off);
+
+    pIn++;
+    pCom++;
+    length--;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_compare_and_replace_if_larger_u2(uint8_t * base,
+                                                uint8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask2 = 0x0c;
+  int8_t n_mask2 = ~ mask2;
+  int8_t mask4 = 0x30;
+  int8_t n_mask4 = ~ mask4;
+  int8_t mask6 = 0xc0;
+  int8_t n_mask6 = ~ mask6;
+  int8_t off2 = 2;
+  int8_t off4 = 4;
+  int8_t off6 = 6;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+  uint8_t *out;
+
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    *((uint32_t *)pIn) = max16(*((uint32_t *)pIn), *((uint32_t *)pCom));
+
+    pIn+=4;
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 2, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 2, 2);
+    uint8_t inA2 = (uint8_t) bitextu((unsigned int) *pIn, 2, 4);
+    uint8_t inA3 = (uint8_t) bitextu((unsigned int) *pIn, 2, 6);
+    v4u inA4 = pack((uint8_t) inA0, (uint8_t) inA1, (uint8_t) inA2, (uint8_t) inA3);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 2, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 2, 2);
+    uint8_t inB2 = (uint8_t) bitextu((unsigned int) *pCom, 2, 4);
+    uint8_t inB3 = (uint8_t) bitextu((unsigned int) *pCom, 2, 6);
+    v4u inB4 = pack((uint8_t) inB0, (uint8_t) inB1, (uint8_t) inB2, (uint8_t) inB3);
+
+    *((v4u*)out) = max4(inA4, inB4);
+
+    uint8_t inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_compare_and_replace_if_larger_i2(int8_t * base,
+                                                int8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask2 = 0x0c;
+  int8_t n_mask2 = ~ mask2;
+  int8_t mask4 = 0x30;
+  int8_t n_mask4 = ~ mask4;
+  int8_t mask6 = 0xc0;
+  int8_t n_mask6 = ~ mask6;
+  int8_t off2 = 2;
+  int8_t off4 = 4;
+  int8_t off6 = 6;
+
+  int8_t *pIn = base;
+  int8_t *pCom = target;
+  int8_t *out;
+
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    *((int32_t *)pIn) = maxs16(*((int32_t *)pIn), *((int32_t *)pCom));
+
+    pIn+=4;
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    int8_t inA0 = (int8_t) bitext((int) *pIn, 2, 0);
+    int8_t inA1 = (int8_t) bitext((int) *pIn, 2, 2);
+    int8_t inA2 = (int8_t) bitext((int) *pIn, 2, 4);
+    int8_t inA3 = (int8_t) bitext((int) *pIn, 2, 6);
+    v4s inA4 = pack((int8_t) inA0, (int8_t) inA1, (int8_t) inA2, (int8_t) inA3);
+    int8_t inB0 = (int8_t) bitext((int) *pCom, 2, 0);
+    int8_t inB1 = (int8_t) bitext((int) *pCom, 2, 2);
+    int8_t inB2 = (int8_t) bitext((int) *pCom, 2, 4);
+    int8_t inB3 = (int8_t) bitext((int) *pCom, 2, 6);
+    v4s inB4 = pack((int8_t) inB0, (int8_t) inB1, (int8_t) inB2, (int8_t) inB3);
+
+    *((v4s*)out) = maxs4(inA4, inB4);
+
+    int8_t inA = (int8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((int8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_avg_and_replace_u2(uint8_t * base,
+                                  uint8_t * target,
+                                  uint16_t length)
+{
+  int8_t mask2 = 0x0c;
+  int8_t n_mask2 = ~ mask2;
+  int8_t mask4 = 0x30;
+  int8_t n_mask4 = ~ mask4;
+  int8_t mask6 = 0xc0;
+  int8_t n_mask6 = ~ mask6;
+  int8_t off2 = 2;
+  int8_t off4 = 4;
+  int8_t off6 = 6;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+  uint8_t *out;
+
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    *((uint32_t *)pIn) = avg16(*((uint32_t *)pIn), *((uint32_t *)pCom));
+
+    pIn+=4;
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 2, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 2, 2);
+    uint8_t inA2 = (uint8_t) bitextu((unsigned int) *pIn, 2, 4);
+    uint8_t inA3 = (uint8_t) bitextu((unsigned int) *pIn, 2, 6);
+    v4u inA4 = pack((uint8_t) inA0, (uint8_t) inA1, (uint8_t) inA2, (uint8_t) inA3);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 2, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 2, 2);
+    uint8_t inB2 = (uint8_t) bitextu((unsigned int) *pCom, 2, 4);
+    uint8_t inB3 = (uint8_t) bitextu((unsigned int) *pCom, 2, 6);
+    v4u inB4 = pack((uint8_t) inB0, (uint8_t) inB1, (uint8_t) inB2, (uint8_t) inB3);
+
+    *((v4u*)out) = avg4(inA4, inB4);
+
+    uint8_t inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+
+#endif
diff --git a/rt_nn_tests/xpnnv2_conv/test.c b/rt_nn_tests/xpnnv2_conv/test.c
new file mode 100644
index 0000000..a332ad8
--- /dev/null
+++ b/rt_nn_tests/xpnnv2_conv/test.c
@@ -0,0 +1,221 @@
+#include <stdio.h>
+#include <stdint.h>
+#include <stdlib.h>
+
+
+#include "pmsis.h"
+
+#include "data_statstest.h"
+//#include "pulp_nn_kernels.h"
+#include "pulp_nn_mix_kernels.h"
+
+
+#define start_cycle_counter()                                   asm volatile("csrw 0xCC0, 0x01;")
+#define stop_cycle_counter()                                    asm volatile("csrw 0xCC0, 0x00;")
+#define read_cycle_counter(x)                                   asm volatile("csrr %0, 0x780;" : "=r" (x))
+#define reset_cycle_counter()                                   asm volatile("csrw 0x780, 0x0;")
+
+uint8_t im2col[IM2COL_DIM] = {0};
+uint8_t outputs[OUTPUT_DIM] = {0};
+
+int32_t outputs_fp[OUTPUT_DIM_FP] = {0};
+
+#ifndef PROFILE
+int num_cycles;
+#else
+int im2col_cycles;
+int hotloop_prep_cycles;
+int hotloop_cycles;
+int threshold_cycles;
+int requant_cycles;
+int hotloop_leftover_cycles;
+int matmul4x2_leftover_cycles;
+#endif
+
+void call_krnl_0();
+void test_0();
+
+int main(int argc, char *argv[])
+{
+#if KRAKEN_PTEST == 1
+  kraken_padframe_aon_pad_gpioa_cfg_rxe_set(24, 0);
+  kraken_padframe_aon_pad_gpioa_cfg_trie_set(24, 0);
+  kraken_padframe_aon_pad_gpioa_cfg_chip2pad_set(24, 0);
+#endif
+  int n_mismatches = 0;
+  int mismatches_tot = 0;
+
+    
+
+  #ifndef PROFILE
+  num_cycles = 0;
+  #endif
+
+  if (get_core_id() == 0) {
+  printf("===> TEST 0: Running xpulp_nn_mix_conv_u4_u8_i2...\n");
+  printf("  dims_in     = [4, 4]\n");
+  printf("  dims_kernel = [3, 3]\n");
+  printf("  ch_in/out   = [16, 32]\n");
+  //printf("  padding_y_top    = [%d]\n", padding_y_top);
+  //printf("  padding_y_bottom = [%d]\n", padding_y_bottom);
+  //printf("  padding_x_left   = [%d]\n", padding_x_left);
+  //printf("  padding_x_right  = [%d]\n", padding_x_right);
+  //printf("  stride_x         = [%d]\n", stride_x);
+  //printf("  stride_y         = [%d]\n", stride_y);
+  }
+  test_0();
+  #ifndef PROFILE
+  stop_cycle_counter();
+  read_cycle_counter(num_cycles);
+  if (get_core_id() == 0) {
+  printf("===> TEST 0: Finished running xpulp_nn_mix_conv_u4_u8_i2\n");
+  printf("num_cycles = %d\n", num_cycles);
+  printf("MACs = 73728\n");
+  printf("MACs/cycle = %.4f\n", 73728/num_cycles);
+  }
+  #endif
+  if (get_core_id() == 0) {
+  printf("Checking for mismatches..\n");
+  n_mismatches = 0;
+
+  for(int i=0; i < 512; i++) {
+    if (outputs[i] != exp_outp_0[i]){
+        printf("***Mismatch in test 0 at iteration %d: Expected: %x, got: %x\n", i, exp_outp_0[i], outputs[i]);
+        n_mismatches++;
+    }
+  }
+  }
+  mismatches_tot += n_mismatches;
+
+  if (get_core_id() == 0) {
+  printf("Got %d mismatches in %d tests\n", mismatches_tot, 1);
+  }
+  return mismatches_tot;
+}
+
+
+void call_krnl_0(void) {
+  uint8_t * pInp;
+  uint8_t * pIm2ColBuffer;
+  int8_t * pBias = NULL;
+  uint8_t * pOut;
+  int8_t * pWeight;
+  uint32_t * pThr;
+  int32_t * pKappa, pLambda;
+  #ifdef PROFILE
+  int32_t im2col_cycles = 0;
+  int32_t hotloop_prep_cycles = 0;
+  int32_t hotloop_cycles = 0;
+  int32_t threshold_cycles = 0;
+  int32_t requant_cycles = 0;
+  int32_t hotloop_leftover_cycles = 0;
+  int32_t matmul4x2_leftover_cycles = 0;
+  #endif
+  pInp = inp_l1;
+  
+  pOut = outp_l1;
+  pIm2ColBuffer = im2col_l1;
+  pWeight = wt_l1;
+  pThr = threshs_l1;
+  pKappa = kappa_l1;
+  pLambda = lambda_l1;
+#if KRAKEN_PTEST == 1
+  if (pi_core_id() == 0) {
+    kraken_padframe_aon_pad_gpioa_cfg_chip2pad_set(24, 1);
+  } 
+#endif
+  xpulp_nn_mix_conv_u4_u8_i2(
+    pInp,
+    pIm2ColBuffer,
+    pBias,
+    pOut,
+    pWeight,
+    pKappa,
+    pLambda,
+    1,
+    13,
+    4,
+    4,
+    16,
+    4,
+    4,
+    32,
+    3,
+    3,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+#ifndef PROFILE
+    1
+#else
+    1,
+    &im2col_cycles,
+    &hotloop_prep_cycles,
+    &hotloop_cycles,
+    &requant_cycles,
+    &hotloop_leftover_cycles,
+    &matmul4x2_leftover_cycles
+#endif
+    );
+
+#if KRAKEN_PTEST == 1
+  if (pi_core_id() == 0) {
+    kraken_padframe_aon_pad_gpioa_cfg_chip2pad_set(24, 0);
+  }
+#endif
+  #ifdef PROFILE
+  if (pi_core_id() == 0) {
+  printf("im2col_cycles = %d\n", im2col_cycles);
+  printf("hotloop_prep_cycles = %d\n", hotloop_prep_cycles);
+  printf("hotloop_cycles = %d\n", hotloop_cycles);
+  printf("requant_cycles = %d\n", requant_cycles);
+  printf("threshold_cycles = %d\n", threshold_cycles);
+  printf("hotloop_leftover_cycles = %d\n", hotloop_leftover_cycles);
+  printf("matmul4x2_leftover_cycles = %d\n", matmul4x2_leftover_cycles);
+  }
+  #endif
+}
+
+void test_0(void) {
+  // DMA transfer inputs from L2 to L1
+  if (pi_core_id() == 0) {
+     plp_dma_memcpy(pIn_0, inp_l1, 128, 1);
+     plp_dma_barrier();
+  }
+  pi_cl_team_barrier(0);
+  if (pi_core_id() == 0) {
+     plp_dma_memcpy(pLambda_0, lambda_l1, 32 * 4, 1); // 4 bytes per lambda item
+     plp_dma_barrier();
+  }
+  if (pi_core_id() == 0) {
+     plp_dma_memcpy(pKappa_0, kappa_l1, 32 * 4, 1); // 4 bytes per lambda item
+     plp_dma_barrier();
+  }
+  pi_cl_team_barrier(0);
+  // transfer weights
+  if (pi_core_id() == 0) {
+     plp_dma_memcpy(pWeight_0, wt_l1, 1152, 1);
+     plp_dma_barrier();
+  }
+  pi_cl_team_barrier(0);
+  call_krnl_0();
+  // get outputs back with DMA
+  if (pi_core_id() == 0) {
+     plp_dma_memcpy(outputs, outp_l1, 512, 0);
+     plp_dma_barrier();
+  }
+}
+
+
+
+
+
+
+
+
+
+
diff --git a/rt_nn_tests/xpnnv2_conv/xpulp_nn_mix_conv_u4_u8_i2.c b/rt_nn_tests/xpnnv2_conv/xpulp_nn_mix_conv_u4_u8_i2.c
new file mode 100644
index 0000000..6055ed9
--- /dev/null
+++ b/rt_nn_tests/xpnnv2_conv/xpulp_nn_mix_conv_u4_u8_i2.c
@@ -0,0 +1,298 @@
+/*
+ * xpulp_nn_mix_conv_u4_u8_i2.c
+ * Nazareno   Bruschi  <nazareno.bruschi@unibo.it>
+ * Alessandro Nadalini <alessandro.nadalini3@unibo.it>
+ *
+ * Copyright (C) 2019-2020 University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include "pmsis.h"
+#include "pulp_nn_utils.h"
+#include "pulp_nn_kernels.h"
+
+
+void __attribute__((noinline)) xpulp_nn_mix_conv_u4_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mult,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batch_norm)
+{
+  uint16_t ch_in_r = PACK_INT4_SIZE(ch_in);
+  uint16_t ch_out_r = PACK_INT8_SIZE(ch_out);
+
+  int core_id = pi_core_id();
+  uint8_t * pIm2ColBase = pIm2ColBuffer + (2 * core_id * PACK_INT4_SIZE(ch_in) * dim_kernel_x * dim_kernel_y);
+  int i_out_y, i_out_x, i_ker_y, i_ker_x;
+  int Log2Core;
+
+  uint8_t extra_chunk = ((dim_out_y & (NUM_CORES-1)) != 0);
+  uint8_t extra_chunk_r;
+  uint16_t dim_out_x_r;
+  uint8_t section;
+  int core_id_r;
+
+  if(extra_chunk && dim_out_x > 1)
+  {
+    Log2Core = log2(NUM_CORES >> 1);
+    core_id_r = (core_id >> 1);
+    dim_out_x_r = (dim_out_x >> 1);
+    section = (core_id & 0x1);
+    extra_chunk_r = ((dim_out_y & ((NUM_CORES >> 1) - 1)) != 0);
+  }
+  else
+  {
+    Log2Core = log2(NUM_CORES);
+    core_id_r = core_id;
+    dim_out_x_r = dim_out_x;
+    section = 0;
+    extra_chunk_r = extra_chunk;
+    extra_chunk = 0;
+  }
+
+  uint8_t flag_dim_out_x_odd = dim_out_x & 0x01;
+
+  int chunk = (dim_out_y >> Log2Core) + extra_chunk_r;
+
+  int start_pixel = min((chunk * core_id_r), dim_out_y);
+  int stop_pixel = min(start_pixel + chunk, dim_out_y);
+
+  uint8_t *pIm2Col = pIm2ColBase;
+  uint8_t *pOutBuffer = pOut + (start_pixel * ch_out_r * dim_out_x) + (section * ch_out_r * dim_out_x_r);
+
+  for (i_out_y = start_pixel; i_out_y < stop_pixel; i_out_y++)
+  {
+    for(i_out_x=(section * dim_out_x_r); i_out_x<(dim_out_x_r + (section * (dim_out_x_r + flag_dim_out_x_odd))); i_out_x++)
+    {
+      if(i_out_y < padding_y_top)
+      {
+        for(i_ker_y=((i_out_y * stride_y) - padding_y_top); i_ker_y<((i_out_y * stride_y) - padding_y_top + dim_kernel_y); i_ker_y++)
+        {
+          for(i_ker_x=((i_out_x * stride_x) - padding_x_left); i_ker_x<((i_out_x * stride_x) - padding_x_left + dim_kernel_x); i_ker_x++)
+          {
+            if((i_ker_y < 0) || (i_ker_y >= dim_in_y) || (i_ker_x < 0) || (i_ker_x >= dim_in_x))
+            {
+              xpulp_nn_zero_mem_u4(pIm2Col, ch_in);
+            }
+            else
+            {
+              xpulp_nn_im2col_u4_to_u4((uint8_t*) (pIn + ((i_ker_y * dim_in_x + i_ker_x) * ch_in_r)), pIm2Col, ch_in);
+            }
+            pIm2Col+=PACK_INT4_SIZE(ch_in);
+          }
+        }
+      }
+      else if(i_out_y < dim_out_y - padding_y_bottom)
+      {
+        if(i_out_x < padding_x_left)
+        {
+          for(i_ker_y=((i_out_y * stride_y) - padding_y_top); i_ker_y<((i_out_y * stride_y) - padding_y_top + dim_kernel_y); i_ker_y++)
+          {
+            for(i_ker_x=((i_out_x * stride_x) - padding_x_left); i_ker_x<((i_out_x * stride_x) - padding_x_left + dim_kernel_x); i_ker_x++)
+            {
+              if((i_ker_x < 0) || (i_ker_x >= dim_in_x))
+              {
+                xpulp_nn_zero_mem_u4(pIm2Col, ch_in);
+              }
+              else
+              {
+                xpulp_nn_im2col_u4_to_u4((uint8_t*) (pIn + ((i_ker_y * dim_in_x + i_ker_x) * ch_in_r)), pIm2Col, ch_in);
+              }
+              pIm2Col+=PACK_INT4_SIZE(ch_in);
+            }
+          }
+        }
+        else if(i_out_x < (dim_out_x - padding_x_right))
+        {
+          for(i_ker_y=((i_out_y * stride_y) - padding_y_top); i_ker_y<((i_out_y * stride_y) - padding_y_top + dim_kernel_y); i_ker_y++)
+          {
+            xpulp_nn_im2col_u4_to_u4((uint8_t*) pIn + (i_ker_y * dim_in_x + i_out_x * stride_x - padding_x_left)*ch_in_r,pIm2Col,ch_in * dim_kernel_x);
+            pIm2Col+=PACK_INT4_SIZE(ch_in * dim_kernel_x);
+          }
+        }
+        else
+        {
+          for(i_ker_y=((i_out_y * stride_y) - padding_y_top); i_ker_y<((i_out_y * stride_y) - padding_y_top + dim_kernel_y); i_ker_y++)
+          {
+            for(i_ker_x = i_out_x * stride_x - padding_x_left; i_ker_x < i_out_x * stride_x - padding_x_left + dim_kernel_x; i_ker_x++)
+            {
+              if((i_ker_x < 0) || (i_ker_x >= dim_in_x))
+              {
+                xpulp_nn_zero_mem_u4(pIm2Col, ch_in);
+              }
+              else
+              {
+                xpulp_nn_im2col_u4_to_u4((uint8_t *)pIn + (i_ker_y*dim_in_x+i_ker_x)* ch_in_r, pIm2Col, ch_in);
+              }
+              pIm2Col+=PACK_INT4_SIZE(ch_in);
+            }
+          }
+        }
+      }
+      else
+      {
+        for(i_ker_y=((i_out_y * stride_y) - padding_y_top); i_ker_y<((i_out_y * stride_y) - padding_y_top + dim_kernel_y); i_ker_y++)
+        {
+          for(i_ker_x = i_out_x * stride_x - padding_x_left; i_ker_x < i_out_x * stride_x - padding_x_left + dim_kernel_x; i_ker_x++)
+          {
+            if(i_ker_y < 0 || (i_ker_y >= dim_in_y) || i_ker_x < 0 || i_ker_x >= dim_in_x)
+            {
+              xpulp_nn_zero_mem_u4(pIm2Col, ch_in);
+            }
+            else
+            {
+              xpulp_nn_im2col_u4_to_u4((uint8_t *) pIn + (i_ker_y * dim_in_x + i_ker_x) * ch_in_r, pIm2Col, ch_in);
+            }
+            pIm2Col+=PACK_INT4_SIZE(ch_in);
+          }
+        }
+      }
+      if(pIm2Col == (pIm2ColBase + ((PACK_INT4_SIZE(ch_in) * dim_kernel_x * dim_kernel_y) << 1)))
+      {
+        pOutBuffer = xpulp_nn_mix_matmul_u4_u8_i2(
+          pIm2ColBase,
+          pBias,
+          pOutBuffer,
+          pOutBuffer + ch_out_r,
+          pWeight,
+          pKappa,
+          pLambda,
+          out_mult,
+          out_shift,
+          (ch_in * dim_kernel_x * dim_kernel_y),
+          ch_out,
+          flag_relu,
+          flag_batch_norm
+          );
+
+        pIm2Col = pIm2ColBase;
+      }
+    }
+
+    if(pIm2Col != pIm2ColBase)
+    {
+
+      MIXED_SKIP("1");
+
+      const int8_t *pA = pWeight;
+      int i;
+      int32_t * k1 = pKappa;
+      int32_t * lambda1 = pLambda;
+      v4s inA[2];
+      uint16_t num_col_im2col = ch_in * dim_kernel_x * dim_kernel_y;
+      uint16_t num_col_im2col_w = PACK_INT2_SIZE(ch_in) * dim_kernel_x * dim_kernel_y;
+
+      for(i = 0; i < ch_out; i++)
+      {
+        int sum = 0;
+        if (pBias != NULL)
+        {
+          sum = *((int*) pBias);
+          pBias += 4;
+        }
+
+        uint8_t *pB = pIm2ColBase;
+
+        int32_t *ptrA  = (int32_t *)pA;
+        uint32_t *ptrB = (uint32_t *)pB;
+
+        for(int j=0; j < (num_col_im2col >> 4); j++)
+        {
+          sum = SumDotp8(*(uint32_t *)ptrB, *(int32_t *)ptrA, sum);
+
+          ptrB++;
+
+          sum = SumDotp8(*(uint32_t *)ptrB, *(int32_t *)ptrA, sum);
+
+          ptrA++;
+          ptrB++;
+        }
+
+        int col_cnt_im2col = num_col_im2col & 0xf;
+
+        if(col_cnt_im2col)
+        {
+
+          uint16_t loop_cnt_im2col_a = (num_col_im2col >> 4) << 3;
+          pB+=loop_cnt_im2col_a;
+
+          do
+          {
+            int8_t inA1 = (int8_t) bitext((int) *pA, 2, 0);
+            uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pB, 4, 0);
+            sum += inA1 * inB1;
+            inA1 = (int8_t) bitext((int) *pA, 2, 2);
+            inB1 = (uint8_t) bitextu((unsigned int) *pB, 4, 4);
+            sum += inA1 * inB1;
+            pB++;
+            inA1 = (int8_t) bitext((int) *pA, 2, 4);
+            inB1 = (uint8_t) bitextu((unsigned int) *pB, 4, 0);
+            sum += inA1 * inB1;
+            inA1 = (int8_t) bitext((int) *pA, 2, 6);
+            inB1 = (uint8_t) bitextu((unsigned int) *pB, 4, 4);
+            sum += inA1 * inB1;
+
+            pA++;
+            pB++;
+            col_cnt_im2col-=4;
+          } while(col_cnt_im2col);
+        }
+        if (flag_batch_norm && flag_relu)
+        {
+          *pOutBuffer = pulp_nn_bn_quant_u8(sum, *k1, *lambda1, out_shift);
+          k1++;
+          lambda1++;
+          pOutBuffer++;
+        }
+        else
+        {
+          if(flag_relu == 1)
+          {
+            *pOutBuffer = pulp_nn_quant_u8(sum, out_mult, out_shift);
+            pOutBuffer++;
+          }
+          else
+          {
+            *pOutBuffer = (uint8_t) clip8(sum >> out_shift);
+            pOutBuffer++;
+          }
+        }
+      }
+    }
+    pOutBuffer+=(extra_chunk * ((dim_out_x_r + ((1 - section) * flag_dim_out_x_odd)) * ch_out_r));
+    pIm2Col = pIm2ColBase;
+  }
+  pi_cl_team_barrier();
+}
diff --git a/rt_nn_tests/xpnnv2_conv/xpulp_nn_mix_matmul_u4_u8_i2.c b/rt_nn_tests/xpnnv2_conv/xpulp_nn_mix_matmul_u4_u8_i2.c
new file mode 100644
index 0000000..bd31a9c
--- /dev/null
+++ b/rt_nn_tests/xpnnv2_conv/xpulp_nn_mix_matmul_u4_u8_i2.c
@@ -0,0 +1,470 @@
+/*
+ * xpulp_nn_mix_matmul_u4_u8_i2.c
+ * Nazareno   Bruschi  <nazareno.bruschi@unibo.it>
+ * Alessandro Nadalini <alessandro.nadalini3@unibo.it>
+ *
+ * Copyright (C) 2019-2020 University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include "pmsis.h"
+#include "pulp_nn_utils.h"
+
+
+uint8_t * __attribute__((noinline)) xpulp_nn_mix_matmul_u4_u8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mult,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batch_norm)
+{
+
+  uint16_t ch_out_r = PACK_INT8_SIZE(ch_out);
+
+  uint16_t num_col_im2col_w = PACK_INT2_SIZE(num_col_im2col);
+  uint16_t num_col_im2col_a = PACK_INT4_SIZE(num_col_im2col);
+
+  int32_t a_rollback = 4 - num_col_im2col_a;
+  int32_t w_rollback = 4 - (num_col_im2col_w + (num_col_im2col_w << 1));
+
+  LEGACY_MODE("0");
+  IVEC_FMT("5");
+  A_STRIDE(num_col_im2col_a);
+  W_STRIDE(num_col_im2col_w);
+  A_ROLLBACK(a_rollback);
+  W_ROLLBACK(w_rollback);
+  A_SKIP("1");
+  W_SKIP("3");
+  MIXED_SKIP("8");
+
+  int8_t *pA = pWeight;
+
+  uint16_t chan_left = ch_out & 0x3;
+
+  for(int i=0; i < (ch_out >> 2); i++)
+  {
+    uint8_t *pB =  pIn;
+
+    uint32_t *ptrB  = (uint32_t *) pB;
+
+    int32_t *ptrA  = (int32_t *) pA ;
+
+    A_ADDRESS(ptrB);
+    W_ADDRESS(ptrA);
+
+    ptrA = MacLoadInit(1, 0, 0, 0, ptrA);
+    ptrA = MacLoadInit(1, 0, 1, 0, ptrA);
+    ptrA = MacLoadInit(1, 0, 2, 0, ptrA);
+    ptrA = MacLoadInit(1, 0, 3, 0, ptrA);
+
+    ptrB = MacLoadInit(0, 1, 0, 0, ptrB);
+
+    int sum  = 0;
+    int sum2 = 0;
+    int sum3 = 0;
+    int sum4 = 0;
+    int sum5 = 0;
+    int sum6 = 0;
+    int sum7 = 0;
+    int sum8 = 0;
+
+
+    if (pBias != NULL)
+    {
+      sum =  *((int*) pBias);
+      pBias += 4;
+      sum2 = *((int*) pBias);
+      pBias += 4;
+      sum3 = *((int*) pBias);
+      pBias += 4;
+      sum4 = *((int*) pBias);
+      pBias += 4;
+
+      sum5 = sum;
+      sum6 = sum2;
+      sum7 = sum3;
+      sum8 = sum4;
+    }
+
+    for(int j=0; j<(num_col_im2col >> 4); j++)
+    {
+      ptrB  = MacLoadInit(0, 1, 0, 1, ptrB);
+
+      sum  = MacLoad8(0, 0, 0, 0, ptrA, sum);
+      sum2 = MacLoad8(0, 0, 1, 0, ptrA, sum2);
+      sum3 = MacLoad8(0, 0, 2, 0, ptrA, sum3);
+      sum4 = MacLoad8(0, 1, 3, 0, ptrB, sum4);
+      ptrB = MacLoadUpdate(ptrB);
+
+      sum5 = MacLoad8(0, 0, 0, 1, ptrA,  sum5);
+      sum6 = MacLoad8(0, 0, 1, 1, ptrA, sum6);
+      sum7 = MacLoad8(0, 0, 2, 1, ptrA, sum7);
+      sum8 = MacLoad8(0, 1, 3, 1, ptrB, sum8);
+      ptrB = MacLoadUpdate(ptrB);
+
+
+      MemoryFence();
+
+      sum  = MacLoad8(0, 0, 0, 0, ptrA, sum);
+      sum2 = MacLoad8(0, 0, 1, 0, ptrA, sum2);
+      sum3 = MacLoad8(0, 0, 2, 0, ptrA, sum3);
+      sum4 = MacLoad8(0, 1, 3, 0, ptrB, sum4);      
+      ptrB = MacLoadUpdate(ptrB);
+
+      sum5 = MacLoad8(1, 0, 0, 1, ptrA, sum5);
+      ptrA = MacLoadUpdate(ptrA);
+
+      sum6 = MacLoad8(1, 0, 1, 1, ptrA, sum6);
+      ptrA = MacLoadUpdate(ptrA);
+
+      sum7 = MacLoad8(1, 0, 2, 1, ptrA, sum7);
+      ptrA = MacLoadUpdate(ptrA);
+
+      sum8 = MacLoad8(1, 0, 3, 1, ptrA, sum8);
+      ptrA = MacLoadUpdate(ptrA);
+    }
+
+    asm volatile ("csrr %0, 0x101" : "=r" (pA));
+    pA-=4;
+    
+    int col_cnt_im2col = num_col_im2col & 0xf;
+
+    if(col_cnt_im2col)
+    {
+
+      uint16_t loop_cnt_im2col_a = (num_col_im2col >> 4) << 3;
+      
+      int8_t *pA2 = (pA  + num_col_im2col_w);
+      int8_t *pA3 = (pA2 + num_col_im2col_w);
+      int8_t *pA4 = (pA3 + num_col_im2col_w);
+
+      pB+=loop_cnt_im2col_a;
+      
+      uint8_t *pB2 = (pB + num_col_im2col_a);
+
+      do
+      {
+        int8_t inA = (int8_t) bitext((int) *pA, 2, 0);
+        int8_t inA2 = (int8_t) bitext((int) *pA2, 2, 0);
+        int8_t inA3 = (int8_t) bitext((int) *pA3, 2, 0);
+        int8_t inA4 = (int8_t) bitext((int) *pA4, 2, 0);
+
+        uint8_t inB = (uint8_t)bitextu((unsigned int) *pB, 4, 0);
+        uint8_t inB2 = (uint8_t)bitextu((unsigned int) *pB2, 4, 0);
+
+        sum += inA * inB;
+        sum2 += inA2 * inB;
+        sum3 += inA3 * inB;
+        sum4 += inA4 * inB;
+
+        sum5 += inA * inB2;
+        sum6 += inA2 * inB2;
+        sum7 += inA3 * inB2;
+        sum8 += inA4 * inB2;
+
+        inA = (int8_t) bitext((int) *pA, 2, 2);
+        inA2 = (int8_t) bitext((int) *pA2, 2, 2);
+        inA3 = (int8_t) bitext((int) *pA3, 2, 2);
+        inA4 = (int8_t) bitext((int) *pA4, 2, 2);
+
+        inB = (uint8_t)bitextu((unsigned int) *pB, 4, 4);
+        inB2 = (uint8_t)bitextu((unsigned int) *pB2, 4, 4);
+
+        sum += inA * inB;
+        sum2 += inA2 * inB;
+        sum3 += inA3 * inB;
+        sum4 += inA4 * inB;
+
+        sum5 += inA * inB2;
+        sum6 += inA2 * inB2;
+        sum7 += inA3 * inB2;
+        sum8 += inA4 * inB2;
+
+        pB++;
+        pB2++;
+
+        inA = (int8_t) bitext((int) *pA, 2, 4);
+        inA2 = (int8_t) bitext((int) *pA2, 2, 4);
+        inA3 = (int8_t) bitext((int) *pA3, 2, 4);
+        inA4 = (int8_t) bitext((int) *pA4, 2, 4);
+
+        inB = (uint8_t)bitextu((unsigned int) *pB, 4, 0);
+        inB2 = (uint8_t)bitextu((unsigned int) *pB2, 4, 0);
+
+        sum += inA * inB;
+        sum2 += inA2 * inB;
+        sum3 += inA3 * inB;
+        sum4 += inA4 * inB;
+
+        sum5 += inA * inB2;
+        sum6 += inA2 * inB2;
+        sum7 += inA3 * inB2;
+        sum8 += inA4 * inB2;
+
+        inA = (int8_t) bitext((int) *pA, 2, 6);
+        inA2 = (int8_t) bitext((int) *pA2, 2, 6);
+        inA3 = (int8_t) bitext((int) *pA3, 2, 6);
+        inA4 = (int8_t) bitext((int) *pA4, 2, 6);
+
+        inB = (uint8_t)bitextu((unsigned int) *pB, 4, 4);
+        inB2 = (uint8_t)bitextu((unsigned int) *pB2, 4, 4);
+
+        sum += inA * inB;
+        sum2 += inA2 * inB;
+        sum3 += inA3 * inB;
+        sum4 += inA4 * inB;
+
+        sum5 += inA * inB2;
+        sum6 += inA2 * inB2;
+        sum7 += inA3 * inB2;
+        sum8 += inA4 * inB2;
+
+        pA++;
+        pA2++;
+        pA3++;
+        pA4++;
+
+        pB++;
+        pB2++;
+
+        col_cnt_im2col-=4;
+      } while(col_cnt_im2col);
+    }
+    if (flag_batch_norm && flag_relu)
+    {
+      *pOut = pulp_nn_bn_quant_u8(sum, *pKappa, *pLambda, out_shift);
+      pOut++;
+      *pOut2 = pulp_nn_bn_quant_u8(sum5, *pKappa, *pLambda, out_shift);
+      pOut2++;
+      pKappa++;
+      pLambda++;
+
+      *pOut = pulp_nn_bn_quant_u8(sum2, *pKappa, *pLambda, out_shift);
+      pOut++;
+      *pOut2 = pulp_nn_bn_quant_u8(sum6, *pKappa, *pLambda, out_shift);
+      pOut2++;
+      pKappa++;
+      pLambda++;
+
+      *pOut = pulp_nn_bn_quant_u8(sum3, *pKappa, *pLambda, out_shift);
+      pOut++;
+      *pOut2 = pulp_nn_bn_quant_u8(sum7, *pKappa, *pLambda, out_shift);
+      pOut2++;
+      pKappa++;
+      pLambda++;
+
+      *pOut = pulp_nn_bn_quant_u8(sum4, *pKappa, *pLambda, out_shift);
+      pOut++;
+      *pOut2 = pulp_nn_bn_quant_u8(sum8, *pKappa, *pLambda, out_shift);
+      pOut2++;
+      pKappa++;
+      pLambda++;
+    }
+    else
+    {
+      if (flag_relu == 1)
+      {
+        *pOut = pulp_nn_quant_u8(sum, out_mult, out_shift);
+        pOut++;
+        *pOut = pulp_nn_quant_u8(sum2, out_mult, out_shift);
+        pOut++;
+        *pOut = pulp_nn_quant_u8(sum3, out_mult, out_shift);
+        pOut++;
+        *pOut = pulp_nn_quant_u8(sum4, out_mult, out_shift);
+        pOut++;
+
+        *pOut2 = pulp_nn_quant_u8(sum5, out_mult, out_shift);
+        pOut2++;
+        *pOut2 = pulp_nn_quant_u8(sum6, out_mult, out_shift);
+        pOut2++;
+        *pOut2 = pulp_nn_quant_u8(sum7, out_mult, out_shift);
+        pOut2++;
+        *pOut2 = pulp_nn_quant_u8(sum8, out_mult, out_shift);
+        pOut2++;
+
+      }
+      else
+      {
+        *pOut = (uint8_t) clip8(sum >> out_shift);
+        pOut++;
+        *pOut = (uint8_t) clip8(sum2 >> out_shift);
+        pOut++;
+        *pOut = (uint8_t) clip8(sum3 >> out_shift);
+        pOut++;
+        *pOut = (uint8_t) clip8(sum4 >> out_shift);
+        pOut++;
+
+        *pOut2 = (uint8_t) clip8(sum5 >> out_shift);
+        pOut2++;
+        *pOut2 = (uint8_t) clip8(sum6 >> out_shift);
+        pOut2++;
+        *pOut2 = (uint8_t) clip8(sum7 >> out_shift);
+        pOut2++;
+        *pOut2 = (uint8_t) clip8(sum8 >> out_shift);
+        pOut2++;
+
+      }
+    }
+    pA+=(3 * num_col_im2col_w);
+  }
+
+  w_rollback = 4;
+  W_ROLLBACK(w_rollback);
+  W_SKIP("0");
+  MIXED_SKIP("2");
+
+  while(chan_left)
+  {
+    uint8_t *pB = pIn;
+
+    int8_t *pA = pWeight + (num_col_im2col_w * (ch_out - chan_left));
+
+    uint32_t *ptrB  = (uint32_t *) pB;
+
+    int32_t *ptrA  = (int32_t *) pA;
+
+    A_ADDRESS(ptrB);
+    W_ADDRESS(ptrA);
+
+    ptrA  = MacLoadInit(1, 0, 0, 0, ptrA);
+
+    ptrB  = MacLoadInit(0, 1, 0, 0, ptrB);
+
+    int sum = 0;
+    if (pBias != NULL)
+    {
+      sum = ((int) (*pBias++));    
+    }
+    int sum2 = sum;
+
+    for(int j=0; j < (num_col_im2col >> 4); j++)
+    {
+      ptrB  = MacLoadInit(0, 1, 0, 1, ptrB);
+
+      sum  = MacLoad8(0, 1, 0, 0, ptrB, sum);
+      ptrB = MacLoadUpdate(ptrB);
+
+      sum2 = MacLoad8(0, 1, 0, 1, ptrB, sum2);
+      ptrB = MacLoadUpdate(ptrB);
+
+      sum  = MacLoad8(0, 1, 0, 0, ptrB, sum);   
+      ptrB = MacLoadUpdate(ptrB);
+
+      sum2  = MacLoad8(1, 0, 0, 1, ptrA, sum2);
+      ptrA  = MacLoadUpdate(ptrA);
+    }
+    asm volatile ("csrr %0, 0x101" : "=r" (pA));
+    pA-=4;
+    int col_cnt_im2col = num_col_im2col & 0xf;
+
+    if(col_cnt_im2col)
+    {
+
+      uint16_t loop_cnt_im2col_a = (num_col_im2col >> 4) << 3;
+      pB+=loop_cnt_im2col_a;
+      
+      uint8_t *pB2 = (pB +loop_cnt_im2col_a);
+
+      int8_t *pA2 = (pA  + num_col_im2col_w);
+      int8_t *pA3 = (pA2 + num_col_im2col_w);
+      int8_t *pA4 = (pA3 + num_col_im2col_w);
+
+      do
+      {
+        int8_t inA = (int8_t) bitext((int) *pA, 2, 0);
+
+        uint8_t inB = (uint8_t)bitextu((unsigned int) *pB, 4, 0);
+        uint8_t inB2 = (uint8_t)bitextu((unsigned int) *pB2, 4, 0);
+
+        sum += inA * inB;
+
+        sum2 += inA * inB2;
+
+        inA = (int8_t) bitext((int) *pA, 2, 2);
+
+        inB = (uint8_t)bitextu((unsigned int) *pB, 4, 4);
+        inB2 = (uint8_t)bitextu((unsigned int) *pB2, 4, 4);
+
+        sum += inA * inB;
+
+        sum2 += inA * inB2;
+
+        pB++;
+        pB2++;
+
+        inA = (int8_t) bitext((int) *pA, 2, 4);
+
+        inB = (uint8_t)bitextu((unsigned int) *pB, 4, 0);
+        inB2 = (uint8_t)bitextu((unsigned int) *pB2, 4, 0);
+
+        sum += inA * inB;
+
+        sum2 += inA * inB2;
+
+        inA = (int8_t) bitext((int) *pA, 2, 6);
+
+        inB = (uint8_t)bitextu((unsigned int) *pB, 4, 4);
+        inB2 = (uint8_t)bitextu((unsigned int) *pB2, 4, 4);
+
+        sum += inA * inB;
+
+        sum2 += inA * inB2;
+
+        pA++;
+
+        pB++;
+        pB2++;
+
+        col_cnt_im2col-=4;
+      } while(col_cnt_im2col);
+    }
+    if (flag_batch_norm && flag_relu)
+    {
+      *pOut = pulp_nn_bn_quant_u8(sum, *pKappa, *pLambda, out_shift);
+      pOut++;
+      *pOut2 = pulp_nn_bn_quant_u8(sum2, *pKappa, *pLambda, out_shift);
+      pOut2++;
+      pKappa++;
+      pLambda++;
+    }
+    else
+    {
+      if (flag_relu == 1)
+      {
+        *pOut = pulp_nn_quant_u8(sum, out_mult, out_shift);
+        pOut++;
+        *pOut2 = pulp_nn_quant_u8(sum2, out_mult, out_shift);
+        pOut2++;
+      }
+      else
+      {
+        *pOut = (uint8_t) clip8(sum >> out_shift);
+        pOut++;
+        *pOut2 = (uint8_t) clip8(sum2 >> out_shift);
+        pOut2++;
+      }
+    }
+    chan_left--;
+  }
+  pOut+=ch_out_r;
+  return pOut;
+}
diff --git a/rt_nn_tests/xptnn_conv/Makefile b/rt_nn_tests/xptnn_conv/Makefile
new file mode 100644
index 0000000..ea4fb4e
--- /dev/null
+++ b/rt_nn_tests/xptnn_conv/Makefile
@@ -0,0 +1,20 @@
+APP = test
+PULP_APP = test
+PULP_APP_SRCS = test.c
+PULP_APP_SRCS += xpulp_tnn_conv_ternary_signed.c
+PULP_APP_SRCS += xpulp_tnn_matmul_ternary_signed.c
+PULP_APP_SRCS += xpulp_tnn_matmul_ternary_signed_4x1.c
+
+CORE=8
+
+
+PULP_CFLAGS += -DNUM_CORES=$(CORE) -I. -O3
+PULP_LDFLAGS += -lc -lm -lgcc -Wl,-print-memory-usage 
+
+PULP_APP_CFLAGS += -DNUM_CORES=$(CORE) -I. -O3
+PULP_APP_LDFLAGS += -lc -lm -lgcc -Wl,-print-memory-usage 
+
+PULP_CFLAGS += -DPULP_RUNTIME -DSINGLE_CORE_DMA -DALWAYS_BLOCK_DMA_TRANSFERS
+
+
+include $(PULP_SDK_HOME)/install/rules/pulp.mk
diff --git a/rt_nn_tests/xptnn_conv/data_statstest.h b/rt_nn_tests/xptnn_conv/data_statstest.h
new file mode 100644
index 0000000..e9baf5e
--- /dev/null
+++ b/rt_nn_tests/xptnn_conv/data_statstest.h
@@ -0,0 +1,1707 @@
+// this file is generated automatically by the python script `generate_statstest.py`
+
+#ifndef _DATA_H
+#define _DATA_H
+
+
+
+#define IM2COL_DIM (36 * NUM_CORES * 2) // for NUM_CORES cores
+#define OUTPUT_DIM 160
+#define OUTPUT_DIM_FP (1)
+uint8_t pWeight_0 [] = {
+0x8d,
+0x38,
+0x18,
+0xbf,
+0x48,
+0xe8,
+0x36,
+0xbd,
+0xa1,
+0x6d,
+0x4d,
+0xed,
+0xaf,
+0x30,
+0x04,
+0x02,
+0xec,
+0x6d,
+0x73,
+0x1c,
+0x33,
+0xc9,
+0x45,
+0x32,
+0x45,
+0x9c,
+0xff,
+0xbd,
+0x0e,
+0x16,
+0x29,
+0x1d,
+0x6a,
+0xb2,
+0xf5,
+0x28,
+0xde,
+0xc7,
+0x71,
+0xdf,
+0xc9,
+0xbc,
+0xfa,
+0x6b,
+0x9c,
+0xa2,
+0x24,
+0x41,
+0x9b,
+0x17,
+0x69,
+0x9a,
+0xfe,
+0xd8,
+0xcb,
+0xfb,
+0xbc,
+0x1f,
+0xbc,
+0xb8,
+0x93,
+0x60,
+0x24,
+0xd9,
+0xa7,
+0x6f,
+0xa0,
+0xe3,
+0x5d,
+0x08,
+0x0b,
+0x09,
+0x80,
+0xfd,
+0xda,
+0x5d,
+0xbb,
+0x12,
+0x47,
+0x7e,
+0xad,
+0x44,
+0x8b,
+0x6c,
+0xd7,
+0x58,
+0xaf,
+0xad,
+0x9b,
+0xb4,
+0x6b,
+0xe3,
+0xaf,
+0xfe,
+0x4b,
+0xd3,
+0x08,
+0xfd,
+0x57,
+0x65,
+0x33,
+0x1e,
+0xa7,
+0xe0,
+0xfa,
+0x2e,
+0x2e,
+0xc7,
+0x0d,
+0x90,
+0x1d,
+0xe3,
+0x22,
+0x1d,
+0x80,
+0x9c,
+0xc7,
+0x48,
+0x2e,
+0xeb,
+0x2c,
+0x07,
+0xe0,
+0xfa,
+0xef,
+0x9c,
+0x02,
+0xc6,
+0x88,
+0xe3,
+0x44,
+0x4d,
+0xa3,
+0xc0,
+0x76,
+0xeb,
+0xe0,
+0x27,
+0xb5,
+0xe4,
+0xeb,
+0xd3,
+0x12,
+0x06,
+0xc7,
+0xf7,
+0x3f,
+0x0f,
+0x45,
+0x57,
+0x57,
+0x49,
+0xd5,
+0xe9,
+0xc1,
+0xc0,
+0x4c,
+0xdf,
+0x3b,
+0xc1,
+0x65,
+0x2d,
+0x35,
+0x6b,
+0x76,
+0x40,
+0xa5,
+0x21,
+0xeb,
+0xb1,
+0xc3,
+0x6e,
+0x2c,
+0xf0,
+0x46,
+0x7e,
+0x58,
+0x2a,
+0x0f,
+0xcc,
+0xe1,
+0xad,
+0xd0,
+0x1d,
+0x5f,
+0x91,
+0x10,
+0x73,
+0xfa,
+0x70,
+0xe0,
+0xac,
+0x04,
+0x5e,
+0x1c,
+0x32,
+0xb8,
+0xea,
+0xc2,
+0x97,
+0xce,
+0xab,
+0xa3,
+0x5f,
+0x9b,
+0x76,
+0x15,
+0x47,
+0x36,
+0x55,
+0xb2,
+0x81,
+0x04,
+0xa8,
+0x4f,
+0x3e,
+0x0f,
+0x5f,
+0xb6,
+0x81,
+0x35,
+0x43,
+0xfa,
+0xcc,
+0x9b,
+0x61,
+0x6d,
+0xd9,
+0x6b,
+0x04,
+0xda,
+0x94,
+0x56,
+0x83,
+0x3d,
+0x6a,
+0xe2,
+0xb2,
+0x49,
+0x0d,
+0x48,
+0x02,
+0xda,
+0xa5,
+0x42,
+0x61,
+0xf4,
+0x6a,
+0xec,
+0x7e,
+0xf5,
+0x7d,
+0x40,
+0x0d,
+0x74,
+0x7e,
+0x28,
+0x01,
+0x80,
+0x05,
+0xa2,
+0xc3,
+0x40,
+0x2a,
+0x6d,
+0xaa,
+0x31,
+0xb6,
+0xc7,
+0x03,
+0x49,
+0xec,
+0x50,
+0x09,
+0x69,
+0x30,
+0xa6,
+0x18,
+0x70,
+0x71,
+0x2d,
+0x5f,
+0x8f,
+0xb6,
+0x60,
+0x70,
+0x1e,
+0x97,
+0x4d,
+0xef,
+0x87,
+0x2a,
+0x3a,
+0x5d,
+0xaf,
+0x08,
+0xc6,
+0x76,
+0x5b,
+0x93,
+0x8f,
+0x21,
+0x3c,
+0xa9,
+0x49,
+0x7d,
+0xa5,
+0xfe,
+0xee,
+0xaf,
+0xd0,
+0x67,
+0x54,
+0x2f,
+0x6b,
+0xaf,
+0x2e,
+0xbc,
+0x1b,
+0xac,
+0xc6,
+0xaf,
+0xa8,
+0xda,
+0x49,
+0x83,
+0xc3,
+0xbb,
+0x61,
+0xc0,
+0xe2,
+0x4b,
+0xeb,
+0x73,
+0xfd,
+0x3e,
+0xb4,
+0x9d,
+0xa8,
+0xcc,
+0xd7,
+0x5d,
+0x82,
+0x40,
+0xde,
+0xc4,
+0x82,
+0x2c,
+0x7f,
+0xf1,
+0x07,
+0x1a,
+0x9d,
+0xc6,
+0x9c,
+0xd4,
+0x49,
+0xde,
+0xde,
+0x74,
+0x17,
+0xe5,
+0xcd,
+0x2a,
+0x2f,
+0xdc,
+0xc7,
+0x3e,
+0x9b,
+0xdd,
+0xc7,
+0x20,
+0xfd,
+0xab,
+0x58,
+0x16,
+0xfe,
+0xc0,
+0x46,
+0x2a,
+0xa3,
+0x0a,
+0x01,
+0x38,
+0x15,
+0xa0,
+0xfc,
+0xd4,
+0x35,
+0x81,
+0xb3,
+0xfb,
+0x2f,
+0x40,
+0x18,
+0x9b,
+0xf1,
+0x17,
+0x23,
+0xd8,
+0x12,
+0x35,
+0x85,
+0xc1,
+0x53,
+0xca,
+0xd5,
+0x1c,
+0x3a,
+0x92,
+0x9f,
+0x1b,
+0x4c,
+0x1d,
+0x5a,
+0xe9,
+0x42,
+0x5c,
+0x20,
+0xe4,
+0xdb,
+0xdb,
+0x8a,
+0xed,
+0xaa,
+0xb2,
+0x89,
+0x1e,
+0x01,
+0x3d,
+0x8f,
+0x16,
+0x41,
+0xaf,
+0x96,
+0x2f,
+0x3d,
+0xbf,
+0xfe,
+0xc3,
+0x2c,
+0x4b,
+0xd4,
+0xad,
+0xf6,
+0xb1,
+0xc5,
+0x22,
+0x2a,
+0xaa,
+0x8c,
+0x47,
+0xa3,
+0xf3,
+0xe1,
+0x3c,
+0x35,
+0x15,
+0x34,
+0x17,
+0x84,
+0x2b,
+0x56,
+0x0a,
+0xfe,
+0x18,
+0x64,
+0xfa,
+0xe9,
+0xea,
+0xde,
+0x39,
+0xfc,
+0x3a,
+0x82,
+0xdf,
+0x0c,
+0x3f,
+0x17,
+0x24,
+0x9f,
+0x36,
+0x82,
+0xef,
+0x71,
+0xbd,
+0x2d,
+0x5d,
+0xe8,
+0x6f,
+0x78,
+0x07,
+0x37,
+0x5c,
+0xb2,
+0x31,
+0x42,
+0x0a,
+0x18,
+0x62,
+0x1d,
+0x06,
+0x27,
+0x9a,
+0x09,
+0xa1,
+0x6b,
+0xad,
+0xb7,
+0x33,
+0x07,
+0xc0,
+0x1a,
+0x7d,
+0xc1,
+0x75,
+0xba,
+0x5e,
+0xdd,
+0xe0,
+0x3a,
+0x49,
+0xd1,
+0x2d,
+0x29,
+0x1d,
+0xd9,
+0xef,
+0xe7,
+0x69,
+0xe9,
+0x41,
+0x10,
+0xb1,
+0x1f,
+0x95,
+0x34,
+0xd6,
+0xa4,
+0x1b,
+0x1b,
+0x36,
+0x24,
+0xb3,
+0x55,
+0xc5,
+0x3b,
+0x85,
+0x7a,
+0xc9,
+0x30,
+0x86,
+0x0d,
+0x28,
+0x15,
+0x1c,
+0x9f,
+0x3e,
+0x56,
+0x12,
+0x68,
+0xf0,
+0xd6,
+0x6d,
+0x4b,
+0x2b,
+0xa5,
+0xd5,
+0xc0,
+0x41,
+0x90,
+0x01,
+0xc4,
+0x30,
+0x66,
+0x8c,
+0x47,
+0x7f,
+0x62,
+0xa3,
+0xd0,
+0x70,
+0x63,
+0x5f,
+0x47,
+0x27,
+0x85,
+0x22,
+0xa1,
+0xaf,
+0x02,
+0x51,
+0x76,
+0x73,
+0x07,
+0x38,
+0xe7,
+0x97,
+0xcc,
+0xa2,
+0xba,
+0x8c,
+0x05,
+0x7c,
+0xe7,
+0xd1,
+0xaf,
+0xe3,
+0x88,
+0x00,
+0xac,
+0xf8,
+0x03,
+0x6c,
+0xc9,
+0x80,
+0x16,
+0x39,
+0x48,
+0x29,
+0x3c,
+0xf8,
+0xe0,
+0x27,
+0xed,
+0x24,
+0x2b,
+0x4c,
+0x3c,
+0x1c,
+0x2e,
+0x07,
+0x63,
+0xca,
+0x0b,
+0x10,
+0xc3,
+0xcc,
+0xbf,
+0x49,
+0x47,
+0x31,
+0x5d,
+0xd0,
+0x11,
+0x86,
+0xc8,
+0x9f,
+0xa5,
+0x11,
+0xe4,
+0x02,
+0x20,
+0x30,
+0xf1,
+0x83,
+0x31,
+0x88,
+0xf0,
+0x07,
+0x13,
+0x21,
+0xb3,
+0x4b,
+0x0d,
+0x01,
+0xf0,
+0x35,
+0x77,
+0x92,
+0x63,
+0xad,
+0x1f,
+0x9a,
+0x68,
+0x03,
+0xcf,
+0xfe,
+0x7b,
+0x05,
+0x29,
+0xde,
+0x66,
+0x39,
+0xa3,
+0x0b,
+0xf1,
+0xe2,
+0x98,
+0xcb,
+0x01,
+0x30,
+0x32,
+0x20,
+0x57,
+0x03,
+0xb1,
+0x41,
+0xb0,
+0xee,
+0xff,
+0xad,
+0x7c,
+0x4a,
+0x60,
+0xf3,
+0xe1,
+0x30,
+0xf1,
+0xea,
+0x41,
+0x9f,
+0xac,
+0x63,
+0x58,
+0x3a,
+0xf1,
+0xf0,
+0x03,
+0xa1,
+0x3e,
+0x1a,
+0x2a,
+0x2d,
+0x40,
+0xa5,
+0x29,
+0x50,
+0xb5,
+0xf5,
+0xe6,
+0xa9,
+0x7f,
+0x2c,
+0xf1,
+0x09,
+0x44,
+0xee,
+0xe0,
+0x12,
+0xbf,
+0x6b,
+0x5d,
+0x38,
+0x36,
+0x73,
+0x7b,
+0xad,
+0x82,
+0x74,
+0x04,
+0xaf,
+0xd1,
+0x0a,
+0x37,
+0xa0,
+0x3b,
+0x5d,
+0x88,
+0x08,
+0xcd,
+0xfe,
+0x86,
+0x45,
+0x31,
+0x6a,
+0xe8,
+0x8b,
+0xfa,
+0xfa,
+0xd8,
+0x54,
+0x8d,
+0x04,
+0x36,
+0xf0,
+0x51,
+0x1a,
+0xa6,
+0xb7,
+0x8d,
+0xea,
+0x54,
+0xa8,
+0xd7,
+0x86,
+0xf1,
+0x3b,
+0x5e,
+0x1f,
+0xd4,
+0x63,
+0x15,
+0x28,
+0xe4,
+0x04,
+0x25,
+0x34,
+0x92,
+0x0e,
+0x2e,
+0xe8,
+0x88,
+0x2a,
+0x40,
+0x55,
+0x25,
+0x8f,
+0x5c,
+0x4a,
+0x4c,
+0x64,
+0x29,
+0xe6,
+0xa3,
+0xc4,
+0x44,
+0x02,
+0xfc,
+0xa3,
+0xc4,
+0x37,
+0xfe,
+0x77,
+0xc1,
+0xf3,
+0x03,
+0x9b,
+0xb1,
+0x3f,
+0xfe,
+0x05,
+0x89,
+0xa9,
+0xaa,
+0x51,
+0xd1,
+0x82,
+0x20,
+0x5e,
+0x7e,
+0xa6,
+0xf8,
+0x92,
+0x60,
+0x60,
+0xab,
+0x85,
+0x24,
+0xd6,
+0xe5,
+0xd0,
+0x16,
+0x22,
+0xcf,
+0xc3,
+0x25,
+0xdc,
+0xaa,
+0x5a,
+0x46,
+0x02,
+0xc6,
+0x71,
+0x1e,
+0x01,
+0xde,
+0xf7,
+0x56,
+0x0d,
+0x4b,
+0x60,
+0x6a,
+0xfd,
+0x08,
+0x6c,
+0x36,
+0x5b,
+0x02,
+0xb8,
+0xd6,
+0x92,
+0xf1,
+0x4f,
+0x2b,
+0xd8,
+0x6c,
+0x26,
+0xf1,
+0x73,
+0xdf,
+0x4b,
+0xe8,
+0x3b,
+0x32,
+0x73,
+0xd7,
+0xe1,
+0x77,
+0x0e,
+0x4f,
+0x44,
+0x26,
+0x60,
+0x43,
+0x08,
+0x67,
+0x3c,
+0x01,
+0x2b,
+0x97,
+0x4a,
+0x4e,
+0x0d,
+0x7a,
+0x8d,
+0x3a,
+0xf5,
+0xb8,
+0xdb,
+0x87,
+0xb1,
+0xa7,
+0xec,
+0x49,
+0x8d,
+0x1a,
+0xe9,
+0x3b,
+0x05,
+0x7a,
+0x47,
+0xff,
+0x68,
+0x5d,
+0x98,
+0x1a,
+0xd0,
+0xec,
+0xde,
+0x89,
+0x23,
+0x82,
+0xc9,
+0x23,
+0x26,
+0x1f,
+0x97,
+0xbb,
+0x4f,
+0xfe,
+0xc6,
+0xdc,
+0x07,
+0xe0,
+0x6b,
+0x3d,
+0x0e,
+0x1c,
+0xc7,
+0x2e,
+0x4c,
+0x01,
+0x0b,
+0xe2,
+0x14,
+0x8f,
+0x63,
+0x2c,
+0x03,
+0x98,
+0xa8,
+0x46,
+0x29,
+0x3f,
+0x65,
+0x9a,
+0xa2,
+0xe3,
+0x95,
+0x91,
+0xe4,
+0x23,
+0x16,
+0xfd,
+0x5d,
+0x9f,
+0xde,
+0x1c,
+0xdc,
+0x60,
+0x9a,
+0xba,
+0x1c,
+0x6f,
+0x07,
+0xa6,
+0xac,
+0x30,
+0xe8,
+0x6e,
+0x71,
+0xb0,
+0xc3,
+0x1a,
+0x92,
+0xe8,
+0x4a,
+0x4c,
+0x3c,
+0x8c,
+0x77,
+0xcb,
+0xf3,
+0x4a,
+0x14,
+0x21,
+0x17,
+0xbb,
+0x73,
+0x38,
+0x7f,
+0xa2,
+0x0a,
+0xcc,
+0xd6,
+0xd7,
+0x23,
+0x07,
+0x1e,
+0x38,
+0xca,
+0x32,
+0xd4,
+0xfb,
+0xb6,
+0x09,
+0x5a,
+0x03,
+0x77,
+0x54,
+0xdd,
+0xd8,
+0x9a,
+0xab,
+0x8f,
+0x51,
+0x54,
+0xa3,
+0x14,
+0x18,
+0xc0,
+0x86,
+0xe4,
+0xb8,
+0x66,
+0xc0,
+0x6d,
+0x8b,
+0xa9,
+0x40,
+0xb3,
+0xf7,
+0x0e,
+0xd3,
+0x27,
+0xe9,
+0x96,
+0xa4,
+0xeb,
+0x0b,
+0x7b,
+0x0e,
+0x74,
+0x9d,
+0x28,
+0xb9,
+0x6b,
+0xfe,
+0xcd,
+0x0b,
+0x01,
+0xf4,
+0xad,
+0x57,
+0x5b,
+0xc5,
+0xad,
+0xc9,
+0xba,
+0xf7,
+0x83,
+0x13,
+0xa9,
+0xf4,
+0x85,
+0x9a,
+0x2a,
+0x6a,
+0x8c,
+0xc4,
+0x24,
+0xfc,
+0x4b,
+0x70,
+0xca,
+0xc7,
+0x58,
+0xb8,
+0x3b,
+0xa1,
+0xd0,
+0x2c,
+0xea,
+0xe9,
+0xe1,
+0x50,
+0x6a,
+0x60,
+0x2d,
+0xd5,
+0xa5,
+0xe4,
+0xce,
+0xcd,
+0xb8,
+0xf3,
+0x42,
+0x43,
+0xc9,
+0x05,
+0x97,
+0xea,
+0xfc,
+0x36,
+0xba,
+0x04,
+0x14,
+0x86,
+0x75,
+0x2b,
+0xc6,
+0x56,
+0x3f,
+0x88,
+0x45,
+0xfa,
+0x1b,
+0xb5,
+0x45,
+0x8a,
+0xa0,
+0x0b,
+0x50,
+0x80,
+0x58,
+0x1f,
+0x9d,
+0x54,
+0xe6,
+0x87,
+0x30,
+0x6e,
+0x0d,
+0x1b,
+0xe6,
+0x91,
+0xa5,
+0xf7,
+0x7a,
+0x7e,
+0x73,
+0x47,
+0x23,
+0x92,
+0xe1,
+0x9b,
+0xd4,
+0xa5,
+0x1e,
+0x08,
+0x17,
+0x4c,
+0xc0,
+0x74,
+0xe3,
+0x45,
+0x8d,
+0x25,
+0x74,
+0x66,
+0x8b,
+0x76,
+0x1a,
+0x0a,
+0x49,
+0x1a,
+0xbd,
+0x65,
+0xc9,
+0x49,
+0xea,
+0x10,
+0x5e,
+0xe7,
+0x5b,
+0xf0,
+0x53,
+0x76,
+0x08,
+0x5b,
+0xcd,
+0x42,
+0xc4,
+0xb5,
+0x1d,
+0x55,
+0x01,
+0x46,
+0x73,
+0xdb,
+0x6b,
+0xee,
+0xc7,
+0xd4,
+0xdf,
+0x61,
+0x61,
+0xb0,
+0x71,
+0x84,
+0xa6,
+0xa8,
+0x86,
+0xa3,
+0x35,
+0xd8,
+0x09,
+0x35,
+0x31,
+0x4c,
+0xa2,
+0xb2,
+0x0f,
+0x1b,
+0xd5,
+0x88,
+0x11,
+0x5b,
+0x69,
+0xe8,
+0x5f,
+0x1b,
+0xe3,
+0xa5,
+0xa8,
+0x82,
+0xc2,
+0x16,
+0xea,
+0xc8,
+0x33,
+0x0e,
+0xcc,
+0xb7,
+0x1a,
+0x34,
+0x61,
+0xdf,
+0x57,
+0xdb,
+0x46,
+0xbb,
+0xcc,
+0x98,
+0xa3,
+0xda,
+0x7e,
+0xfe,
+0xa7,
+0x58,
+0x3d,
+0x5a,
+0xe5,
+0x16,
+0xd9,
+0xc7,
+0xfc,
+0x0f,
+0x46,
+0x23,
+0x60,
+0x04,
+0x9c,
+0xd5,
+0x11,
+0x43,
+0xc0,
+0x6a,
+0xf3,
+0x35,
+0x7a,
+0x84,
+0xa7,
+0x91,
+0x1d,
+0xd6,
+0x91,
+0xe3,
+0xfb,
+0x73,
+0x88,
+0xcf,
+0x9a,
+0xd4,
+0x5e,
+0xad,
+0xb3,
+0xdd,
+0x2b,
+0xf1,
+0x65,
+0x61,
+0xe0,
+0xfe,
+0x00,
+0x44,
+0x76,
+0x0d,
+0xd9,
+0xfc,
+0x03,
+0x75,
+0x37,
+0x81,
+0x91,
+0x5c,
+0x3c,
+0xfc,
+0x63,
+0x90,
+0x48,
+0xef,
+0x14,
+0xf8,
+0xd7,
+0xdf,
+0xf4,
+0x6c,
+0x13,
+0x20,
+0xd9,
+0x3e,
+0x2a,
+0x40,
+0x2c,
+0x49,
+0x3d,
+0x5e,
+0xdd,
+0xf8,
+0x11,
+0x84,
+0x7c,
+0x3d,
+0x1a,
+0x5e,
+0xc5,
+0x7d,
+0x32,
+0xa5,
+0xab,
+0xd4,
+0x69,
+0x08,
+0x86,
+0xa8,
+0x74,
+0x74,
+0x25,
+0x68,
+0x4b,
+0x24,
+0x40,
+0x4d,
+0xcb,
+0xf7,
+0x05,
+0x09,
+0xc2,
+0x39,
+0xa9,
+0xdd,
+0x00,
+0xc8,
+0x94,
+0x0a,
+0x5d,
+0x13,
+0x8d,
+0xa5,
+0xaa,
+0xff,
+0x29,
+0x36,
+0xfc,
+0x58,
+0x6d,
+0x6a,
+0x7e,
+0xab,
+0xf0,
+0x86,
+0x6f,
+0xc2,
+0x45,
+0x87,
+0xa7,
+0xb8,
+0xb5,
+0x16,
+0x9d,
+0xea,
+0x2a,
+0x25,
+0xb0,
+0x60,
+0xe9,
+0xb5,
+0x06,
+0x17,
+0x62,
+0x64,
+0xf7,
+0xcb,
+0x84,
+0x7c,
+0xb6,
+0xa0,
+0x26
+};
+
+uint8_t pIn_0 [] = {
+0x05,
+0xaa,
+0xc9,
+0x2c,
+0x10,
+0x4f,
+0x05,
+0x8a,
+0x9b,
+0xbf,
+0xc9,
+0xed,
+0x5a,
+0x84,
+0x4a,
+0x85,
+0x1d,
+0x18,
+0xe1,
+0x88,
+0xa4,
+0x4c,
+0x33,
+0x9c,
+0x65,
+0xf3,
+0x2f,
+0x58,
+0x0a,
+0xf6,
+0xa5,
+0xcd,
+0x5e,
+0x01,
+0x33,
+0xe6,
+0x5c,
+0xab,
+0x9d,
+0x09,
+0xc4,
+0x7d,
+0x95,
+0x57,
+0xeb,
+0x3d,
+0x01,
+0x17,
+0x86,
+0x8a,
+0xbd,
+0x1d,
+0x97,
+0x84,
+0x75,
+0x46,
+0xa7,
+0xfe,
+0xf6,
+0x69,
+0xca,
+0x41,
+0xc2,
+0xdc
+};
+
+uint32_t pThr_0 [] = {
+0xff63ff84,
+0xff7dff97,
+0xffa5ffbf,
+0x00350065,
+0xff6dff7a,
+0xff5fffa1,
+0x001f0020,
+0x000d0051,
+0x003c004d,
+0xff6fffa3,
+0x007800cf,
+0x00240061,
+0x0025003e,
+0x00150040,
+0xffc6fff4,
+0x00890099,
+0xff51ff92,
+0x00620063,
+0xffda0023,
+0x001e004e,
+0x00550089,
+0xff5cff9e,
+0xffd4fff9,
+0x0039003d,
+0xff99ffa4,
+0x000c0038,
+0xffc2fff1,
+0x003c005e,
+0x0079009a,
+0xff8bffc3,
+0xffe7001e,
+0x00560060,
+0xff5eff80,
+0x009800ae,
+0xff6eff93,
+0x00930094,
+0x009800ef,
+0xffcc0008,
+0x000c0031,
+0x00430095
+};
+
+
+/* -----------EXPECTED OUTPUTS----------- */
+uint8_t exp_outp_0 [] = {
+0x47,
+0x77,
+0x7f,
+0xd0,
+0x4f,
+0x6f,
+0x53,
+0xf4,
+0x47,
+0x77,
+0x7f,
+0xd0,
+0x4f,
+0x6f,
+0x53,
+0xf4,
+0x47,
+0x77,
+0x7f,
+0xd0,
+0x4f,
+0x6f,
+0x53,
+0xf4,
+0x47,
+0x77,
+0x7f,
+0xd0,
+0x4f,
+0x6f,
+0x53,
+0xf4,
+0x47,
+0x77,
+0x7f,
+0xd0,
+0x4f,
+0x6f,
+0x53,
+0xf4,
+0x47,
+0x77,
+0x7f,
+0xd0,
+0x4f,
+0x6f,
+0x53,
+0xf4,
+0x47,
+0x77,
+0x7f,
+0xd0,
+0x4f,
+0x6f,
+0x53,
+0xf4,
+0x47,
+0x77,
+0x7f,
+0xd0,
+0x4f,
+0x6f,
+0x53,
+0xf4,
+0x47,
+0x77,
+0x7f,
+0xd0,
+0x4f,
+0x6f,
+0x53,
+0xf4,
+0x47,
+0x77,
+0x7f,
+0xd0,
+0x4f,
+0x6f,
+0x53,
+0xbc,
+0x47,
+0x77,
+0x7f,
+0xd0,
+0x4f,
+0x6f,
+0x53,
+0xf4,
+0x47,
+0x77,
+0x7f,
+0xd0,
+0x4f,
+0x6f,
+0x53,
+0xf4,
+0x47,
+0x77,
+0x7f,
+0xd0,
+0x4f,
+0x6f,
+0x53,
+0xdf,
+0x47,
+0x77,
+0x7f,
+0xd0,
+0x4f,
+0x63,
+0x53,
+0xf4,
+0x47,
+0x77,
+0x7f,
+0xd0,
+0x4f,
+0x6f,
+0x53,
+0xbc,
+0x47,
+0x77,
+0x7f,
+0xd0,
+0x4f,
+0x6f,
+0x53,
+0xf4
+};
+
+
+
+L1_DATA uint32_t threshs_l1 [40] = {0};
+L1_DATA int32_t kappa_l1 [40] = {0};
+L1_DATA int32_t lambda_l1 [40] = {0};
+L1_DATA uint8_t inp_l1 [64] = {0};
+L1_DATA uint8_t outp_l1 [128] = {0};
+L1_DATA int8_t wt_l1 [1440] = {0};
+L1_DATA uint8_t im2col_l1 [IM2COL_DIM] = {0};
+
+#endif
diff --git a/rt_nn_tests/xptnn_conv/pulp_nn_kernels.h b/rt_nn_tests/xptnn_conv/pulp_nn_kernels.h
new file mode 100644
index 0000000..f4ec977
--- /dev/null
+++ b/rt_nn_tests/xptnn_conv/pulp_nn_kernels.h
@@ -0,0 +1,26528 @@
+/*
+ * pulp_nn_kernels.h
+ * Nazareno Bruschi <nazareno.bruschi@unibo.it>
+ *
+ * Copyright (C) 2019-2020 University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef __XPULPNN_KERNELS__
+#define __XPULPNN_KERNELS__
+
+void xpulp_nn_conv_u8_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+#ifndef PROFILE
+                        uint8_t flag_batch_norm);
+#else
+  uint8_t flag_batch_norm,
+    uint32_t *im2col_cycles,
+    uint32_t *hotloop_prep_cycles,
+    uint32_t *hotloop_cycles,
+    uint32_t *requant_cycles,
+    uint32_t *hotloop_leftover_cycles,
+    uint32_t *matmul4x2_leftover_cycles);
+#endif
+                        
+
+void xpulp_nn_conv_i2_i2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_i8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_i8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_i8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_i4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_i4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_i4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_i2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_i2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_i2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_i8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_i8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_i8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_i4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_i4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_i4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_i2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_i2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_i2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_i8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_i8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_i8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_i4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_i4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_i4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_i2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_i2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_i2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u8_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i8_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u8_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i8_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u8_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i8_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u4_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i4_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u4_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i4_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u4_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i4_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u2_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i2_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u2_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i2_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u2_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i2_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u8_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i8_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u8_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i8_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u8_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i8_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u4_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i4_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u4_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i4_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u4_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i4_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u2_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i2_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u2_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i2_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u2_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i2_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u8_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i8_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u8_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i8_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u8_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i8_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u4_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i4_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u4_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i4_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u4_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i4_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u2_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i2_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u2_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i2_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+#ifndef PROFILE
+                        uint8_t flag_batch_norm);
+#else
+  uint8_t flag_batch_norm,
+    uint32_t *hotloop_prep_cycles,
+    uint32_t *hotloop_cycles,
+    uint32_t *requant_cycles,
+    uint32_t *hotloop_leftover_cycles);
+#endif
+
+uint8_t *xpulp_nn_matmul_u2_u2_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i2_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_i8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_i8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_i8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_i4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_i4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_i4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_i2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_i2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_i2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_i8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_i8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_i8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_i4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_i4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_i4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_i2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_i2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_i2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_i8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_i8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_i8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_i4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_i4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_i4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_i2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_i2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_i2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_i32_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_i8_i32_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_u8_i32_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_i8_i32_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_u8_i32_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_i8_i32_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_u4_i32_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_i4_i32_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_u4_i32_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_i4_i32_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_u4_i32_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_i4_i32_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_u2_i32_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_i2_i32_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_u2_i32_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_i2_i32_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_u2_i32_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_i2_i32_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_u8_u8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_i8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_u8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_i8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_u8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_i8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_u4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_i4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_u4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_i4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_u4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_i4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_u2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_i2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_u2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_i2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_u2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_i2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_u8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_i8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_u8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_i8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_u8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_i8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_u4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_i4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_u4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_i4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_u4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_i4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_u2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_i2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_u2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_i2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_u2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_i2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_u8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_i8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_u8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_i8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_u8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_i8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_u4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_i4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_u4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_i4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_u4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_i4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_u2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_i2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_u2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_i2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_u2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_i2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_maxpool_u8(
+                        uint8_t * pIn,
+                        uint8_t * pOut,
+                        uint16_t  dim_im_in_x,
+                        uint16_t  dim_im_in_y,
+                        uint16_t  ch_im_in,
+                        uint16_t  dim_im_out_x,
+                        uint16_t  dim_im_out_y,
+                        uint16_t  dim_kernel_x,
+                        uint16_t  dim_kernel_y,
+                        uint16_t  padding_t,
+                        uint16_t  padding_b,
+                        uint16_t  padding_l,
+                        uint16_t  padding_r,
+                        uint16_t  stride_x,
+                        uint16_t  stride_y);
+
+void xpulp_nn_maxpool_i8(
+                        int8_t * pIn,
+                        int8_t * pOut,
+                        uint16_t  dim_im_in_x,
+                        uint16_t  dim_im_in_y,
+                        uint16_t  ch_im_in,
+                        uint16_t  dim_im_out_x,
+                        uint16_t  dim_im_out_y,
+                        uint16_t  dim_kernel_x,
+                        uint16_t  dim_kernel_y,
+                        uint16_t  padding_t,
+                        uint16_t  padding_b,
+                        uint16_t  padding_l,
+                        uint16_t  padding_r,
+                        uint16_t  stride_x,
+                        uint16_t  stride_y);
+
+void xpulp_nn_maxpool_u4(
+                        uint8_t * pIn,
+                        uint8_t * pOut,
+                        uint16_t  dim_im_in_x,
+                        uint16_t  dim_im_in_y,
+                        uint16_t  ch_im_in,
+                        uint16_t  dim_im_out_x,
+                        uint16_t  dim_im_out_y,
+                        uint16_t  dim_kernel_x,
+                        uint16_t  dim_kernel_y,
+                        uint16_t  padding_t,
+                        uint16_t  padding_b,
+                        uint16_t  padding_l,
+                        uint16_t  padding_r,
+                        uint16_t  stride_x,
+                        uint16_t  stride_y);
+
+void xpulp_nn_maxpool_i4(
+                        int8_t * pIn,
+                        int8_t * pOut,
+                        uint16_t  dim_im_in_x,
+                        uint16_t  dim_im_in_y,
+                        uint16_t  ch_im_in,
+                        uint16_t  dim_im_out_x,
+                        uint16_t  dim_im_out_y,
+                        uint16_t  dim_kernel_x,
+                        uint16_t  dim_kernel_y,
+                        uint16_t  padding_t,
+                        uint16_t  padding_b,
+                        uint16_t  padding_l,
+                        uint16_t  padding_r,
+                        uint16_t  stride_x,
+                        uint16_t  stride_y);
+
+void xpulp_nn_maxpool_u2(
+                        uint8_t * pIn,
+                        uint8_t * pOut,
+                        uint16_t  dim_im_in_x,
+                        uint16_t  dim_im_in_y,
+                        uint16_t  ch_im_in,
+                        uint16_t  dim_im_out_x,
+                        uint16_t  dim_im_out_y,
+                        uint16_t  dim_kernel_x,
+                        uint16_t  dim_kernel_y,
+                        uint16_t  padding_t,
+                        uint16_t  padding_b,
+                        uint16_t  padding_l,
+                        uint16_t  padding_r,
+                        uint16_t  stride_x,
+                        uint16_t  stride_y);
+
+void xpulp_nn_maxpool_i2(
+                        int8_t * pIn,
+                        int8_t * pOut,
+                        uint16_t  dim_im_in_x,
+                        uint16_t  dim_im_in_y,
+                        uint16_t  ch_im_in,
+                        uint16_t  dim_im_out_x,
+                        uint16_t  dim_im_out_y,
+                        uint16_t  dim_kernel_x,
+                        uint16_t  dim_kernel_y,
+                        uint16_t  padding_t,
+                        uint16_t  padding_b,
+                        uint16_t  padding_l,
+                        uint16_t  padding_r,
+                        uint16_t  stride_x,
+                        uint16_t  stride_y);
+
+void xpulp_nn_avgpool_u8_u8(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u8_i8(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i8_u8(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i8_i8(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u8_u4(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u8_i4(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i8_u4(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i8_i4(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u8_u2(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u8_i2(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i8_u2(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i8_i2(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u4_u8(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u4_i8(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i4_u8(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i4_i8(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u4_u4(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u4_i4(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i4_u4(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i4_i4(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u4_u2(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u4_i2(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i4_u2(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i4_i2(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u2_u8(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u2_i8(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i2_u8(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i2_i8(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u2_u4(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u2_i4(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i2_u4(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i2_i4(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u2_u2(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u2_i2(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i2_u2(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i2_i2(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_add_u8_u8(
+                        uint8_t * pIn1,
+                        uint8_t * pIn2,
+                        uint8_t * pOut,
+                        uint16_t out_mult1,
+                        uint16_t out_mult2,
+                        uint16_t out_shift,
+                        uint16_t dim_im_in_x,
+                        uint16_t dim_im_in_y,
+                        uint16_t ch_im_in);
+
+void xpulp_nn_add_u8_u8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u4(
+                        uint8_t * pIn1,
+                        uint8_t * pIn2,
+                        uint8_t * pOut,
+                        uint16_t out_mult1,
+                        uint16_t out_mult2,
+                        uint16_t out_shift,
+                        uint16_t dim_im_in_x,
+                        uint16_t dim_im_in_y,
+                        uint16_t ch_im_in);
+
+void xpulp_nn_add_u8_u4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u2(
+                        uint8_t * pIn1,
+                        uint8_t * pIn2,
+                        uint8_t * pOut,
+                        uint16_t out_mult1,
+                        uint16_t out_mult2,
+                        uint16_t out_shift,
+                        uint16_t dim_im_in_x,
+                        uint16_t dim_im_in_y,
+                        uint16_t ch_im_in);
+
+void xpulp_nn_add_u8_u2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u4(
+                        uint8_t * pIn1,
+                        uint8_t * pIn2,
+                        uint8_t * pOut,
+                        uint16_t out_mult1,
+                        uint16_t out_mult2,
+                        uint16_t out_shift,
+                        uint16_t dim_im_in_x,
+                        uint16_t dim_im_in_y,
+                        uint16_t ch_im_in);
+
+void xpulp_nn_add_u4_u4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u2(
+                        uint8_t * pIn1,
+                        uint8_t * pIn2,
+                        uint8_t * pOut,
+                        uint16_t out_mult1,
+                        uint16_t out_mult2,
+                        uint16_t out_shift,
+                        uint16_t dim_im_in_x,
+                        uint16_t dim_im_in_y,
+                        uint16_t ch_im_in);
+
+void xpulp_nn_add_u4_u2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u2(
+                        uint8_t * pIn1,
+                        uint8_t * pIn2,
+                        uint8_t * pOut,
+                        uint16_t out_mult1,
+                        uint16_t out_mult2,
+                        uint16_t out_shift,
+                        uint16_t dim_im_in_x,
+                        uint16_t dim_im_in_y,
+                        uint16_t ch_im_in);
+
+void xpulp_nn_add_u2_u2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+
+
+#endif
diff --git a/rt_nn_tests/xptnn_conv/pulp_nn_mix_kernels.h b/rt_nn_tests/xptnn_conv/pulp_nn_mix_kernels.h
new file mode 100644
index 0000000..8b2a3c6
--- /dev/null
+++ b/rt_nn_tests/xptnn_conv/pulp_nn_mix_kernels.h
@@ -0,0 +1,7093 @@
+/*
+ * pulp_nn_kernels.h
+ * Nazareno Bruschi <nazareno.bruschi@unibo.it>
+ *
+ * Copyright (C) 2019-2020 University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef __XPULPNN-MIXED_KERNELS__
+#define __XPULPNN-MIXED_KERNELS__
+
+void xpulp_nn_mix_conv_u8_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u8_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u8_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u8_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u4_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u4_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u4_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u2_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u2_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u2_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u8_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u8_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u8_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u4_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u4_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u4_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u2_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u2_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u2_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u8_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u8_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u8_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u4_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u4_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u4_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u2_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u2_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u2_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_avgpool_u8_u8(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u8_i8(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i8_u8(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i8_i8(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u8_u4(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u8_i4(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i8_u4(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i8_i4(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u8_u2(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u8_i2(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i8_u2(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i8_i2(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u4_u8(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u4_i8(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i4_u8(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i4_i8(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u4_u4(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u4_i4(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i4_u4(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i4_i4(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u4_u2(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u4_i2(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i4_u2(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i4_i2(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u2_u8(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u2_i8(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i2_u8(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i2_i8(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u2_u4(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u2_i4(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i2_u4(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i2_i4(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u2_u2(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u2_i2(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i2_u2(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i2_i2(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_add_u8_u8(
+                        uint8_t * pIn1,
+                        uint8_t * pIn2,
+                        uint8_t * pOut,
+                        uint16_t out_mult1,
+                        uint16_t out_mult2,
+                        uint16_t out_shift,
+                        uint16_t dim_im_in_x,
+                        uint16_t dim_im_in_y,
+                        uint16_t ch_im_in);
+
+void xpulp_nn_add_u8_u8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u4(
+                        uint8_t * pIn1,
+                        uint8_t * pIn2,
+                        uint8_t * pOut,
+                        uint16_t out_mult1,
+                        uint16_t out_mult2,
+                        uint16_t out_shift,
+                        uint16_t dim_im_in_x,
+                        uint16_t dim_im_in_y,
+                        uint16_t ch_im_in);
+
+void xpulp_nn_add_u8_u4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u2(
+                        uint8_t * pIn1,
+                        uint8_t * pIn2,
+                        uint8_t * pOut,
+                        uint16_t out_mult1,
+                        uint16_t out_mult2,
+                        uint16_t out_shift,
+                        uint16_t dim_im_in_x,
+                        uint16_t dim_im_in_y,
+                        uint16_t ch_im_in);
+
+void xpulp_nn_add_u8_u2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u4(
+                        uint8_t * pIn1,
+                        uint8_t * pIn2,
+                        uint8_t * pOut,
+                        uint16_t out_mult1,
+                        uint16_t out_mult2,
+                        uint16_t out_shift,
+                        uint16_t dim_im_in_x,
+                        uint16_t dim_im_in_y,
+                        uint16_t ch_im_in);
+
+void xpulp_nn_add_u4_u4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u2(
+                        uint8_t * pIn1,
+                        uint8_t * pIn2,
+                        uint8_t * pOut,
+                        uint16_t out_mult1,
+                        uint16_t out_mult2,
+                        uint16_t out_shift,
+                        uint16_t dim_im_in_x,
+                        uint16_t dim_im_in_y,
+                        uint16_t ch_im_in);
+
+void xpulp_nn_add_u4_u2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u2(
+                        uint8_t * pIn1,
+                        uint8_t * pIn2,
+                        uint8_t * pOut,
+                        uint16_t out_mult1,
+                        uint16_t out_mult2,
+                        uint16_t out_shift,
+                        uint16_t dim_im_in_x,
+                        uint16_t dim_im_in_y,
+                        uint16_t ch_im_in);
+
+void xpulp_nn_add_u2_u2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+
+
+#endif
\ No newline at end of file
diff --git a/rt_nn_tests/xptnn_conv/pulp_nn_utils.h b/rt_nn_tests/xptnn_conv/pulp_nn_utils.h
new file mode 100644
index 0000000..44d2251
--- /dev/null
+++ b/rt_nn_tests/xptnn_conv/pulp_nn_utils.h
@@ -0,0 +1,2079 @@
+/*
+ * pulp_nn_utils.h
+ * Nazareno   Bruschi  <nazareno.bruschi@unibo.it>
+ * Alessandro Nadalini <alessandro.nadalini3@unibo.it>
+ * Georg Rutishauser   <georgr@iis.ee.ethz.ch>
+ *
+ * Copyright (C) 2019-2020 ETH Zurich & University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef __PULPNN_UTILS__
+#define __PULPNN_UTILS__
+
+#include <stdint.h>
+
+typedef   signed short v2s __attribute__((vector_size (4)));
+
+
+
+#define min(a,b)                                             ((a)<(b)?(a):(b))
+#define log2(x)                                              __builtin_pulp_fl1(x)
+#define PACK_INT2_SIZE(x)                                    ((x) >> 2)
+
+#define MemoryFence()                                        asm volatile("":::"memory")
+
+#define CHANS_DECOMPR(x)                                     (5*x >> 2) // equivalent to division by 0.8
+
+/* Functions for Compressed MAC */
+#define CompressedMAC(sum, ptr, config) asm volatile(                \
+    "pv.smlsdotsp.t %[shum], %[phtr], %[chonfig];"                    \
+    : [shum] "+r" (sum), [phtr] "+r" (ptr): [chonfig] "I" (config))
+
+#define CompressedMACUnsigned(sum, ptr, config) asm volatile(                   \
+    "pv.smlsdotsup.t %[shum], %[phtr], %[chonfig];" \
+    : [shum] "+r" (sum), [phtr] "+r" (ptr): [chonfig] "I" (config))
+
+#define InitNNRF(ptr, config) asm volatile(        \
+    "pv.smlsdotsp.t x0, %[phtr], %[chonfig];"       \
+    : [phtr] "+r" (ptr) : [chonfig] "I" (config))
+
+#define ThresholdCompress(res, val, thrs) asm volatile(                                                \
+    "pv.thrc %[rhes], %[vhal], %[thhrs];" : [rhes] "+r" (res) : [vhal] "r" (val), [thhrs] "r" (thrs))
+
+#define GetConfig(a_update, b_update, a_reg, b_reg) a_update << 4 | b_update << 3 | a_reg << 1 | b_reg
+
+/* Functions for threshold&compress */
+#define check_store(res, pOut)            \
+  if ((res & 0xe0000000) == 0x00000000) { \
+    *pOut = res & 0xff;                   \
+    pOut++;                               \
+    incr_val=ch_out_r; }
+
+#define check_store_4x1(res, pOut)        \
+  if ((res & 0xe0000000) == 0x00000000) { \
+    *pOut = res & 0xff;                   \
+    pOut++; }
+
+#define reset_currThr()                           \
+  if ((uint32_t *) currThr == (uint32_t *) (pThr + ch_out)) { \
+    currThr = (v2s *) pThr;                                    \
+  }
+
+#define MacLoads20(a_update, b_update, a_reg, b_reg, ptr, sum)  __builtin_pulp_mlsdotsp20_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+
+#define MacLoad20(a_update, b_update, a_reg, b_reg, ptr, sum)  __builtin_pulp_mlsdotsup20_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+
+/* Functions for compressed min/max */
+#define CompressedMax(res, in1, in2) asm volatile( \
+                                                   "pv.max.t %[res_], %[in1_], %[in2_];" : [res_] "=r" (res) : [in1_] "r" (in1), [in2_] "r" (in2))
+
+#define CompressedMin(res, in1, in2) asm volatile(                      \
+                                                   "pv.min.t %[res_], %[in1_], %[in2_];" : [res_] "=r" (res) : [in1_] "r" (in1), [in2_] "r" (in2))
+
+#define Max16(res, in1, in2) asm volatile(                      \
+                                                   "pv.max.c %[res_], %[in1_], %[in2_];" : [res_] "=r" (res) : [in1_] "r" (in1), [in2_] "r" (in2))
+
+#define Min16(res, in1, in2) asm volatile(                      \
+                                                                        "pv.min.c %[res_], %[in1_], %[in2_];" : [res_] "=r" (res) : [in1_] "r" (in1), [in2_] "r" (in2))
+
+#define thr_cmp(state, val, threshs)  __builtin_pulp_thresh_compr(state, val, threshs)
+
+typedef unsigned char  v4u __attribute__((vector_size (4)));
+typedef   signed char  v4s __attribute__((vector_size (4)));
+
+#define bitext(x,size,off)                                      __builtin_pulp_bextract(x,size,off)
+#define bitextu(x,size,off)                                     __builtin_pulp_bextractu(x,size,off)
+#ifdef __clang__
+#define bitins(dst,not_mask_imm,src,mask_imm,off)               __builtin_binsert(dst,not_mask_imm,src,mask_imm,off)
+#else
+#define bitins(dst,not_mask_imm,src,mask_imm,off)               __builtin_pulp_binsert(dst,not_mask_imm,src,mask_imm,off)
+#endif
+#define pack(x,y,z,t)                                           __builtin_pulp_pack4(x,y,z,t)
+#define max4(a,b)                                               __builtin_pulp_maxu4(a,b)
+#define maxs4(a, b)                                             __builtin_pulp_max4(a, b)
+#define max8(a, b)                                              __builtin_pulp_maxu8(a, b)
+#define maxs8(a, b)                                             __builtin_pulp_max8(a, b)
+#define max16(a, b)                                             __builtin_pulp_maxu16(a, b)
+#define maxs16(a, b)                                            __builtin_pulp_max16(a, b)
+#define maxs20(a, b)                                            __builtin_pulp_max20(a, b)
+#define max32(a,b)                                              __builtin_pulp_maxusi(a,b)
+#define maxs32(a,b)                                             __builtin_pulp_maxsi(a,b)
+#define min32(a,b)                                              __builtin_pulp_minusi(a,b)
+#define mins32(a,b)                                             __builtin_pulp_minsi(a,b)
+#define min4(a, b)                                              __builtin_pulp_minu4(a, b)
+#define mins4(a, b)                                             __builtin_pulp_min4(a, b)
+#define min8(a, b)                                              __builtin_pulp_minu8(a, b)
+#define mins8(a, b)                                             __builtin_pulp_min8(a, b)
+#define min16(a, b)                                             __builtin_pulp_minu16(a, b)
+#define mins16(a, b)                                            __builtin_pulp_min16(a, b)
+#define mins20(a, b)                                            __builtin_pulp_min20(a, b)
+#define avg4(a,b)                                               __builtin_pulp_avgu4(a,b)
+#define avg8(a,b)                                               __builtin_pulp_avgu8(a,b)
+#define avg16(a,b)                                              __builtin_pulp_avgu16(a,b)
+#define log2(x)                                                 __builtin_pulp_fl1(x)
+#define min(a,b)                                                ((a)<(b)?(a):(b))
+#define SumDotp4(a, b, c)                                       __builtin_pulp_sdotusp4(a, b, c)
+#define SumDotp8(a, b, c)                                       __builtin_pulp_sdotusp8(a, b, c)
+#define SumDotp16(a, b, c)                                      __builtin_pulp_sdotusp16(a, b, c)
+#define SumDotps4(a, b, c)                                      __builtin_pulp_sdotsp4(a, b, c)
+#define SumDotps8(a, b, c)                                      __builtin_pulp_sdotsp8(a, b, c)
+#define SumDotps16(a, b, c)                                     __builtin_pulp_sdotsp16(a, b, c)
+#define clip4(x)                                                __builtin_pulp_clipu_r(x, 15)
+#define clip2(x)                                                __builtin_pulp_clipu_r(x, 3)
+#define clip8(x)                                                __builtin_pulp_clipu_r(x, 255)
+
+#define clips4(x)                                               __builtin_pulp_clip_r(x, 7)
+#define clips2(x)                                               __builtin_pulp_clip_r(x, 1)
+#define clips8(x)                                               __builtin_pulp_clip_r(x, 127)
+#define MacLoadInit(a_update, b_update, a_reg, b_reg, ptr)      __builtin_pulp_mlinitspr_v3(a_update, b_update, a_reg, b_reg, ptr)
+#define MacLoadUpdate(ptr)                                      __builtin_pulp_mlupdatespr_v3(ptr)
+#define MacLoadAssign(ptr)                                      __builtin_pulp_mlassignspr_v3(ptr)
+#define MacLoad4(a_update, b_update, a_reg, b_reg, ptr, sum)    __builtin_pulp_mlsdotsup4_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+#define MacLoad8(a_update, b_update, a_reg, b_reg, ptr, sum)    __builtin_pulp_mlsdotsup8_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+#define MacLoad16(a_update, b_update, a_reg, b_reg, ptr, sum)   __builtin_pulp_mlsdotsup16_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+#define MacLoads4(a_update, b_update, a_reg, b_reg, ptr, sum)   __builtin_pulp_mlsdotsp4_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+#define MacLoads8(a_update, b_update, a_reg, b_reg, ptr, sum)   __builtin_pulp_mlsdotsp8_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+#define MacLoads16(a_update, b_update, a_reg, b_reg, ptr, sum)  __builtin_pulp_mlsdotsp16_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+
+#define PACK_INT8_SIZE(x)                                    (x)
+#define PACK_INT4_SIZE(x)                                    ((x) >> 1)
+#define PACK_INT2_SIZE(x)                                    ((x) >> 2)
+
+#define MemoryFence()                                        asm volatile("":::"memory")
+
+#define LEGACY_MODE(x)                                       asm volatile ("csrwi 0x010," x)
+#define IVEC_FMT(x)                                          asm volatile ("csrwi 0x00D," x)
+#define MIXED_SKIP(x)                                        asm volatile ("csrwi 0x00F," x)
+#define A_ADDRESS(x)                                         asm volatile ("csrw 0x100, %0" :: "r" (x))
+#define W_ADDRESS(x)                                         asm volatile ("csrw 0x101, %0" :: "r" (x))
+#define A_STRIDE(x)                                          asm volatile ("csrw 0x102, %0":: "r" (x))
+#define W_STRIDE(x)                                          asm volatile ("csrw 0x103, %0":: "r" (x))
+#define A_ROLLBACK(x)                                        asm volatile ("csrw 0x104, %0":: "r" (x))
+#define W_ROLLBACK(x)                                        asm volatile ("csrw 0x105, %0":: "r" (x))
+#define A_SKIP(x)                                            asm volatile ("csrwi 0x106," x)
+#define W_SKIP(x)                                            asm volatile ("csrwi 0x107," x)
+
+static uint8_t __attribute__((noinline)) pulp_nn_quant_u2 (
+  int32_t phi,
+  int16_t m,
+  int8_t d
+  ) {
+  int32_t x = (m * phi) >> d;
+  uint8_t res = clip2(x);
+  return res;
+}
+static uint8_t __attribute__((noinline)) pulp_nn_bn_quant_u2 (
+  int32_t phi,
+  int32_t k,
+  int32_t lambda,
+  int8_t d
+  ) {
+  int32_t integer_image_phi = (k * phi) + lambda;
+  int32_t x = (integer_image_phi) >> d;
+  uint8_t res = clip2(x);
+  return res;
+}
+static uint8_t __attribute__((noinline)) pulp_nn_add_quant_u2 (
+  uint8_t pix1,
+  uint8_t pix2,
+  int16_t m1,
+  int16_t m2,
+  int8_t d
+  ) {
+  uint32_t integer_image = pix1*m1 + pix2*m2;
+  uint32_t x = (integer_image) >> d;
+  uint8_t res = clip2(x);
+  return res;
+}
+
+static int8_t __attribute__((noinline)) pulp_nn_quant_i2 (
+  int32_t phi,
+  int16_t m,
+  int8_t d
+  ) {
+  int32_t x = (m * phi) >> d;
+  int8_t res = clips2(x);
+  return res;
+}
+static int8_t __attribute__((noinline)) pulp_nn_bn_quant_i2 (
+  int32_t phi,
+  int32_t k,
+  int32_t lambda,
+  int8_t d
+  ) {
+  int32_t integer_image_phi = (k * phi) + lambda;
+  int32_t x = (integer_image_phi) >> d;
+  int8_t res = clips2(x);
+  return res;
+}
+static int8_t __attribute__((noinline)) pulp_nn_add_quant_i2 (
+  int8_t pix1,
+  int8_t pix2,
+  int16_t m1,
+  int16_t m2,
+  int8_t d
+  ) {
+  int32_t integer_image = pix1*m1 + pix2*m2;
+  int32_t x = (integer_image) >> d;
+  int8_t res = clips2(x);
+  return res;
+}
+
+static uint8_t __attribute__((noinline)) pulp_nn_quant_u4 (
+  int32_t phi,
+  int16_t m,
+  int8_t d
+  ) {
+  int32_t x = (m * phi) >> d;
+  uint8_t res = clip4(x);
+  return res;
+}
+static uint8_t __attribute__((noinline)) pulp_nn_bn_quant_u4 (
+  int32_t phi,
+  int32_t k,
+  int32_t lambda,
+  int8_t d
+  ) {
+  int32_t integer_image_phi = (k * phi) + lambda;
+  int32_t x = (integer_image_phi) >> d;
+  uint8_t res = clip4(x);
+  return res;
+}
+static uint8_t __attribute__((noinline)) pulp_nn_add_quant_u4 (
+  uint8_t pix1,
+  uint8_t pix2,
+  int16_t m1,
+  int16_t m2,
+  int8_t d
+  ) {
+  uint32_t integer_image = pix1*m1 + pix2*m2;
+  uint32_t x = (integer_image) >> d;
+  uint8_t res = clip4(x);
+  return res;
+}
+
+static int8_t __attribute__((noinline)) pulp_nn_quant_i4 (
+  int32_t phi,
+  int16_t m,
+  int8_t d
+  ) {
+  int32_t x = (m * phi) >> d;
+  int8_t res = clips4(x);
+  return res;
+}
+static int8_t __attribute__((noinline)) pulp_nn_bn_quant_i4 (
+  int32_t phi,
+  int32_t k,
+  int32_t lambda,
+  int8_t d
+  ) {
+  int32_t integer_image_phi = (k * phi) + lambda;
+  int32_t x = (integer_image_phi) >> d;
+  int8_t res = clips4(x);
+  return res;
+}
+static int8_t __attribute__((noinline)) pulp_nn_add_quant_i4 (
+  int8_t pix1,
+  int8_t pix2,
+  int16_t m1,
+  int16_t m2,
+  int8_t d
+  ) {
+  int32_t integer_image = pix1*m1 + pix2*m2;
+  int32_t x = (integer_image) >> d;
+  int8_t res = clips4(x);
+  return res;
+}
+
+static uint8_t __attribute__((noinline)) pulp_nn_quant_u8 (
+  int32_t phi,
+  int16_t m,
+  int8_t d
+  ) {
+  int32_t x = (m * phi) >> d;
+  uint8_t res = clip8(x);
+  return res;
+}
+static uint8_t __attribute__((noinline)) pulp_nn_bn_quant_u8 (
+  int32_t phi,
+  int32_t k,
+  int32_t lambda,
+  int8_t d
+  ) {
+  int32_t integer_image_phi = (k * phi) + lambda;
+  int32_t x = (integer_image_phi) >> d;
+  uint8_t res = clip8(x);
+  return res;
+}
+static uint8_t __attribute__((noinline)) pulp_nn_add_quant_u8 (
+  uint8_t pix1,
+  uint8_t pix2,
+  int16_t m1,
+  int16_t m2,
+  int8_t d
+  ) {
+  uint32_t integer_image = pix1*m1 + pix2*m2;
+  uint32_t x = (integer_image) >> d;
+  uint8_t res = clip8(x);
+  return res;
+}
+
+static int8_t __attribute__((noinline)) pulp_nn_quant_i8 (
+  int32_t phi,
+  int16_t m,
+  int8_t d
+  ) {
+  int32_t x = (m * phi) >> d;
+  int8_t res = clips8(x);
+  return res;
+}
+static int8_t __attribute__((noinline)) pulp_nn_bn_quant_i8 (
+  int32_t phi,
+  int32_t k,
+  int32_t lambda,
+  int8_t d
+  ) {
+  int32_t integer_image_phi = (k * phi) + lambda;
+  int32_t x = (integer_image_phi) >> d;
+  int8_t res = clips8(x);
+  return res;
+}
+static int8_t __attribute__((noinline)) pulp_nn_add_quant_i8 (
+  int8_t pix1,
+  int8_t pix2,
+  int16_t m1,
+  int16_t m2,
+  int8_t d
+  ) {
+  int32_t integer_image = pix1*m1 + pix2*m2;
+  int32_t x = (integer_image) >> d;
+  int8_t res = clips8(x);
+  return res;
+}
+
+
+static uint8_t __attribute__((noinline)) pulp_nn_u4_quant(int input, int16_t * pThr)
+{
+  if(input <= pThr[7] )
+  {
+    if( input <= pThr[3])
+    {
+      if( input <= pThr[1])
+      {
+        if( input <= pThr[0])
+          return 0;
+        else
+          return 1;
+      }
+      else
+      {
+        if( input <= pThr[2])
+          return 2;
+        else
+          return 3;
+      }
+    }
+    else
+    {
+      if( input <= pThr[5])
+      {
+        if( input <= pThr[4])
+          return 4;
+        else
+          return 5;
+      }
+      else
+      {
+        if( input <= pThr[6])
+          return 6;
+        else
+          return 7;
+      }
+    }
+  }
+  else
+  {
+    if( input <= pThr[11])
+    {
+      if( input <= pThr[9])
+      {
+        if( input <= pThr[8])
+          return 8;
+        else
+          return 9;
+      }
+      else
+      {
+        if( input <= pThr[10])
+          return 10;
+        else
+          return 11;
+      }
+    }
+    else
+    {
+      if( input <= pThr[13])
+      {
+        if( input <= pThr[12])
+          return 12;
+        else
+          return 13;
+      }
+      else
+      {
+        if( input <= pThr[14])
+          return 14;
+        else
+          return 15;
+      }
+    }
+  }
+}
+
+static uint8_t __attribute__((noinline)) pulp_nn_u2_quant(int input, int16_t * pThr)
+{
+  if( input <= pThr[1])
+  {
+    if( input <= pThr[0])
+        {
+      return 0;
+    }
+        else
+        {
+      return 1;
+    }
+  }
+    else
+    {
+    if( input <= pThr[2])
+        {
+      return 2;
+    }
+        else
+        {
+      return 3;
+    }
+  }
+}
+
+/*
+ * Common
+ */
+
+
+static v4s __attribute__((noinline)) pulp_nn_i4_to_i8_r( int8_t *pSrc)
+{
+  v4s Src = *((v4s*) pSrc);
+  int8_t bext1, bext2, bext3, bext4;
+
+  bext1 = (int8_t) bitext((int) Src, 4, 0);
+  bext2 = (int8_t) bitext((int) Src, 4, 4);
+  bext3 = (int8_t) bitext((int) Src, 4, 8);
+  bext4 = (int8_t) bitext((int) Src, 4, 12);
+  v4s res = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+
+  return res;
+}
+
+static v4u __attribute__((noinline)) pulp_nn_u4_to_u8_r(uint8_t *pSrc)
+{
+  v4u Src = *((v4u*) pSrc);
+  uint8_t bext1, bext2, bext3, bext4;
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 4, 0);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 4, 4);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 4, 8);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 4, 12);
+  v4u res = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+
+  return res;
+}
+
+static v4s __attribute__((noinline)) pulp_nn_i2_to_i8_r( int8_t *pSrc)
+{
+  v4s Src = *((v4s*) pSrc);
+  int8_t bext1, bext2, bext3, bext4;
+
+  bext1 = (int8_t) bitext((int) Src, 2, 0);
+  bext2 = (int8_t) bitext((int) Src, 2, 2);
+  bext3 = (int8_t) bitext((int) Src, 2, 4);
+  bext4 = (int8_t) bitext((int) Src, 2, 6);
+  v4s res = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+
+  return res;
+}
+
+static v4u __attribute__((noinline)) pulp_nn_u2_to_u8_r(uint8_t *pSrc)
+{
+  v4u Src = *((v4u*) pSrc);
+  uint8_t bext1, bext2, bext3, bext4;
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 0);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 2);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 4);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 6);
+  v4u res = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+
+  return res;
+}
+
+static v4s __attribute__((noinline)) pulp_nn_i2_to_i4_r(int8_t *pSrc)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  v4s Src = *((v4s*) pSrc);
+  int8_t bext1, bext2, bext3, bext4;
+  int8_t out1, out2, out3, out4;
+
+  bext1 = (int8_t) bitextu((int) Src, 2, 0);
+  bext2 = (int8_t) bitextu((int) Src, 2, 2);
+  bext3 = (int8_t) bitextu((int) Src, 2, 4);
+  bext4 = (int8_t) bitextu((int) Src, 2, 6);
+
+  out1 = bitins(bext1, n_mask, bext2, mask, off);
+  out2 = bitins(bext3, n_mask, bext4, mask, off);
+
+  bext1 = (int8_t) bitextu((int) Src, 2, 8);
+  bext2 = (int8_t) bitextu((int) Src, 2, 10);
+  bext3 = (int8_t) bitextu((int) Src, 2, 12);
+  bext4 = (int8_t) bitextu((int) Src, 2, 14);
+
+  out3 = bitins(bext1, n_mask, bext2, mask, off);
+  out4 = bitins(bext3, n_mask, bext4, mask, off);
+
+  v4s res = pack((int8_t) out1, (int8_t) out2, (int8_t) out3, (int8_t) out4);
+
+  return res;
+}
+
+static v4u __attribute__((noinline)) pulp_nn_u2_to_u4_r(uint8_t *pSrc)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  v4u Src = *((v4u*) pSrc);
+  uint8_t bext1, bext2, bext3, bext4;
+  uint8_t out1, out2, out3, out4;
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 0);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 2);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 4);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 6);
+
+  out1 = bitins(bext1, n_mask, bext2, mask, off);
+  out2 = bitins(bext3, n_mask, bext4, mask, off);
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 8);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 10);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 12);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 14);
+
+  out3 = bitins(bext1, n_mask, bext2, mask, off);
+  out4 = bitins(bext3, n_mask, bext4, mask, off);
+
+  v4u res = pack((uint8_t) out1, (uint8_t) out2, (uint8_t) out3, (uint8_t) out4);
+
+  return res;
+}
+
+static int8_t *__attribute__((always_inline)) pulp_nn_i4_to_i8( int8_t *pSrc, int8_t *pDst)
+{
+  v4s Src = *((v4s*) pSrc);
+  int8_t bext1, bext2, bext3, bext4;
+
+  pSrc+=4;
+
+  bext1 = (int8_t) bitext((int) Src, 4, 0);
+  bext2 = (int8_t) bitext((int) Src, 4, 4);
+  bext3 = (int8_t) bitext((int) Src, 4, 8);
+  bext4 = (int8_t) bitext((int) Src, 4, 12);
+  *((v4s*)pDst) = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+   MemoryFence();
+  pDst+=4;
+  bext1 = (int8_t) bitext((int) Src, 4, 16);
+  bext2 = (int8_t) bitext((int) Src, 4, 20);
+  bext3 = (int8_t) bitext((int) Src, 4, 24);
+  bext4 = (int8_t) bitext((int) Src, 4, 28);
+  *((v4s*)pDst) = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+
+  return pSrc;
+}
+
+static uint8_t *__attribute__((always_inline)) pulp_nn_u4_to_u8(uint8_t *pSrc, uint8_t *pDst)
+{
+  v4u Src = *((v4u*) pSrc);
+  uint8_t bext1, bext2, bext3, bext4;
+
+  pSrc+=4;
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 4, 0);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 4, 4);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 4, 8);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 4, 12);
+  *((v4u*)pDst) = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 4, 16);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 4, 20);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 4, 24);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 4, 28);
+  *((v4u*)pDst) = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+
+  return pSrc;
+}
+
+static int8_t *__attribute__((always_inline)) pulp_nn_i2_to_i8( int8_t * pSrc, int8_t * pDst)
+{
+  v4s Src = *((v4s*) pSrc);
+  int8_t bext1, bext2, bext3, bext4;
+
+  pSrc+=4;
+
+  bext1 = (int8_t) bitext((int) Src, 2, 0);
+  bext2 = (int8_t) bitext((int) Src, 2, 2);
+  bext3 = (int8_t) bitext((int) Src, 2, 4);
+  bext4 = (int8_t) bitext((int) Src, 2, 6);
+  *((v4s*)pDst) = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (int8_t) bitext((int) Src, 2, 8);
+  bext2 = (int8_t) bitext((int) Src, 2, 10);
+  bext3 = (int8_t) bitext((int) Src, 2, 12);
+  bext4 = (int8_t) bitext((int) Src, 2, 14);
+  *((v4s*)pDst) = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (int8_t) bitext((int) Src, 2, 16);
+  bext2 = (int8_t) bitext((int) Src, 2, 18);
+  bext3 = (int8_t) bitext((int) Src, 2, 20);
+  bext4 = (int8_t) bitext((int) Src, 2, 22);
+  *((v4s*)pDst) = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (int8_t) bitext((int) Src, 2, 24);
+  bext2 = (int8_t) bitext((int) Src, 2, 26);
+  bext3 = (int8_t) bitext((int) Src, 2, 28);
+  bext4 = (int8_t) bitext((int) Src, 2, 30);
+  *((v4s*)pDst) = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+
+  return pSrc;
+}
+
+static uint8_t *__attribute__((always_inline)) pulp_nn_u2_to_u8(uint8_t * pSrc, uint8_t * pDst)
+{
+  v4u Src = *((v4u*) pSrc);
+  uint8_t bext1, bext2, bext3, bext4;
+
+  pSrc+=4;
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 0);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 2);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 4);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 6);
+  *((v4u*)pDst) = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 8);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 10);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 12);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 14);
+  *((v4u*)pDst) = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 16);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 18);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 20);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 22);
+  *((v4u*)pDst) = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 24);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 26);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 28);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 30);
+  *((v4u*)pDst) = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+
+  return pSrc;
+}
+
+static int8_t *__attribute__((always_inline)) pulp_nn_i2_to_i4( int8_t * pSrc, int8_t * pDst)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  v4s Src = *((v4s*) pSrc);
+  int8_t bext1, bext2, bext3, bext4;
+  int8_t out1, out2, out3, out4;
+
+  pSrc+=4;
+
+  bext1 = (int8_t) bitext((int) Src, 2, 0);
+  bext2 = (int8_t) bitext((int) Src, 2, 2);
+  bext3 = (int8_t) bitext((int) Src, 2, 4);
+  bext4 = (int8_t) bitext((int) Src, 2, 6);
+
+  out1 = bitins(bext1, n_mask, bext2, mask, off);
+  out2 = bitins(bext3, n_mask, bext4, mask, off);
+  MemoryFence();
+
+  bext1 = (int8_t) bitext((int) Src, 2, 8);
+  bext2 = (int8_t) bitext((int) Src, 2, 10);
+  bext3 = (int8_t) bitext((int) Src, 2, 12);
+  bext4 = (int8_t) bitext((int) Src, 2, 14);
+
+  out3 = bitins(bext1, n_mask, bext2, mask, off);
+  out4 = bitins(bext3, n_mask, bext4, mask, off);
+
+  *((v4s*)pDst) = pack((int8_t) out1, (int8_t) out2, (int8_t) out3, (int8_t) out4);
+  MemoryFence();
+
+  pDst+=4;
+  bext1 = (int8_t) bitext((int) Src, 2, 16);
+  bext2 = (int8_t) bitext((int) Src, 2, 18);
+  bext3 = (int8_t) bitext((int) Src, 2, 20);
+  bext4 = (int8_t) bitext((int) Src, 2, 22);
+
+  out1 = bitins(bext1, n_mask, bext2, mask, off);
+  out2 = bitins(bext3, n_mask, bext4, mask, off);
+  MemoryFence();
+
+  bext1 = (int8_t) bitext((int) Src, 2, 24);
+  bext2 = (int8_t) bitext((int) Src, 2, 26);
+  bext3 = (int8_t) bitext((int) Src, 2, 28);
+  bext4 = (int8_t) bitext((int) Src, 2, 30);
+
+  out3 = bitins(bext1, n_mask, bext2, mask, off);
+  out4 = bitins(bext3, n_mask, bext4, mask, off);
+
+  *((v4s*)pDst) = pack((int8_t) out1, (int8_t) out2, (int8_t) out3, (int8_t) out4);
+
+  return pSrc;
+}
+
+static uint8_t *__attribute__((always_inline)) pulp_nn_u2_to_u4( uint8_t * pSrc, uint8_t * pDst)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  v4u Src = *((v4u*) pSrc);
+  uint8_t bext1, bext2, bext3, bext4;
+  uint8_t out1, out2, out3, out4;
+
+  pSrc+=4;
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 0);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 2);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 4);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 6);
+
+  out1 = bitins(bext1, n_mask, bext2, mask, off);
+  out2 = bitins(bext3, n_mask, bext4, mask, off);
+  MemoryFence();
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 8);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 10);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 12);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 14);
+
+  out3 = bitins(bext1, n_mask, bext2, mask, off);
+  out4 = bitins(bext3, n_mask, bext4, mask, off);
+
+  *((v4u*)pDst) = pack((uint8_t) out1, (uint8_t) out2, (uint8_t) out3, (uint8_t) out4);
+  MemoryFence();
+
+  pDst+=4;
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 16);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 18);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 20);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 22);
+
+  out1 = bitins(bext1, n_mask, bext2, mask, off);
+  out2 = bitins(bext3, n_mask, bext4, mask, off);
+  MemoryFence();
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 24);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 26);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 28);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 30);
+
+  out3 = bitins(bext1, n_mask, bext2, mask, off);
+  out4 = bitins(bext3, n_mask, bext4, mask, off);
+  
+  *((v4u*)pDst) = pack((uint8_t) out1, (uint8_t) out2, (uint8_t) out3, (uint8_t) out4);
+
+  return pSrc;
+}
+
+/*
+ * XpulpV2
+ */
+
+static void __attribute__((noinline)) pulp_zero_mem(uint8_t * pBuffer, unsigned int size)
+{
+  int lfover = size &0x3;
+  for (int i=0; i<(size>>2); i++)
+  {
+    *((v4u *)pBuffer) = (v4u){0,0,0,0};
+    MemoryFence();
+    pBuffer+=4;
+  }
+  while(lfover)
+  {
+    *pBuffer++=0;
+    lfover--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_im2col_u2_to_u8(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_u2_to_u8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 16;
+  }
+  while (lfover)
+  {
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 0);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 2);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 4);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 6);
+    pOutput++;
+    pInput++;
+    lfover -= 4;
+  }
+}
+static void __attribute__((noinline)) pulp_nn_im2col_i2_to_i8(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_i2_to_i8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 16;
+  }
+  while (lfover)
+  {
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 0);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 2);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 4);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 6);
+    pOutput++;
+    pInput++;
+    lfover -= 4;
+  }
+}
+static void __attribute__((noinline)) pulp_nn_im2col_u4_to_u8(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 3u;
+  int lfover = blockSize & 0x07;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_u4_to_u8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 8;
+  }
+  while (lfover)
+  {
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 4, 0);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 4, 4);
+    pOutput++;
+    pInput++;
+    lfover -= 2;
+  }
+}
+static void __attribute__((noinline)) pulp_nn_im2col_i4_to_i8(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 3u;
+  int lfover = blockSize & 0x07;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_i4_to_i8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 8;
+  }
+  while (lfover)
+  {
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 4, 0);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 4, 4);
+    pOutput++;
+    pInput++;
+    lfover -= 2;
+  }
+}
+static void __attribute__((noinline)) pulp_nn_im2col_u8_to_u8(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 2u;
+  int lfover = blockSize & 0x03;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4u*)pOutput) = *((v4u*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    *((uint8_t*)pOutput) = *((uint8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover--;
+  }
+}
+static void __attribute__((noinline)) pulp_nn_im2col_i8_to_i8(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 2u;
+  int lfover = blockSize & 0x03;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4s*)pOutput) = *((v4s*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    *((int8_t*)pOutput) = *((int8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover--;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_im2col_u2_to_u2(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4u*)pOutput) = *((v4u*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    uint8_t extr;
+    *((uint8_t*)pOutput) = *((uint8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover -= 4;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_i2_to_i2(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4s*)pOutput) = *((v4s*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    int8_t extr;
+    *((int8_t*)pOutput) = *((int8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover -= 4;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_u2_to_u4(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_u2_to_u4(pInput, pOutput);
+    MemoryFence();
+    pOutput += 8;
+  }
+  while (lfover)
+  {
+    uint8_t extr;
+    // TODO: this is potentially dangerous/wrong if the number of channels
+    // is not a multiple of 8!
+    *((v4u *)pOutput) = pulp_nn_u2_to_u4_r(pInput);
+    pInput += 2;
+    pOutput += 4;
+    lfover -= 8;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_i2_to_i4(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_i2_to_i4(pInput, pOutput);
+    MemoryFence();
+    pOutput += 8;
+  }
+  while (lfover)
+  {
+    int8_t extr;
+    // TODO: this is potentially dangerous/wrong if the number of channels
+    // is not a multiple of 8!
+    *((v4s *)pOutput) = pulp_nn_i2_to_i4_r(pInput);
+    pInput += 2;
+    pOutput += 4;
+    lfover -= 8;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_u2_to_u8(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_u2_to_u8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 16;
+  }
+  while (lfover)
+  {
+    uint8_t extr;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 0);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 2);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 4);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 6);
+    pOutput++;
+    pInput++;
+    lfover -= 4;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_i2_to_i8(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_i2_to_i8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 16;
+  }
+  while (lfover)
+  {
+    int8_t extr;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 0);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 2);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 4);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 6);
+    pOutput++;
+    pInput++;
+    lfover -= 4;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_u4_to_u4(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 3u;
+  int lfover = blockSize & 0x07;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4u*)pOutput) = *((v4u*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    uint8_t extr;
+    *((uint8_t*)pOutput) = *((uint8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover -= 2;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_i4_to_i4(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 3u;
+  int lfover = blockSize & 0x07;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4s*)pOutput) = *((v4s*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    int8_t extr;
+    *((int8_t*)pOutput) = *((int8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover -= 2;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_u4_to_u8(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 3u;
+  int lfover = blockSize & 0x07;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_u4_to_u8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 8;
+  }
+  while (lfover)
+  {
+    uint8_t extr;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 4, 0);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 4, 4);
+    pOutput++;
+    pInput++;
+    lfover -= 2;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_i4_to_i8(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 3u;
+  int lfover = blockSize & 0x07;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_i4_to_i8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 8;
+  }
+  while (lfover)
+  {
+    int8_t extr;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 4, 0);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 4, 4);
+    pOutput++;
+    pInput++;
+    lfover -= 2;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_u8_to_u8(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 2u;
+  int lfover = blockSize & 0x03;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4u*)pOutput) = *((v4u*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    uint8_t extr;
+    *((uint8_t*)pOutput) = *((uint8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover -= 1;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_i8_to_i8(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 2u;
+  int lfover = blockSize & 0x03;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4s*)pOutput) = *((v4s*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    int8_t extr;
+    *((int8_t*)pOutput) = *((int8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover -= 1;
+  }
+}
+
+
+
+static void __attribute__((noinline)) pulp_nn_compare_and_replace_if_larger_u8(uint8_t * base,
+                                                uint8_t * target,
+                                                uint16_t length)
+{
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+  v4u inp;
+  v4u com;
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    inp = *((v4u*)pIn);
+    com = *((v4u*)pCom);
+
+    *((v4u*)pIn) = max4(inp, com);
+
+    pCom+=4;
+    pIn+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    if(*pIn<*pCom)
+      *pIn=*pCom;
+    
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_compare_and_replace_if_larger_i8(
+    int8_t * base, int8_t * target, uint16_t length) {
+  int8_t *pIn = base;
+  int8_t *pCom = target;
+  v4s inp;
+  v4s com;
+  int cnt = length >> 2;
+
+  while (cnt > 0u) {
+    inp = *((v4s *)pIn);
+    com = *((v4s *)pCom);
+
+    *((v4s *)pIn) = maxs4(inp, com);
+
+    pCom += 4;
+    pIn += 4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left > 0u) {
+    if (*pIn < *pCom)
+      *pIn = *pCom;
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_avg_and_replace_u8(uint8_t * base,
+                                  uint8_t * target,
+                                  uint16_t length)
+{
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+
+  while (length>0u)
+  {
+    *pIn = ((*pIn + *pCom) >> 1);
+
+    pIn++;
+    pCom++;
+    length--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_compare_and_replace_if_larger_u4(uint8_t * base,
+                                                uint8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+  v4u inp[2];
+  v4u com[2];
+  uint8_t *out;
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    pulp_nn_u4_to_u8(pIn, (uint8_t *)inp);
+    pulp_nn_u4_to_u8(pCom, (uint8_t *)com);
+
+    *((v4u *)out) = max4(inp[0], com[0]);
+
+    *((uint8_t*)pIn) = bitins(*out, n_mask, *(out + 1), mask, off);
+    pIn++;
+    *((uint8_t*)pIn) = bitins(*(out + 2), n_mask, *(out + 3), mask, off);
+    pIn++;
+
+    *((v4u *)out) = max4(inp[1], com[1]);
+
+    *((uint8_t*)pIn) = bitins(*out, n_mask, *(out + 1), mask, off);
+    pIn++;
+    *((uint8_t*)pIn) = bitins(*(out + 2), n_mask, *(out + 3), mask, off);
+    pIn++;
+
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 4, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 4, 4);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 4, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 4, 4);
+
+    if(inA0<inB0)
+      inA0=inB0;
+
+    if(inA1<inB1)
+      inA1=inB1;
+
+    *((uint8_t*)pIn) = bitins(inA0, n_mask, inA1, mask, off);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_compare_and_replace_if_larger_i4(int8_t * base,
+                                                int8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  int8_t *pIn = base;
+  int8_t *pCom = target;
+  v4s inp[2];
+  v4s com[2];
+  int8_t *out;
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    pulp_nn_i4_to_i8(pIn, (int8_t *)inp);
+    pulp_nn_i4_to_i8(pCom, (int8_t *)com);
+
+    *((v4s *)out) = maxs4(inp[0], com[0]);
+
+    *((int8_t*)pIn) = bitins(*out, n_mask, *(out + 1), mask, off);
+    pIn++;
+    *((int8_t*)pIn) = bitins(*(out + 2), n_mask, *(out + 3), mask, off);
+    pIn++;
+
+    *((v4s *)out) = maxs4(inp[1], com[1]);
+
+    *((int8_t*)pIn) = bitins(*out, n_mask, *(out + 1), mask, off);
+    pIn++;
+    *((int8_t*)pIn) = bitins(*(out + 2), n_mask, *(out + 3), mask, off);
+    pIn++;
+
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    int8_t inA0 = (int8_t) bitext((int) *pIn, 4, 0);
+    int8_t inA1 = (int8_t) bitext((int) *pIn, 4, 4);
+    int8_t inB0 = (int8_t) bitext((int) *pCom, 4, 0);
+    int8_t inB1 = (int8_t) bitext((int) *pCom, 4, 4);
+
+    if(inA0<inB0)
+      inA0=inB0;
+
+    if(inA1<inB1)
+      inA1=inB1;
+
+    *((int8_t*)pIn) = bitins(inA0, n_mask, inA1, mask, off);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_avg_and_replace_u4(uint8_t * base,
+                                  uint8_t * target,
+                                  uint16_t length)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+
+  while (length>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 4, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 4, 4);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 4, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 4, 4);
+
+    inA0 = ((inA0 + inB0) >> 1);
+    inA1 = ((inA1 + inB1) >> 1);
+
+    *((uint8_t*)pIn) = bitins(inA0, n_mask, inA1, mask, off);
+
+    pIn++;
+    pCom++;
+    length--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_compare_and_replace_if_larger_u2(uint8_t * base,
+                                                uint8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask2 = 0x0c;
+  int8_t n_mask2 = ~ mask2;
+  int8_t mask4 = 0x30;
+  int8_t n_mask4 = ~ mask4;
+  int8_t mask6 = 0xc0;
+  int8_t n_mask6 = ~ mask6;
+  int8_t off2 = 2;
+  int8_t off4 = 4;
+  int8_t off6 = 6;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+  v4u inp[4];
+  v4u com[4];
+  uint8_t *out;
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    pulp_nn_u2_to_u8(pIn, inp);
+    pulp_nn_u2_to_u8(pCom, com);
+
+    *((v4u*)out) = max4(inp[0], com[0]);
+
+    uint8_t inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    *((v4u*)out) = max4(inp[1], com[1]);
+
+    inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    *((v4u*)out) = max4(inp[2], com[2]);
+
+    inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    *((v4u*)out) = max4(inp[3], com[3]);
+
+    inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 2, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 2, 2);
+    uint8_t inA2 = (uint8_t) bitextu((unsigned int) *pIn, 2, 4);
+    uint8_t inA3 = (uint8_t) bitextu((unsigned int) *pIn, 2, 6);
+    v4u inA4 = pack((uint8_t) inA0, (uint8_t) inA1, (uint8_t) inA2, (uint8_t) inA3);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 2, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 2, 2);
+    uint8_t inB2 = (uint8_t) bitextu((unsigned int) *pCom, 2, 4);
+    uint8_t inB3 = (uint8_t) bitextu((unsigned int) *pCom, 2, 6);
+    v4u inB4 = pack((uint8_t) inB0, (uint8_t) inB1, (uint8_t) inB2, (uint8_t) inB3);
+
+    *((v4u*)out) = max4(inA4, inB4);
+
+    uint8_t inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+
+static void __attribute__((noinline)) pulp_nn_compare_and_replace_if_larger_i2(int8_t * base,
+                                                int8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask2 = 0x0c;
+  int8_t n_mask2 = ~ mask2;
+  int8_t mask4 = 0x30;
+  int8_t n_mask4 = ~ mask4;
+  int8_t mask6 = 0xc0;
+  int8_t n_mask6 = ~ mask6;
+  int8_t off2 = 2;
+  int8_t off4 = 4;
+  int8_t off6 = 6;
+
+  int8_t *pIn = base;
+  int8_t *pCom = target;
+  v4s inp[4];
+  v4s com[4];
+  int8_t *out;
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    pulp_nn_i2_to_i8(pIn, inp);
+    pulp_nn_i2_to_i8(pCom, com);
+
+    *((v4s*)out) = maxs4(inp[0], com[0]);
+
+    int8_t inA = (int8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((int8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    *((v4s*)out) = maxs4(inp[1], com[1]);
+
+    inA = (int8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((int8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    *((v4s*)out) = maxs4(inp[2], com[2]);
+
+    inA = (int8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((int8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    *((v4s*)out) = maxs4(inp[3], com[3]);
+
+    inA = (int8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((int8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    int8_t inA0 = (int8_t) bitext((unsigned int) *pIn, 2, 0);
+    int8_t inA1 = (int8_t) bitext((unsigned int) *pIn, 2, 2);
+    int8_t inA2 = (int8_t) bitext((unsigned int) *pIn, 2, 4);
+    int8_t inA3 = (int8_t) bitext((unsigned int) *pIn, 2, 6);
+    v4s inA4 = pack((int8_t) inA0, (int8_t) inA1, (int8_t) inA2, (int8_t) inA3);
+    int8_t inB0 = (int8_t) bitext((unsigned int) *pCom, 2, 0);
+    int8_t inB1 = (int8_t) bitext((unsigned int) *pCom, 2, 2);
+    int8_t inB2 = (int8_t) bitext((unsigned int) *pCom, 2, 4);
+    int8_t inB3 = (int8_t) bitext((unsigned int) *pCom, 2, 6);
+    v4s inB4 = pack((int8_t) inB0, (int8_t) inB1, (int8_t) inB2, (int8_t) inB3);
+
+    *((v4s*)out) = maxs4(inA4, inB4);
+
+    int8_t inA = (int8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((int8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_avg_and_replace_u2(uint8_t * base,
+                                  uint8_t * target,
+                                  uint16_t length)
+{
+  int8_t mask2 = 0x0c;
+  int8_t n_mask2 = ~ mask2;
+  int8_t mask4 = 0x30;
+  int8_t n_mask4 = ~ mask4;
+  int8_t mask6 = 0xc0;
+  int8_t n_mask6 = ~ mask6;
+  int8_t off2 = 2;
+  int8_t off4 = 4;
+  int8_t off6 = 6;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+
+  while (length>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 2, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 2, 2);
+    uint8_t inA2 = (uint8_t) bitextu((unsigned int) *pIn, 2, 4);
+    uint8_t inA3 = (uint8_t) bitextu((unsigned int) *pIn, 2, 6);
+
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 2, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 2, 2);
+    uint8_t inB2 = (uint8_t) bitextu((unsigned int) *pCom, 2, 4);
+    uint8_t inB3 = (uint8_t) bitextu((unsigned int) *pCom, 2, 6);
+
+    inA0 = ((inA0 + inB0) >> 1);
+    inA1 = ((inA1 + inB1) >> 1);
+    inA2 = ((inA2 + inB2) >> 1);
+    inA3 = ((inA3 + inB3) >> 1);
+
+    uint8_t inA = (uint8_t) bitins(inA0, n_mask2, inA1, mask2, off2);
+    inA = bitins(inA, n_mask4, inA2, mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, inA3, mask6, off6);
+
+    pIn++;
+    pCom++;
+    length--;
+  }
+}
+
+/*
+ * XpulpNN
+ */
+
+static void __attribute__((noinline)) xpulp_nn_zero_mem_u8(uint8_t * pBuffer, unsigned int size)
+{
+  int lfover = size &0x3;
+  for (int i=0; i<(size>>2); i++)
+  {
+    *((v4u *)pBuffer) = (v4u){0,0,0,0};
+    MemoryFence();
+    pBuffer+=4;
+  }
+  while(lfover)
+  {
+    *pBuffer++=0;
+    lfover--;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_zero_mem_u4(uint8_t * pBuffer, unsigned int size)
+{
+  int lfover = size &0x7;
+  for (int i=0; i<(size>>3); i++)
+  {
+    *((v4u *)pBuffer) = (v4u){0,0,0,0};
+    MemoryFence();
+    pBuffer+=4;
+  }
+  while(lfover)
+  {
+    *pBuffer++=0;
+    lfover-=2;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_zero_mem_u2(uint8_t * pBuffer, unsigned int size)
+{
+  int lfover = size &0xf;
+  for (int i=0; i<(size>>4); i++)
+  {
+    *((v4u *)pBuffer) = (v4u){0,0,0,0};
+    MemoryFence();
+    pBuffer+=4;
+  }
+  while(lfover)
+  {
+    *pBuffer++=0;
+    lfover-=4;
+  }
+}
+
+
+static void __attribute__((noinline)) xpulp_tnn_zero_mem_ternary(uint8_t * pBuffer, unsigned int size, unsigned int uns)
+{
+  uint8_t pad_val = 0xd9;
+  uint32_t pad_vec = 0xd9d9d9d9;
+  if (uns) {
+    // if we are using an unsigned kernel, we need to pad with -1 because the hardware will add a +1 to ALL values!
+    pad_val = 0xff;
+    pad_vec = 0xffffffff;
+  }
+  int lfover = size &0xf;
+  for (int i=0; i<(size>>4); i++)
+    {
+      *((v4u *)pBuffer) = (v4u)pad_vec;
+      MemoryFence();
+      pBuffer+=4;
+    }
+  while(lfover)
+    {
+      *pBuffer++=pad_val;
+      lfover-=4;
+    }
+}
+
+static void __attribute__((noinline)) xpulp_nn_compare_and_replace_if_larger_u4(uint8_t * base,
+                                                uint8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    *((uint32_t *)pIn) = max8(*((uint32_t *)pIn), *((uint32_t *)pCom));
+
+    pIn+=4;
+    pCom+=4;
+
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 4, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 4, 4);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 4, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 4, 4);
+
+    if(inA0<inB0)
+      inA0=inB0;
+
+    if(inA1<inB1)
+      inA1=inB1;
+
+    *((uint8_t*)pIn) = bitins(inA0, n_mask, inA1, mask, off);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_compare_and_replace_if_larger_i4(
+    int8_t * base, int8_t * target, uint16_t length) {
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~mask;
+  int8_t off = 0x04;
+
+  int8_t *pIn = base;
+  int8_t *pCom = target;
+
+  int cnt = length >> 2;
+
+  while (cnt > 0u) {
+    *((int32_t *)pIn) = maxs8(*((int32_t *)pIn), *((int32_t *)pCom));
+
+    pIn += 4;
+    pCom += 4;
+
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left > 0u) {
+    int8_t inA0 = (int8_t)bitext((int)*pIn, 4, 0);
+    int8_t inA1 = (int8_t)bitext((int)*pIn, 4, 4);
+    int8_t inB0 = (int8_t)bitext((int)*pCom, 4, 0);
+    int8_t inB1 = (int8_t)bitext((int)*pCom, 4, 4);
+
+    if (inA0 < inB0)
+      inA0 = inB0;
+
+    if (inA1 < inB1)
+      inA1 = inB1;
+
+    *((int8_t *)pIn) = bitins(inA0, n_mask, inA1, mask, off);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_avg_and_replace_u4(uint8_t * base,
+                                  uint8_t * target,
+                                  uint16_t length)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+
+  int cnt = length >> 2;
+
+  while (cnt > 0u)
+  {
+    *((uint32_t *)pIn) = avg8(*((uint32_t *)pIn), *((uint32_t *)pCom));
+
+    pIn+=4;
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+
+  while (left>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 4, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 4, 4);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 4, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 4, 4);
+
+    inA0 = ((inA0 + inB0) >> 1);
+    inA1 = ((inA1 + inB1) >> 1);
+
+    *((uint8_t*)pIn) = bitins(inA0, n_mask, inA1, mask, off);
+
+    pIn++;
+    pCom++;
+    length--;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_compare_and_replace_if_larger_u2(uint8_t * base,
+                                                uint8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask2 = 0x0c;
+  int8_t n_mask2 = ~ mask2;
+  int8_t mask4 = 0x30;
+  int8_t n_mask4 = ~ mask4;
+  int8_t mask6 = 0xc0;
+  int8_t n_mask6 = ~ mask6;
+  int8_t off2 = 2;
+  int8_t off4 = 4;
+  int8_t off6 = 6;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+  uint8_t *out;
+
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    *((uint32_t *)pIn) = max16(*((uint32_t *)pIn), *((uint32_t *)pCom));
+
+    pIn+=4;
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 2, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 2, 2);
+    uint8_t inA2 = (uint8_t) bitextu((unsigned int) *pIn, 2, 4);
+    uint8_t inA3 = (uint8_t) bitextu((unsigned int) *pIn, 2, 6);
+    v4u inA4 = pack((uint8_t) inA0, (uint8_t) inA1, (uint8_t) inA2, (uint8_t) inA3);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 2, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 2, 2);
+    uint8_t inB2 = (uint8_t) bitextu((unsigned int) *pCom, 2, 4);
+    uint8_t inB3 = (uint8_t) bitextu((unsigned int) *pCom, 2, 6);
+    v4u inB4 = pack((uint8_t) inB0, (uint8_t) inB1, (uint8_t) inB2, (uint8_t) inB3);
+
+    *((v4u*)out) = max4(inA4, inB4);
+
+    uint8_t inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_compare_and_replace_if_larger_i2(int8_t * base,
+                                                int8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask2 = 0x0c;
+  int8_t n_mask2 = ~ mask2;
+  int8_t mask4 = 0x30;
+  int8_t n_mask4 = ~ mask4;
+  int8_t mask6 = 0xc0;
+  int8_t n_mask6 = ~ mask6;
+  int8_t off2 = 2;
+  int8_t off4 = 4;
+  int8_t off6 = 6;
+
+  int8_t *pIn = base;
+  int8_t *pCom = target;
+  int8_t *out;
+
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    *((int32_t *)pIn) = maxs16(*((int32_t *)pIn), *((int32_t *)pCom));
+
+    pIn+=4;
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    int8_t inA0 = (int8_t) bitext((int) *pIn, 2, 0);
+    int8_t inA1 = (int8_t) bitext((int) *pIn, 2, 2);
+    int8_t inA2 = (int8_t) bitext((int) *pIn, 2, 4);
+    int8_t inA3 = (int8_t) bitext((int) *pIn, 2, 6);
+    v4s inA4 = pack((int8_t) inA0, (int8_t) inA1, (int8_t) inA2, (int8_t) inA3);
+    int8_t inB0 = (int8_t) bitext((int) *pCom, 2, 0);
+    int8_t inB1 = (int8_t) bitext((int) *pCom, 2, 2);
+    int8_t inB2 = (int8_t) bitext((int) *pCom, 2, 4);
+    int8_t inB3 = (int8_t) bitext((int) *pCom, 2, 6);
+    v4s inB4 = pack((int8_t) inB0, (int8_t) inB1, (int8_t) inB2, (int8_t) inB3);
+
+    *((v4s*)out) = maxs4(inA4, inB4);
+
+    int8_t inA = (int8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((int8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_tnn_compare_and_replace_if_larger_ternary(int8_t * base,
+                                                int8_t * target,
+                                                uint16_t length)
+{
+  uint8_t mask2 = 0x0c;
+  uint8_t n_mask2 = ~ mask2;
+  uint8_t mask4 = 0x30;
+  uint8_t n_mask4 = ~ mask4;
+  uint8_t mask6 = 0xc0;
+  uint8_t n_mask6 = ~ mask6;
+  uint8_t off2 = 2;
+  uint8_t off4 = 4;
+  uint8_t off6 = 6;
+
+  uint8_t *pIn = (uint8_t *) base;
+  uint8_t *pCom = (uint8_t *) target;
+  uint8_t *out;
+
+  int cnt = length >> 2;
+  uint32_t result;
+
+  while(cnt > 0u)
+  {
+    uint32_t in1 = *((uint32_t *)pIn);
+    uint32_t in2 = *((int32_t *)pCom);
+    result = maxs20(in1, in2);
+    *((uint32_t *)pIn) = result;
+
+    pIn+=4;
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  if (left>0u)
+  {
+    // do the vector max on the whole word - we won't use the leftover bytes
+    uint32_t in1 = *((uint32_t *)pIn);
+    uint32_t in2 = *((int32_t *)pCom);
+    result = maxs20(in1, in2);
+
+    // ...and copy back the relevant bytes of the result to pIn
+    for (int i=0; i<left; i++)
+      *((uint8_t*)(pIn + i)) = (uint8_t) (result >> (8*i));
+
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_avg_and_replace_u2(uint8_t * base,
+                                  uint8_t * target,
+                                  uint16_t length)
+{
+  int8_t mask2 = 0x0c;
+  int8_t n_mask2 = ~ mask2;
+  int8_t mask4 = 0x30;
+  int8_t n_mask4 = ~ mask4;
+  int8_t mask6 = 0xc0;
+  int8_t n_mask6 = ~ mask6;
+  int8_t off2 = 2;
+  int8_t off4 = 4;
+  int8_t off6 = 6;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+  uint8_t *out;
+
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    *((uint32_t *)pIn) = avg16(*((uint32_t *)pIn), *((uint32_t *)pCom));
+
+    pIn+=4;
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 2, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 2, 2);
+    uint8_t inA2 = (uint8_t) bitextu((unsigned int) *pIn, 2, 4);
+    uint8_t inA3 = (uint8_t) bitextu((unsigned int) *pIn, 2, 6);
+    v4u inA4 = pack((uint8_t) inA0, (uint8_t) inA1, (uint8_t) inA2, (uint8_t) inA3);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 2, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 2, 2);
+    uint8_t inB2 = (uint8_t) bitextu((unsigned int) *pCom, 2, 4);
+    uint8_t inB3 = (uint8_t) bitextu((unsigned int) *pCom, 2, 6);
+    v4u inB4 = pack((uint8_t) inB0, (uint8_t) inB1, (uint8_t) inB2, (uint8_t) inB3);
+
+    *((v4u*)out) = avg4(inA4, inB4);
+
+    uint8_t inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+#endif
diff --git a/rt_nn_tests/xptnn_conv/pulp_nn_utils_xpnn.h b/rt_nn_tests/xptnn_conv/pulp_nn_utils_xpnn.h
new file mode 100644
index 0000000..0c783ae
--- /dev/null
+++ b/rt_nn_tests/xptnn_conv/pulp_nn_utils_xpnn.h
@@ -0,0 +1,1937 @@
+/*
+ * pulp_nn_utils.h
+ * Nazareno   Bruschi  <nazareno.bruschi@unibo.it>
+ * Alessandro Nadalini <alessandro.nadalini3@unibo.it>
+ * Georg Rutishauser   <georgr@iis.ee.ethz.ch>
+ *
+ * Copyright (C) 2019-2020 ETH Zurich & University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef __PULPNN_UTILS__
+#define __PULPNN_UTILS__
+
+#include "pmsis.h"
+#ifdef GAP_SDK
+#include "pulp.h"
+#endif
+
+#define bitext(x,size,off)                                      __builtin_pulp_bextract(x,size,off)
+#define bitextu(x,size,off)                                     __builtin_pulp_bextractu(x,size,off)
+#define bitins(dst,not_mask_imm,src,mask_imm,off)               __builtin_pulp_binsert(dst,not_mask_imm,src,mask_imm,off)
+#define pack(x,y,z,t)                                           __builtin_pulp_pack4(x,y,z,t)
+#define max4(a,b)                                               __builtin_pulp_maxu4(a,b)
+#define maxs4(a, b)                                             __builtin_pulp_max4(a, b)
+#define max8(a, b)                                              __builtin_pulp_maxu8(a, b)
+#define maxs8(a, b)                                             __builtin_pulp_max8(a, b)
+#define max16(a, b)                                             __builtin_pulp_maxu16(a, b)
+#define maxs16(a, b)                                            __builtin_pulp_max16(a, b)
+#define max32(a,b)                                              __builtin_pulp_maxusi(a,b)
+#define maxs32(a,b)                                             __builtin_pulp_maxsi(a,b)
+#define min32(a,b)                                              __builtin_pulp_minusi(a,b)
+#define mins32(a,b)                                             __builtin_pulp_minsi(a,b)
+#define min4(a, b)                                              __builtin_pulp_minu4(a, b)
+#define mins4(a, b)                                             __builtin_pulp_min4(a, b)
+#define min8(a, b)                                              __builtin_pulp_minu8(a, b)
+#define mins8(a, b)                                             __builtin_pulp_min8(a, b)
+#define min16(a, b)                                             __builtin_pulp_minu16(a, b)
+#define mins16(a, b)                                            __builtin_pulp_min16(a, b)
+#define avg4(a,b)                                               __builtin_pulp_avgu4(a,b)
+#define avg8(a,b)                                               __builtin_pulp_avgu8(a,b)
+#define avg16(a,b)                                              __builtin_pulp_avgu16(a,b)
+#define log2(x)                                                 __builtin_pulp_fl1(x)
+#define min(a,b)                                                ((a)<(b)?(a):(b))
+#define SumDotp4(a, b, c)                                       __builtin_pulp_sdotusp4(a, b, c)
+#define SumDotp8(a, b, c)                                       __builtin_pulp_sdotusp8(a, b, c)
+#define SumDotp16(a, b, c)                                      __builtin_pulp_sdotusp16(a, b, c)
+#define SumDotps4(a, b, c)                                      __builtin_pulp_sdotsp4(a, b, c)
+#define SumDotps8(a, b, c)                                      __builtin_pulp_sdotsp8(a, b, c)
+#define SumDotps16(a, b, c)                                     __builtin_pulp_sdotsp16(a, b, c)
+#define clip4(x)                                                __builtin_pulp_clipu_r(x, 15)
+#define clip2(x)                                                __builtin_pulp_clipu_r(x, 3)
+#define clip8(x)                                                __builtin_pulp_clipu_r(x, 255)
+
+#define clips4(x)                                               __builtin_pulp_clip_r(x, 7)
+#define clips2(x)                                               __builtin_pulp_clip_r(x, 1)
+#define clips8(x)                                               __builtin_pulp_clip_r(x, 127)
+#define MacLoadInit(a_update, b_update, a_reg, b_reg, ptr)      __builtin_pulp_mlinitspr_v3(a_update, b_update, a_reg, b_reg, ptr)
+#define MacLoadUpdate(ptr)                                      __builtin_pulp_mlupdatespr_v3(ptr)
+#define MacLoadAssign(ptr)                                      __builtin_pulp_mlassignspr_v3(ptr)
+#define MacLoad4(a_update, b_update, a_reg, b_reg, ptr, sum)    __builtin_pulp_mlsdotsup4_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+#define MacLoad8(a_update, b_update, a_reg, b_reg, ptr, sum)    __builtin_pulp_mlsdotsup8_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+#define MacLoad16(a_update, b_update, a_reg, b_reg, ptr, sum)   __builtin_pulp_mlsdotsup16_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+#define MacLoads4(a_update, b_update, a_reg, b_reg, ptr, sum)   __builtin_pulp_mlsdotsp4_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+#define MacLoads8(a_update, b_update, a_reg, b_reg, ptr, sum)   __builtin_pulp_mlsdotsp8_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+#define MacLoads16(a_update, b_update, a_reg, b_reg, ptr, sum)  __builtin_pulp_mlsdotsp16_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+
+#define PACK_INT8_SIZE(x)                                    (x)
+#define PACK_INT4_SIZE(x)                                    ((x) >> 1)
+#define PACK_INT2_SIZE(x)                                    ((x) >> 2)
+
+#define MemoryFence()                                        asm volatile("":::"memory")
+
+#define LEGACY_MODE(x)                                       asm volatile ("csrwi 0x010," x)
+#define IVEC_FMT(x)                                          asm volatile ("csrwi 0x00D," x)
+#define MIXED_SKIP(x)                                        asm volatile ("csrwi 0x00F," x)
+#define A_ADDRESS(x)                                         asm volatile ("csrw 0x100, %0" :: "r" (x))
+#define W_ADDRESS(x)                                         asm volatile ("csrw 0x101, %0" :: "r" (x))
+#define A_STRIDE(x)                                          asm volatile ("csrw 0x102, %0":: "r" (x))
+#define W_STRIDE(x)                                          asm volatile ("csrw 0x103, %0":: "r" (x))
+#define A_ROLLBACK(x)                                        asm volatile ("csrw 0x104, %0":: "r" (x))
+#define W_ROLLBACK(x)                                        asm volatile ("csrw 0x105, %0":: "r" (x))
+#define A_SKIP(x)                                            asm volatile ("csrwi 0x106," x)
+#define W_SKIP(x)                                            asm volatile ("csrwi 0x107," x)
+
+static uint8_t __attribute__((noinline)) pulp_nn_quant_u2 (
+  int32_t phi,
+  int16_t m,
+  int8_t d
+  ) {
+  int32_t x = (m * phi) >> d;
+  uint8_t res = clip2(x);
+  return res;
+}
+static uint8_t __attribute__((noinline)) pulp_nn_bn_quant_u2 (
+  int32_t phi,
+  int32_t k,
+  int32_t lambda,
+  int8_t d
+  ) {
+  int32_t integer_image_phi = (k * phi) + lambda;
+  int32_t x = (integer_image_phi) >> d;
+  uint8_t res = clip2(x);
+  return res;
+}
+static uint8_t __attribute__((noinline)) pulp_nn_add_quant_u2 (
+  uint8_t pix1,
+  uint8_t pix2,
+  int16_t m1,
+  int16_t m2,
+  int8_t d
+  ) {
+  uint32_t integer_image = pix1*m1 + pix2*m2;
+  uint32_t x = (integer_image) >> d;
+  uint8_t res = clip2(x);
+  return res;
+}
+
+static int8_t __attribute__((noinline)) pulp_nn_quant_i2 (
+  int32_t phi,
+  int16_t m,
+  int8_t d
+  ) {
+  int32_t x = (m * phi) >> d;
+  int8_t res = clips2(x);
+  return res;
+}
+static int8_t __attribute__((noinline)) pulp_nn_bn_quant_i2 (
+  int32_t phi,
+  int32_t k,
+  int32_t lambda,
+  int8_t d
+  ) {
+  int32_t integer_image_phi = (k * phi) + lambda;
+  int32_t x = (integer_image_phi) >> d;
+  int8_t res = clips2(x);
+  return res;
+}
+static int8_t __attribute__((noinline)) pulp_nn_add_quant_i2 (
+  int8_t pix1,
+  int8_t pix2,
+  int16_t m1,
+  int16_t m2,
+  int8_t d
+  ) {
+  int32_t integer_image = pix1*m1 + pix2*m2;
+  int32_t x = (integer_image) >> d;
+  int8_t res = clips2(x);
+  return res;
+}
+
+static uint8_t __attribute__((noinline)) pulp_nn_quant_u4 (
+  int32_t phi,
+  int16_t m,
+  int8_t d
+  ) {
+  int32_t x = (m * phi) >> d;
+  uint8_t res = clip4(x);
+  return res;
+}
+static uint8_t __attribute__((noinline)) pulp_nn_bn_quant_u4 (
+  int32_t phi,
+  int32_t k,
+  int32_t lambda,
+  int8_t d
+  ) {
+  int32_t integer_image_phi = (k * phi) + lambda;
+  int32_t x = (integer_image_phi) >> d;
+  uint8_t res = clip4(x);
+  return res;
+}
+static uint8_t __attribute__((noinline)) pulp_nn_add_quant_u4 (
+  uint8_t pix1,
+  uint8_t pix2,
+  int16_t m1,
+  int16_t m2,
+  int8_t d
+  ) {
+  uint32_t integer_image = pix1*m1 + pix2*m2;
+  uint32_t x = (integer_image) >> d;
+  uint8_t res = clip4(x);
+  return res;
+}
+
+static int8_t __attribute__((noinline)) pulp_nn_quant_i4 (
+  int32_t phi,
+  int16_t m,
+  int8_t d
+  ) {
+  int32_t x = (m * phi) >> d;
+  int8_t res = clips4(x);
+  return res;
+}
+static int8_t __attribute__((noinline)) pulp_nn_bn_quant_i4 (
+  int32_t phi,
+  int32_t k,
+  int32_t lambda,
+  int8_t d
+  ) {
+  int32_t integer_image_phi = (k * phi) + lambda;
+  int32_t x = (integer_image_phi) >> d;
+  int8_t res = clips4(x);
+  return res;
+}
+static int8_t __attribute__((noinline)) pulp_nn_add_quant_i4 (
+  int8_t pix1,
+  int8_t pix2,
+  int16_t m1,
+  int16_t m2,
+  int8_t d
+  ) {
+  int32_t integer_image = pix1*m1 + pix2*m2;
+  int32_t x = (integer_image) >> d;
+  int8_t res = clips4(x);
+  return res;
+}
+
+static uint8_t __attribute__((noinline)) pulp_nn_quant_u8 (
+  int32_t phi,
+  int16_t m,
+  int8_t d
+  ) {
+  int32_t x = (m * phi) >> d;
+  uint8_t res = clip8(x);
+  return res;
+}
+static uint8_t __attribute__((noinline)) pulp_nn_bn_quant_u8 (
+  int32_t phi,
+  int32_t k,
+  int32_t lambda,
+  int8_t d
+  ) {
+  int32_t integer_image_phi = (k * phi) + lambda;
+  int32_t x = (integer_image_phi) >> d;
+  uint8_t res = clip8(x);
+  return res;
+}
+static uint8_t __attribute__((noinline)) pulp_nn_add_quant_u8 (
+  uint8_t pix1,
+  uint8_t pix2,
+  int16_t m1,
+  int16_t m2,
+  int8_t d
+  ) {
+  uint32_t integer_image = pix1*m1 + pix2*m2;
+  uint32_t x = (integer_image) >> d;
+  uint8_t res = clip8(x);
+  return res;
+}
+
+static int8_t __attribute__((noinline)) pulp_nn_quant_i8 (
+  int32_t phi,
+  int16_t m,
+  int8_t d
+  ) {
+  int32_t x = (m * phi) >> d;
+  int8_t res = clips8(x);
+  return res;
+}
+static int8_t __attribute__((noinline)) pulp_nn_bn_quant_i8 (
+  int32_t phi,
+  int32_t k,
+  int32_t lambda,
+  int8_t d
+  ) {
+  int32_t integer_image_phi = (k * phi) + lambda;
+  int32_t x = (integer_image_phi) >> d;
+  int8_t res = clips8(x);
+  return res;
+}
+static int8_t __attribute__((noinline)) pulp_nn_add_quant_i8 (
+  int8_t pix1,
+  int8_t pix2,
+  int16_t m1,
+  int16_t m2,
+  int8_t d
+  ) {
+  int32_t integer_image = pix1*m1 + pix2*m2;
+  int32_t x = (integer_image) >> d;
+  int8_t res = clips8(x);
+  return res;
+}
+
+
+static uint8_t __attribute__((noinline)) pulp_nn_u4_quant(int input, int16_t * pThr)
+{
+  if(input <= pThr[7] )
+  {
+    if( input <= pThr[3])
+    {
+      if( input <= pThr[1])
+      {
+        if( input <= pThr[0])
+          return 0;
+        else
+          return 1;
+      }
+      else
+      {
+        if( input <= pThr[2])
+          return 2;
+        else
+          return 3;
+      }
+    }
+    else
+    {
+      if( input <= pThr[5])
+      {
+        if( input <= pThr[4])
+          return 4;
+        else
+          return 5;
+      }
+      else
+      {
+        if( input <= pThr[6])
+          return 6;
+        else
+          return 7;
+      }
+    }
+  }
+  else
+  {
+    if( input <= pThr[11])
+    {
+      if( input <= pThr[9])
+      {
+        if( input <= pThr[8])
+          return 8;
+        else
+          return 9;
+      }
+      else
+      {
+        if( input <= pThr[10])
+          return 10;
+        else
+          return 11;
+      }
+    }
+    else
+    {
+      if( input <= pThr[13])
+      {
+        if( input <= pThr[12])
+          return 12;
+        else
+          return 13;
+      }
+      else
+      {
+        if( input <= pThr[14])
+          return 14;
+        else
+          return 15;
+      }
+    }
+  }
+}
+
+static uint8_t __attribute__((noinline)) pulp_nn_u2_quant(int input, int16_t * pThr)
+{
+  if( input <= pThr[1])
+  {
+    if( input <= pThr[0])
+        {
+      return 0;
+    }
+        else
+        {
+      return 1;
+    }
+  }
+    else
+    {
+    if( input <= pThr[2])
+        {
+      return 2;
+    }
+        else
+        {
+      return 3;
+    }
+  }
+}
+
+/*
+ * Common
+ */
+
+
+static v4s __attribute__((noinline)) pulp_nn_i4_to_i8_r( int8_t *pSrc)
+{
+  v4s Src = *((v4s*) pSrc);
+  int8_t bext1, bext2, bext3, bext4;
+
+  bext1 = (int8_t) bitext((int) Src, 4, 0);
+  bext2 = (int8_t) bitext((int) Src, 4, 4);
+  bext3 = (int8_t) bitext((int) Src, 4, 8);
+  bext4 = (int8_t) bitext((int) Src, 4, 12);
+  v4s res = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+
+  return res;
+}
+
+static v4u __attribute__((noinline)) pulp_nn_u4_to_u8_r(uint8_t *pSrc)
+{
+  v4u Src = *((v4u*) pSrc);
+  uint8_t bext1, bext2, bext3, bext4;
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 4, 0);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 4, 4);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 4, 8);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 4, 12);
+  v4u res = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+
+  return res;
+}
+
+static v4s __attribute__((noinline)) pulp_nn_i2_to_i8_r( int8_t *pSrc)
+{
+  v4s Src = *((v4s*) pSrc);
+  int8_t bext1, bext2, bext3, bext4;
+
+  bext1 = (int8_t) bitext((int) Src, 2, 0);
+  bext2 = (int8_t) bitext((int) Src, 2, 2);
+  bext3 = (int8_t) bitext((int) Src, 2, 4);
+  bext4 = (int8_t) bitext((int) Src, 2, 6);
+  v4s res = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+
+  return res;
+}
+
+static v4u __attribute__((noinline)) pulp_nn_u2_to_u8_r(uint8_t *pSrc)
+{
+  v4u Src = *((v4u*) pSrc);
+  uint8_t bext1, bext2, bext3, bext4;
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 0);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 2);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 4);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 6);
+  v4u res = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+
+  return res;
+}
+
+static v4s __attribute__((noinline)) pulp_nn_i2_to_i4_r(int8_t *pSrc)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  v4s Src = *((v4s*) pSrc);
+  int8_t bext1, bext2, bext3, bext4;
+  int8_t out1, out2, out3, out4;
+
+  bext1 = (int8_t) bitextu((int) Src, 2, 0);
+  bext2 = (int8_t) bitextu((int) Src, 2, 2);
+  bext3 = (int8_t) bitextu((int) Src, 2, 4);
+  bext4 = (int8_t) bitextu((int) Src, 2, 6);
+
+  out1 = bitins(bext1, n_mask, bext2, mask, off);
+  out2 = bitins(bext3, n_mask, bext4, mask, off);
+
+  bext1 = (int8_t) bitextu((int) Src, 2, 8);
+  bext2 = (int8_t) bitextu((int) Src, 2, 10);
+  bext3 = (int8_t) bitextu((int) Src, 2, 12);
+  bext4 = (int8_t) bitextu((int) Src, 2, 14);
+
+  out3 = bitins(bext1, n_mask, bext2, mask, off);
+  out4 = bitins(bext3, n_mask, bext4, mask, off);
+
+  v4s res = pack((int8_t) out1, (int8_t) out2, (int8_t) out3, (int8_t) out4);
+
+  return res;
+}
+
+static v4u __attribute__((noinline)) pulp_nn_u2_to_u4_r(uint8_t *pSrc)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  v4u Src = *((v4u*) pSrc);
+  uint8_t bext1, bext2, bext3, bext4;
+  uint8_t out1, out2, out3, out4;
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 0);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 2);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 4);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 6);
+
+  out1 = bitins(bext1, n_mask, bext2, mask, off);
+  out2 = bitins(bext3, n_mask, bext4, mask, off);
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 8);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 10);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 12);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 14);
+
+  out3 = bitins(bext1, n_mask, bext2, mask, off);
+  out4 = bitins(bext3, n_mask, bext4, mask, off);
+
+  v4u res = pack((uint8_t) out1, (uint8_t) out2, (uint8_t) out3, (uint8_t) out4);
+
+  return res;
+}
+
+static int8_t *__attribute__((always_inline)) pulp_nn_i4_to_i8( int8_t *pSrc, int8_t *pDst)
+{
+  v4s Src = *((v4s*) pSrc);
+  int8_t bext1, bext2, bext3, bext4;
+
+  pSrc+=4;
+
+  bext1 = (int8_t) bitext((int) Src, 4, 0);
+  bext2 = (int8_t) bitext((int) Src, 4, 4);
+  bext3 = (int8_t) bitext((int) Src, 4, 8);
+  bext4 = (int8_t) bitext((int) Src, 4, 12);
+  *((v4s*)pDst) = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+   MemoryFence();
+  pDst+=4;
+  bext1 = (int8_t) bitext((int) Src, 4, 16);
+  bext2 = (int8_t) bitext((int) Src, 4, 20);
+  bext3 = (int8_t) bitext((int) Src, 4, 24);
+  bext4 = (int8_t) bitext((int) Src, 4, 28);
+  *((v4s*)pDst) = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+
+  return pSrc;
+}
+
+static uint8_t *__attribute__((always_inline)) pulp_nn_u4_to_u8(uint8_t *pSrc, uint8_t *pDst)
+{
+  v4u Src = *((v4u*) pSrc);
+  uint8_t bext1, bext2, bext3, bext4;
+
+  pSrc+=4;
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 4, 0);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 4, 4);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 4, 8);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 4, 12);
+  *((v4u*)pDst) = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 4, 16);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 4, 20);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 4, 24);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 4, 28);
+  *((v4u*)pDst) = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+
+  return pSrc;
+}
+
+static int8_t *__attribute__((always_inline)) pulp_nn_i2_to_i8( int8_t * pSrc, int8_t * pDst)
+{
+  v4s Src = *((v4s*) pSrc);
+  int8_t bext1, bext2, bext3, bext4;
+
+  pSrc+=4;
+
+  bext1 = (int8_t) bitext((int) Src, 2, 0);
+  bext2 = (int8_t) bitext((int) Src, 2, 2);
+  bext3 = (int8_t) bitext((int) Src, 2, 4);
+  bext4 = (int8_t) bitext((int) Src, 2, 6);
+  *((v4s*)pDst) = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (int8_t) bitext((int) Src, 2, 8);
+  bext2 = (int8_t) bitext((int) Src, 2, 10);
+  bext3 = (int8_t) bitext((int) Src, 2, 12);
+  bext4 = (int8_t) bitext((int) Src, 2, 14);
+  *((v4s*)pDst) = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (int8_t) bitext((int) Src, 2, 16);
+  bext2 = (int8_t) bitext((int) Src, 2, 18);
+  bext3 = (int8_t) bitext((int) Src, 2, 20);
+  bext4 = (int8_t) bitext((int) Src, 2, 22);
+  *((v4s*)pDst) = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (int8_t) bitext((int) Src, 2, 24);
+  bext2 = (int8_t) bitext((int) Src, 2, 26);
+  bext3 = (int8_t) bitext((int) Src, 2, 28);
+  bext4 = (int8_t) bitext((int) Src, 2, 30);
+  *((v4s*)pDst) = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+
+  return pSrc;
+}
+
+static uint8_t *__attribute__((always_inline)) pulp_nn_u2_to_u8(uint8_t * pSrc, uint8_t * pDst)
+{
+  v4u Src = *((v4u*) pSrc);
+  uint8_t bext1, bext2, bext3, bext4;
+
+  pSrc+=4;
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 0);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 2);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 4);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 6);
+  *((v4u*)pDst) = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 8);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 10);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 12);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 14);
+  *((v4u*)pDst) = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 16);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 18);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 20);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 22);
+  *((v4u*)pDst) = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 24);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 26);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 28);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 30);
+  *((v4u*)pDst) = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+
+  return pSrc;
+}
+
+static int8_t *__attribute__((always_inline)) pulp_nn_i2_to_i4( int8_t * pSrc, int8_t * pDst)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  v4s Src = *((v4s*) pSrc);
+  int8_t bext1, bext2, bext3, bext4;
+  int8_t out1, out2, out3, out4;
+
+  pSrc+=4;
+
+  bext1 = (int8_t) bitext((int) Src, 2, 0);
+  bext2 = (int8_t) bitext((int) Src, 2, 2);
+  bext3 = (int8_t) bitext((int) Src, 2, 4);
+  bext4 = (int8_t) bitext((int) Src, 2, 6);
+
+  out1 = bitins(bext1, n_mask, bext2, mask, off);
+  out2 = bitins(bext3, n_mask, bext4, mask, off);
+  MemoryFence();
+
+  bext1 = (int8_t) bitext((int) Src, 2, 8);
+  bext2 = (int8_t) bitext((int) Src, 2, 10);
+  bext3 = (int8_t) bitext((int) Src, 2, 12);
+  bext4 = (int8_t) bitext((int) Src, 2, 14);
+
+  out3 = bitins(bext1, n_mask, bext2, mask, off);
+  out4 = bitins(bext3, n_mask, bext4, mask, off);
+
+  *((v4s*)pDst) = pack((int8_t) out1, (int8_t) out2, (int8_t) out3, (int8_t) out4);
+  MemoryFence();
+
+  pDst+=4;
+  bext1 = (int8_t) bitext((int) Src, 2, 16);
+  bext2 = (int8_t) bitext((int) Src, 2, 18);
+  bext3 = (int8_t) bitext((int) Src, 2, 20);
+  bext4 = (int8_t) bitext((int) Src, 2, 22);
+
+  out1 = bitins(bext1, n_mask, bext2, mask, off);
+  out2 = bitins(bext3, n_mask, bext4, mask, off);
+  MemoryFence();
+
+  bext1 = (int8_t) bitext((int) Src, 2, 24);
+  bext2 = (int8_t) bitext((int) Src, 2, 26);
+  bext3 = (int8_t) bitext((int) Src, 2, 28);
+  bext4 = (int8_t) bitext((int) Src, 2, 30);
+
+  out3 = bitins(bext1, n_mask, bext2, mask, off);
+  out4 = bitins(bext3, n_mask, bext4, mask, off);
+
+  *((v4s*)pDst) = pack((int8_t) out1, (int8_t) out2, (int8_t) out3, (int8_t) out4);
+
+  return pSrc;
+}
+
+static uint8_t *__attribute__((always_inline)) pulp_nn_u2_to_u4( uint8_t * pSrc, uint8_t * pDst)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  v4u Src = *((v4u*) pSrc);
+  uint8_t bext1, bext2, bext3, bext4;
+  uint8_t out1, out2, out3, out4;
+
+  pSrc+=4;
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 0);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 2);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 4);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 6);
+
+  out1 = bitins(bext1, n_mask, bext2, mask, off);
+  out2 = bitins(bext3, n_mask, bext4, mask, off);
+  MemoryFence();
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 8);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 10);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 12);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 14);
+
+  out3 = bitins(bext1, n_mask, bext2, mask, off);
+  out4 = bitins(bext3, n_mask, bext4, mask, off);
+
+  *((v4u*)pDst) = pack((uint8_t) out1, (uint8_t) out2, (uint8_t) out3, (uint8_t) out4);
+  MemoryFence();
+
+  pDst+=4;
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 16);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 18);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 20);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 22);
+
+  out1 = bitins(bext1, n_mask, bext2, mask, off);
+  out2 = bitins(bext3, n_mask, bext4, mask, off);
+  MemoryFence();
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 24);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 26);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 28);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 30);
+
+  out3 = bitins(bext1, n_mask, bext2, mask, off);
+  out4 = bitins(bext3, n_mask, bext4, mask, off);
+  
+  *((v4u*)pDst) = pack((uint8_t) out1, (uint8_t) out2, (uint8_t) out3, (uint8_t) out4);
+
+  return pSrc;
+}
+
+/*
+ * XpulpV2
+ */
+
+static void __attribute__((noinline)) pulp_zero_mem(uint8_t * pBuffer, unsigned int size)
+{
+  int lfover = size &0x3;
+  for (int i=0; i<(size>>2); i++)
+  {
+    *((v4u *)pBuffer) = (v4u){0,0,0,0};
+    MemoryFence();
+    pBuffer+=4;
+  }
+  while(lfover)
+  {
+    *pBuffer++=0;
+    lfover--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_im2col_u2_to_u8(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_u2_to_u8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 16;
+  }
+  while (lfover)
+  {
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 0);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 2);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 4);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 6);
+    pOutput++;
+    pInput++;
+    lfover -= 4;
+  }
+}
+static void __attribute__((noinline)) pulp_nn_im2col_i2_to_i8(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_i2_to_i8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 16;
+  }
+  while (lfover)
+  {
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 0);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 2);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 4);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 6);
+    pOutput++;
+    pInput++;
+    lfover -= 4;
+  }
+}
+static void __attribute__((noinline)) pulp_nn_im2col_u4_to_u8(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 3u;
+  int lfover = blockSize & 0x07;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_u4_to_u8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 8;
+  }
+  while (lfover)
+  {
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 4, 0);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 4, 4);
+    pOutput++;
+    pInput++;
+    lfover -= 2;
+  }
+}
+static void __attribute__((noinline)) pulp_nn_im2col_i4_to_i8(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 3u;
+  int lfover = blockSize & 0x07;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_i4_to_i8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 8;
+  }
+  while (lfover)
+  {
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 4, 0);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 4, 4);
+    pOutput++;
+    pInput++;
+    lfover -= 2;
+  }
+}
+static void __attribute__((noinline)) pulp_nn_im2col_u8_to_u8(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 2u;
+  int lfover = blockSize & 0x03;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4u*)pOutput) = *((v4u*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    *((uint8_t*)pOutput) = *((uint8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover--;
+  }
+}
+static void __attribute__((noinline)) pulp_nn_im2col_i8_to_i8(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 2u;
+  int lfover = blockSize & 0x03;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4s*)pOutput) = *((v4s*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    *((int8_t*)pOutput) = *((int8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover--;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_im2col_u2_to_u2(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4u*)pOutput) = *((v4u*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    uint8_t extr;
+    *((uint8_t*)pOutput) = *((uint8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover -= 4;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_i2_to_i2(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4s*)pOutput) = *((v4s*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    int8_t extr;
+    *((int8_t*)pOutput) = *((int8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover -= 4;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_u2_to_u4(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_u2_to_u4(pInput, pOutput);
+    MemoryFence();
+    pOutput += 8;
+  }
+  while (lfover)
+  {
+    uint8_t extr;
+    // TODO: this is potentially dangerous/wrong if the number of channels
+    // is not a multiple of 8!
+    *((v4u *)pOutput) = pulp_nn_u2_to_u4_r(pInput);
+    pInput += 2;
+    pOutput += 4;
+    lfover -= 8;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_i2_to_i4(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_i2_to_i4(pInput, pOutput);
+    MemoryFence();
+    pOutput += 8;
+  }
+  while (lfover)
+  {
+    int8_t extr;
+    // TODO: this is potentially dangerous/wrong if the number of channels
+    // is not a multiple of 8!
+    *((v4s *)pOutput) = pulp_nn_i2_to_i4_r(pInput);
+    pInput += 2;
+    pOutput += 4;
+    lfover -= 8;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_u2_to_u8(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_u2_to_u8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 16;
+  }
+  while (lfover)
+  {
+    uint8_t extr;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 0);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 2);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 4);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 6);
+    pOutput++;
+    pInput++;
+    lfover -= 4;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_i2_to_i8(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_i2_to_i8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 16;
+  }
+  while (lfover)
+  {
+    int8_t extr;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 0);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 2);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 4);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 6);
+    pOutput++;
+    pInput++;
+    lfover -= 4;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_u4_to_u4(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 3u;
+  int lfover = blockSize & 0x07;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4u*)pOutput) = *((v4u*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    uint8_t extr;
+    *((uint8_t*)pOutput) = *((uint8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover -= 2;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_i4_to_i4(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 3u;
+  int lfover = blockSize & 0x07;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4s*)pOutput) = *((v4s*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    int8_t extr;
+    *((int8_t*)pOutput) = *((int8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover -= 2;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_u4_to_u8(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 3u;
+  int lfover = blockSize & 0x07;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_u4_to_u8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 8;
+  }
+  while (lfover)
+  {
+    uint8_t extr;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 4, 0);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 4, 4);
+    pOutput++;
+    pInput++;
+    lfover -= 2;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_i4_to_i8(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 3u;
+  int lfover = blockSize & 0x07;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_i4_to_i8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 8;
+  }
+  while (lfover)
+  {
+    int8_t extr;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 4, 0);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 4, 4);
+    pOutput++;
+    pInput++;
+    lfover -= 2;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_u8_to_u8(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 2u;
+  int lfover = blockSize & 0x03;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4u*)pOutput) = *((v4u*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    uint8_t extr;
+    *((uint8_t*)pOutput) = *((uint8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover -= 1;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_i8_to_i8(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 2u;
+  int lfover = blockSize & 0x03;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4s*)pOutput) = *((v4s*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    int8_t extr;
+    *((int8_t*)pOutput) = *((int8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover -= 1;
+  }
+}
+
+
+
+static void __attribute__((noinline)) pulp_nn_compare_and_replace_if_larger_u8(uint8_t * base,
+                                                uint8_t * target,
+                                                uint16_t length)
+{
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+  v4u inp;
+  v4u com;
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    inp = *((v4u*)pIn);
+    com = *((v4u*)pCom);
+
+    *((v4u*)pIn) = max4(inp, com);
+
+    pCom+=4;
+    pIn+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    if(*pIn<*pCom)
+      *pIn=*pCom;
+    
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_compare_and_replace_if_larger_i8(
+    int8_t * base, int8_t * target, uint16_t length) {
+  int8_t *pIn = base;
+  int8_t *pCom = target;
+  v4s inp;
+  v4s com;
+  int cnt = length >> 2;
+
+  while (cnt > 0u) {
+    inp = *((v4s *)pIn);
+    com = *((v4s *)pCom);
+
+    *((v4s *)pIn) = maxs4(inp, com);
+
+    pCom += 4;
+    pIn += 4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left > 0u) {
+    if (*pIn < *pCom)
+      *pIn = *pCom;
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_avg_and_replace_u8(uint8_t * base,
+                                  uint8_t * target,
+                                  uint16_t length)
+{
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+
+  while (length>0u)
+  {
+    *pIn = ((*pIn + *pCom) >> 1);
+
+    pIn++;
+    pCom++;
+    length--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_compare_and_replace_if_larger_u4(uint8_t * base,
+                                                uint8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+  v4u inp[2];
+  v4u com[2];
+  uint8_t *out;
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    pulp_nn_u4_to_u8(pIn, (uint8_t *)inp);
+    pulp_nn_u4_to_u8(pCom, (uint8_t *)com);
+
+    *((v4u *)out) = max4(inp[0], com[0]);
+
+    *((uint8_t*)pIn) = bitins(*out, n_mask, *(out + 1), mask, off);
+    pIn++;
+    *((uint8_t*)pIn) = bitins(*(out + 2), n_mask, *(out + 3), mask, off);
+    pIn++;
+
+    *((v4u *)out) = max4(inp[1], com[1]);
+
+    *((uint8_t*)pIn) = bitins(*out, n_mask, *(out + 1), mask, off);
+    pIn++;
+    *((uint8_t*)pIn) = bitins(*(out + 2), n_mask, *(out + 3), mask, off);
+    pIn++;
+
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 4, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 4, 4);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 4, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 4, 4);
+
+    if(inA0<inB0)
+      inA0=inB0;
+
+    if(inA1<inB1)
+      inA1=inB1;
+
+    *((uint8_t*)pIn) = bitins(inA0, n_mask, inA1, mask, off);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_compare_and_replace_if_larger_i4(int8_t * base,
+                                                int8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  int8_t *pIn = base;
+  int8_t *pCom = target;
+  v4s inp[2];
+  v4s com[2];
+  int8_t *out;
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    pulp_nn_i4_to_i8(pIn, (int8_t *)inp);
+    pulp_nn_i4_to_i8(pCom, (int8_t *)com);
+
+    *((v4s *)out) = maxs4(inp[0], com[0]);
+
+    *((int8_t*)pIn) = bitins(*out, n_mask, *(out + 1), mask, off);
+    pIn++;
+    *((int8_t*)pIn) = bitins(*(out + 2), n_mask, *(out + 3), mask, off);
+    pIn++;
+
+    *((v4s *)out) = maxs4(inp[1], com[1]);
+
+    *((int8_t*)pIn) = bitins(*out, n_mask, *(out + 1), mask, off);
+    pIn++;
+    *((int8_t*)pIn) = bitins(*(out + 2), n_mask, *(out + 3), mask, off);
+    pIn++;
+
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    int8_t inA0 = (int8_t) bitext((int) *pIn, 4, 0);
+    int8_t inA1 = (int8_t) bitext((int) *pIn, 4, 4);
+    int8_t inB0 = (int8_t) bitext((int) *pCom, 4, 0);
+    int8_t inB1 = (int8_t) bitext((int) *pCom, 4, 4);
+
+    if(inA0<inB0)
+      inA0=inB0;
+
+    if(inA1<inB1)
+      inA1=inB1;
+
+    *((int8_t*)pIn) = bitins(inA0, n_mask, inA1, mask, off);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_avg_and_replace_u4(uint8_t * base,
+                                  uint8_t * target,
+                                  uint16_t length)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+
+  while (length>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 4, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 4, 4);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 4, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 4, 4);
+
+    inA0 = ((inA0 + inB0) >> 1);
+    inA1 = ((inA1 + inB1) >> 1);
+
+    *((uint8_t*)pIn) = bitins(inA0, n_mask, inA1, mask, off);
+
+    pIn++;
+    pCom++;
+    length--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_compare_and_replace_if_larger_u2(uint8_t * base,
+                                                uint8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask2 = 0x0c;
+  int8_t n_mask2 = ~ mask2;
+  int8_t mask4 = 0x30;
+  int8_t n_mask4 = ~ mask4;
+  int8_t mask6 = 0xc0;
+  int8_t n_mask6 = ~ mask6;
+  int8_t off2 = 2;
+  int8_t off4 = 4;
+  int8_t off6 = 6;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+  v4u inp[4];
+  v4u com[4];
+  uint8_t *out;
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    pulp_nn_u2_to_u8(pIn, inp);
+    pulp_nn_u2_to_u8(pCom, com);
+
+    *((v4u*)out) = max4(inp[0], com[0]);
+
+    uint8_t inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    *((v4u*)out) = max4(inp[1], com[1]);
+
+    inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    *((v4u*)out) = max4(inp[2], com[2]);
+
+    inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    *((v4u*)out) = max4(inp[3], com[3]);
+
+    inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 2, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 2, 2);
+    uint8_t inA2 = (uint8_t) bitextu((unsigned int) *pIn, 2, 4);
+    uint8_t inA3 = (uint8_t) bitextu((unsigned int) *pIn, 2, 6);
+    v4u inA4 = pack((uint8_t) inA0, (uint8_t) inA1, (uint8_t) inA2, (uint8_t) inA3);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 2, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 2, 2);
+    uint8_t inB2 = (uint8_t) bitextu((unsigned int) *pCom, 2, 4);
+    uint8_t inB3 = (uint8_t) bitextu((unsigned int) *pCom, 2, 6);
+    v4u inB4 = pack((uint8_t) inB0, (uint8_t) inB1, (uint8_t) inB2, (uint8_t) inB3);
+
+    *((v4u*)out) = max4(inA4, inB4);
+
+    uint8_t inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+
+static void __attribute__((noinline)) pulp_nn_compare_and_replace_if_larger_i2(int8_t * base,
+                                                int8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask2 = 0x0c;
+  int8_t n_mask2 = ~ mask2;
+  int8_t mask4 = 0x30;
+  int8_t n_mask4 = ~ mask4;
+  int8_t mask6 = 0xc0;
+  int8_t n_mask6 = ~ mask6;
+  int8_t off2 = 2;
+  int8_t off4 = 4;
+  int8_t off6 = 6;
+
+  int8_t *pIn = base;
+  int8_t *pCom = target;
+  v4s inp[4];
+  v4s com[4];
+  int8_t *out;
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    pulp_nn_i2_to_i8(pIn, inp);
+    pulp_nn_i2_to_i8(pCom, com);
+
+    *((v4s*)out) = maxs4(inp[0], com[0]);
+
+    int8_t inA = (int8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((int8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    *((v4s*)out) = maxs4(inp[1], com[1]);
+
+    inA = (int8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((int8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    *((v4s*)out) = maxs4(inp[2], com[2]);
+
+    inA = (int8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((int8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    *((v4s*)out) = maxs4(inp[3], com[3]);
+
+    inA = (int8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((int8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    int8_t inA0 = (int8_t) bitext((unsigned int) *pIn, 2, 0);
+    int8_t inA1 = (int8_t) bitext((unsigned int) *pIn, 2, 2);
+    int8_t inA2 = (int8_t) bitext((unsigned int) *pIn, 2, 4);
+    int8_t inA3 = (int8_t) bitext((unsigned int) *pIn, 2, 6);
+    v4s inA4 = pack((int8_t) inA0, (int8_t) inA1, (int8_t) inA2, (int8_t) inA3);
+    int8_t inB0 = (int8_t) bitext((unsigned int) *pCom, 2, 0);
+    int8_t inB1 = (int8_t) bitext((unsigned int) *pCom, 2, 2);
+    int8_t inB2 = (int8_t) bitext((unsigned int) *pCom, 2, 4);
+    int8_t inB3 = (int8_t) bitext((unsigned int) *pCom, 2, 6);
+    v4s inB4 = pack((int8_t) inB0, (int8_t) inB1, (int8_t) inB2, (int8_t) inB3);
+
+    *((v4s*)out) = maxs4(inA4, inB4);
+
+    int8_t inA = (int8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((int8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_avg_and_replace_u2(uint8_t * base,
+                                  uint8_t * target,
+                                  uint16_t length)
+{
+  int8_t mask2 = 0x0c;
+  int8_t n_mask2 = ~ mask2;
+  int8_t mask4 = 0x30;
+  int8_t n_mask4 = ~ mask4;
+  int8_t mask6 = 0xc0;
+  int8_t n_mask6 = ~ mask6;
+  int8_t off2 = 2;
+  int8_t off4 = 4;
+  int8_t off6 = 6;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+
+  while (length>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 2, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 2, 2);
+    uint8_t inA2 = (uint8_t) bitextu((unsigned int) *pIn, 2, 4);
+    uint8_t inA3 = (uint8_t) bitextu((unsigned int) *pIn, 2, 6);
+
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 2, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 2, 2);
+    uint8_t inB2 = (uint8_t) bitextu((unsigned int) *pCom, 2, 4);
+    uint8_t inB3 = (uint8_t) bitextu((unsigned int) *pCom, 2, 6);
+
+    inA0 = ((inA0 + inB0) >> 1);
+    inA1 = ((inA1 + inB1) >> 1);
+    inA2 = ((inA2 + inB2) >> 1);
+    inA3 = ((inA3 + inB3) >> 1);
+
+    uint8_t inA = (uint8_t) bitins(inA0, n_mask2, inA1, mask2, off2);
+    inA = bitins(inA, n_mask4, inA2, mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, inA3, mask6, off6);
+
+    pIn++;
+    pCom++;
+    length--;
+  }
+}
+
+/*
+ * XpulpNN
+ */
+
+static void __attribute__((noinline)) xpulp_nn_zero_mem_u8(uint8_t * pBuffer, unsigned int size)
+{
+  int lfover = size &0x3;
+  for (int i=0; i<(size>>2); i++)
+  {
+    *((v4u *)pBuffer) = (v4u){0,0,0,0};
+    MemoryFence();
+    pBuffer+=4;
+  }
+  while(lfover)
+  {
+    *pBuffer++=0;
+    lfover--;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_zero_mem_u4(uint8_t * pBuffer, unsigned int size)
+{
+  int lfover = size &0x7;
+  for (int i=0; i<(size>>3); i++)
+  {
+    *((v4u *)pBuffer) = (v4u){0,0,0,0};
+    MemoryFence();
+    pBuffer+=4;
+  }
+  while(lfover)
+  {
+    *pBuffer++=0;
+    lfover-=2;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_zero_mem_u2(uint8_t * pBuffer, unsigned int size)
+{
+  int lfover = size &0xf;
+  for (int i=0; i<(size>>4); i++)
+  {
+    *((v4u *)pBuffer) = (v4u){0,0,0,0};
+    MemoryFence();
+    pBuffer+=4;
+  }
+  while(lfover)
+  {
+    *pBuffer++=0;
+    lfover-=4;
+  }
+}
+
+
+static void __attribute__((noinline)) xpulp_nn_compare_and_replace_if_larger_u4(uint8_t * base,
+                                                uint8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    *((uint32_t *)pIn) = max8(*((uint32_t *)pIn), *((uint32_t *)pCom));
+
+    pIn+=4;
+    pCom+=4;
+
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 4, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 4, 4);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 4, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 4, 4);
+
+    if(inA0<inB0)
+      inA0=inB0;
+
+    if(inA1<inB1)
+      inA1=inB1;
+
+    *((uint8_t*)pIn) = bitins(inA0, n_mask, inA1, mask, off);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_compare_and_replace_if_larger_i4(
+    int8_t * base, int8_t * target, uint16_t length) {
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~mask;
+  int8_t off = 0x04;
+
+  int8_t *pIn = base;
+  int8_t *pCom = target;
+
+  int cnt = length >> 2;
+
+  while (cnt > 0u) {
+    *((int32_t *)pIn) = maxs8(*((int32_t *)pIn), *((int32_t *)pCom));
+
+    pIn += 4;
+    pCom += 4;
+
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left > 0u) {
+    int8_t inA0 = (int8_t)bitext((int)*pIn, 4, 0);
+    int8_t inA1 = (int8_t)bitext((int)*pIn, 4, 4);
+    int8_t inB0 = (int8_t)bitext((int)*pCom, 4, 0);
+    int8_t inB1 = (int8_t)bitext((int)*pCom, 4, 4);
+
+    if (inA0 < inB0)
+      inA0 = inB0;
+
+    if (inA1 < inB1)
+      inA1 = inB1;
+
+    *((int8_t *)pIn) = bitins(inA0, n_mask, inA1, mask, off);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_avg_and_replace_u4(uint8_t * base,
+                                  uint8_t * target,
+                                  uint16_t length)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+
+  int cnt = length >> 2;
+
+  while (cnt > 0u)
+  {
+    *((uint32_t *)pIn) = avg8(*((uint32_t *)pIn), *((uint32_t *)pCom));
+
+    pIn+=4;
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+
+  while (left>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 4, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 4, 4);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 4, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 4, 4);
+
+    inA0 = ((inA0 + inB0) >> 1);
+    inA1 = ((inA1 + inB1) >> 1);
+
+    *((uint8_t*)pIn) = bitins(inA0, n_mask, inA1, mask, off);
+
+    pIn++;
+    pCom++;
+    length--;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_compare_and_replace_if_larger_u2(uint8_t * base,
+                                                uint8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask2 = 0x0c;
+  int8_t n_mask2 = ~ mask2;
+  int8_t mask4 = 0x30;
+  int8_t n_mask4 = ~ mask4;
+  int8_t mask6 = 0xc0;
+  int8_t n_mask6 = ~ mask6;
+  int8_t off2 = 2;
+  int8_t off4 = 4;
+  int8_t off6 = 6;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+  uint8_t *out;
+
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    *((uint32_t *)pIn) = max16(*((uint32_t *)pIn), *((uint32_t *)pCom));
+
+    pIn+=4;
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 2, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 2, 2);
+    uint8_t inA2 = (uint8_t) bitextu((unsigned int) *pIn, 2, 4);
+    uint8_t inA3 = (uint8_t) bitextu((unsigned int) *pIn, 2, 6);
+    v4u inA4 = pack((uint8_t) inA0, (uint8_t) inA1, (uint8_t) inA2, (uint8_t) inA3);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 2, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 2, 2);
+    uint8_t inB2 = (uint8_t) bitextu((unsigned int) *pCom, 2, 4);
+    uint8_t inB3 = (uint8_t) bitextu((unsigned int) *pCom, 2, 6);
+    v4u inB4 = pack((uint8_t) inB0, (uint8_t) inB1, (uint8_t) inB2, (uint8_t) inB3);
+
+    *((v4u*)out) = max4(inA4, inB4);
+
+    uint8_t inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_compare_and_replace_if_larger_i2(int8_t * base,
+                                                int8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask2 = 0x0c;
+  int8_t n_mask2 = ~ mask2;
+  int8_t mask4 = 0x30;
+  int8_t n_mask4 = ~ mask4;
+  int8_t mask6 = 0xc0;
+  int8_t n_mask6 = ~ mask6;
+  int8_t off2 = 2;
+  int8_t off4 = 4;
+  int8_t off6 = 6;
+
+  int8_t *pIn = base;
+  int8_t *pCom = target;
+  int8_t *out;
+
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    *((int32_t *)pIn) = maxs16(*((int32_t *)pIn), *((int32_t *)pCom));
+
+    pIn+=4;
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    int8_t inA0 = (int8_t) bitext((int) *pIn, 2, 0);
+    int8_t inA1 = (int8_t) bitext((int) *pIn, 2, 2);
+    int8_t inA2 = (int8_t) bitext((int) *pIn, 2, 4);
+    int8_t inA3 = (int8_t) bitext((int) *pIn, 2, 6);
+    v4s inA4 = pack((int8_t) inA0, (int8_t) inA1, (int8_t) inA2, (int8_t) inA3);
+    int8_t inB0 = (int8_t) bitext((int) *pCom, 2, 0);
+    int8_t inB1 = (int8_t) bitext((int) *pCom, 2, 2);
+    int8_t inB2 = (int8_t) bitext((int) *pCom, 2, 4);
+    int8_t inB3 = (int8_t) bitext((int) *pCom, 2, 6);
+    v4s inB4 = pack((int8_t) inB0, (int8_t) inB1, (int8_t) inB2, (int8_t) inB3);
+
+    *((v4s*)out) = maxs4(inA4, inB4);
+
+    int8_t inA = (int8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((int8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_avg_and_replace_u2(uint8_t * base,
+                                  uint8_t * target,
+                                  uint16_t length)
+{
+  int8_t mask2 = 0x0c;
+  int8_t n_mask2 = ~ mask2;
+  int8_t mask4 = 0x30;
+  int8_t n_mask4 = ~ mask4;
+  int8_t mask6 = 0xc0;
+  int8_t n_mask6 = ~ mask6;
+  int8_t off2 = 2;
+  int8_t off4 = 4;
+  int8_t off6 = 6;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+  uint8_t *out;
+
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    *((uint32_t *)pIn) = avg16(*((uint32_t *)pIn), *((uint32_t *)pCom));
+
+    pIn+=4;
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 2, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 2, 2);
+    uint8_t inA2 = (uint8_t) bitextu((unsigned int) *pIn, 2, 4);
+    uint8_t inA3 = (uint8_t) bitextu((unsigned int) *pIn, 2, 6);
+    v4u inA4 = pack((uint8_t) inA0, (uint8_t) inA1, (uint8_t) inA2, (uint8_t) inA3);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 2, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 2, 2);
+    uint8_t inB2 = (uint8_t) bitextu((unsigned int) *pCom, 2, 4);
+    uint8_t inB3 = (uint8_t) bitextu((unsigned int) *pCom, 2, 6);
+    v4u inB4 = pack((uint8_t) inB0, (uint8_t) inB1, (uint8_t) inB2, (uint8_t) inB3);
+
+    *((v4u*)out) = avg4(inA4, inB4);
+
+    uint8_t inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+
+#endif
diff --git a/rt_nn_tests/xptnn_conv/test.c b/rt_nn_tests/xptnn_conv/test.c
new file mode 100644
index 0000000..ea71523
--- /dev/null
+++ b/rt_nn_tests/xptnn_conv/test.c
@@ -0,0 +1,224 @@
+#include <stdio.h>
+#include <stdint.h>
+#include <stdlib.h>
+
+
+#include "xpulp_tnn_matmul_ternary.h"
+#include "xpulp_tnn_matmul_ternary_4x1.h"
+#include "xpulp_tnn_conv_ternary.h"
+#include "xpulp_tnn_conv_ternary_signed.h"
+#ifndef PROFILE // HACKYTIME
+#include "xpulp_tnn_conv1d_ternary.h"
+#include "xpulp_tnn_conv1d_ternary_signed.h"
+#endif
+#include "xpulp_tnn_maxpool_ternary.h"
+#include "xpulp_tnn_linear_ternary_i32_signed.h"
+#include "xpulp_tnn_linear_ternary_i32.h"
+#include "pmsis.h"
+
+#include "data_statstest.h"
+//#include "pulp_nn_kernels.h"
+#include "pulp_nn_mix_kernels.h"
+
+
+#define start_cycle_counter()                                   asm volatile("csrw 0xCC0, 0x01;")
+#define stop_cycle_counter()                                    asm volatile("csrw 0xCC0, 0x00;")
+#define read_cycle_counter(x)                                   asm volatile("csrr %0, 0x780;" : "=r" (x))
+#define reset_cycle_counter()                                   asm volatile("csrw 0x780, 0x0;")
+
+uint8_t im2col[IM2COL_DIM] = {0};
+uint8_t outputs[OUTPUT_DIM] = {0};
+
+int32_t outputs_fp[OUTPUT_DIM_FP] = {0};
+
+#ifndef PROFILE
+int num_cycles;
+#else
+int im2col_cycles;
+int hotloop_prep_cycles;
+int hotloop_cycles;
+int threshold_cycles;
+int requant_cycles;
+int hotloop_leftover_cycles;
+int matmul4x2_leftover_cycles;
+#endif
+
+void call_krnl_0();
+void test_0();
+
+int main(int argc, char *argv[])
+{
+#if KRAKEN_PTEST == 1
+  kraken_padframe_aon_pad_gpioa_cfg_rxe_set(24, 0);
+  kraken_padframe_aon_pad_gpioa_cfg_trie_set(24, 0);
+  kraken_padframe_aon_pad_gpioa_cfg_chip2pad_set(24, 0);
+#endif
+  int n_mismatches = 0;
+  int mismatches_tot = 0;
+
+    
+
+  #ifndef PROFILE
+  num_cycles = 0;
+  #endif
+
+  if (get_core_id() == 0) {
+  printf("===> TEST 0: Running xpulp_tnn_conv_ternary_signed...\n");
+  printf("  dims_in     = [4, 4]\n");
+  printf("  dims_kernel = [3, 3]\n");
+  printf("  ch_in/out   = [20, 40]\n");
+  //printf("  padding_y_top    = [%d]\n", padding_y_top);
+  //printf("  padding_y_bottom = [%d]\n", padding_y_bottom);
+  //printf("  padding_x_left   = [%d]\n", padding_x_left);
+  //printf("  padding_x_right  = [%d]\n", padding_x_right);
+  //printf("  stride_x         = [%d]\n", stride_x);
+  //printf("  stride_y         = [%d]\n", stride_y);
+  }
+  test_0();
+  #ifndef PROFILE
+  stop_cycle_counter();
+  read_cycle_counter(num_cycles);
+  if (get_core_id() == 0) {
+  printf("===> TEST 0: Finished running xpulp_tnn_conv_ternary_signed\n");
+  printf("num_cycles = %d\n", num_cycles);
+  printf("MACs = 115200\n");
+  printf("MACs/cycle = %.4f\n", 115200/num_cycles);
+  }
+  #endif
+  if (get_core_id() == 0) {
+  printf("Checking for mismatches..\n");
+  n_mismatches = 0;
+
+  for(int i=0; i < 128; i++) {
+    if (outputs[i] != exp_outp_0[i]){
+        printf("***Mismatch in test 0 at iteration %d: Expected: %x, got: %x\n", i, exp_outp_0[i], outputs[i]);
+        n_mismatches++;
+    }
+  }
+  }
+  mismatches_tot += n_mismatches;
+
+  if (get_core_id() == 0) {
+  printf("Got %d mismatches in %d tests\n", mismatches_tot, 1);
+  }
+  return mismatches_tot;
+}
+
+
+void call_krnl_0(void) {
+  uint8_t * pInp;
+  uint8_t * pIm2ColBuffer;
+  int8_t * pBias = NULL;
+  uint8_t * pOut;
+  int8_t * pWeight;
+  uint32_t * pThr;
+  int32_t * pKappa, pLambda;
+  #ifdef PROFILE
+  int32_t im2col_cycles = 0;
+  int32_t hotloop_prep_cycles = 0;
+  int32_t hotloop_cycles = 0;
+  int32_t threshold_cycles = 0;
+  int32_t requant_cycles = 0;
+  int32_t hotloop_leftover_cycles = 0;
+  int32_t matmul4x2_leftover_cycles = 0;
+  #endif
+  pInp = inp_l1;
+  
+  pOut = outp_l1;
+  pIm2ColBuffer = im2col_l1;
+  pWeight = wt_l1;
+  pThr = threshs_l1;
+  pKappa = kappa_l1;
+  pLambda = lambda_l1;
+#if KRAKEN_PTEST == 1
+  if (pi_core_id() == 0) {
+    kraken_padframe_aon_pad_gpioa_cfg_chip2pad_set(24, 1);
+  } 
+#endif
+  xpulp_tnn_conv_ternary_signed(
+    pInp,
+    pIm2ColBuffer,
+    pBias,
+    pOut,
+    pWeight,
+    pThr,
+    4,
+    4,
+    20,
+    4,
+    4,
+    40,
+    3,
+    3,
+    1,
+    1,
+    1,
+    1,
+    1,
+#ifndef PROFILE
+    1
+#else
+    1,
+    &im2col_cycles,
+    &hotloop_prep_cycles,
+    &hotloop_cycles,
+    &threshold_cycles,
+    &hotloop_leftover_cycles,
+    &matmul4x2_leftover_cycles
+#endif
+    );
+
+#if KRAKEN_PTEST == 1
+  if (pi_core_id() == 0) {
+    kraken_padframe_aon_pad_gpioa_cfg_chip2pad_set(24, 0);
+  }
+#endif
+  #ifdef PROFILE
+  if (pi_core_id() == 0) {
+  printf("im2col_cycles = %d\n", im2col_cycles);
+  printf("hotloop_prep_cycles = %d\n", hotloop_prep_cycles);
+  printf("hotloop_cycles = %d\n", hotloop_cycles);
+  printf("requant_cycles = %d\n", requant_cycles);
+  printf("threshold_cycles = %d\n", threshold_cycles);
+  printf("hotloop_leftover_cycles = %d\n", hotloop_leftover_cycles);
+  printf("matmul4x2_leftover_cycles = %d\n", matmul4x2_leftover_cycles);
+  }
+  #endif
+}
+
+void test_0(void) {
+  // DMA transfer inputs from L2 to L1
+  if (pi_core_id() == 0) {
+     plp_dma_memcpy(pIn_0, inp_l1, 64, 1);
+     plp_dma_barrier();
+  }
+  pi_cl_team_barrier(0);
+  // transfer thresholds
+  if (pi_core_id() == 0) {
+     plp_dma_memcpy(pThr_0, threshs_l1, 40 * 4, 1); // 4 bytes per set of 2 thresholds
+     plp_dma_barrier();
+  }
+  pi_cl_team_barrier(0);
+  // transfer weights
+  if (pi_core_id() == 0) {
+     plp_dma_memcpy(pWeight_0, wt_l1, 1440, 1);
+     plp_dma_barrier();
+  }
+  pi_cl_team_barrier(0);
+  call_krnl_0();
+  // get outputs back with DMA
+  if (pi_core_id() == 0) {
+     plp_dma_memcpy(outputs, outp_l1, 128, 0);
+     plp_dma_barrier();
+  }
+}
+
+
+
+
+
+
+
+
+
+
diff --git a/rt_nn_tests/xptnn_conv/xpulp_tnn_conv1d_ternary.h b/rt_nn_tests/xptnn_conv/xpulp_tnn_conv1d_ternary.h
new file mode 100644
index 0000000..f1a08a6
--- /dev/null
+++ b/rt_nn_tests/xptnn_conv/xpulp_tnn_conv1d_ternary.h
@@ -0,0 +1,35 @@
+#ifndef __XPULP_NN_CONV1D_TERNARY_H
+#define __XPULP_NN_CONV1D_TERNARY_H
+#include <stdio.h>
+#include "pulp_nn_utils.h"
+
+
+
+// TODO: review argument order
+void __attribute__((noinline)) xpulp_tnn_conv1d_ternary(
+                                                        uint8_t  *pIn,
+                                                        uint8_t  *pIm2ColBuffer,
+                                                        int8_t   *pBias,
+                                                        uint8_t  *pOut,
+                                                        uint8_t  *pWeight,
+                                                        uint32_t *pThr,
+                                                        uint16_t  dim_in_x,
+                                                        uint16_t  ch_in,
+                                                        uint16_t  dim_out_x,
+                                                        uint16_t  ch_out,
+                                                        uint16_t  dim_kernel_x,
+                                                        uint16_t  padding_x_left,
+                                                        uint16_t  padding_x_right,
+                                                        uint16_t  stride_x,
+#ifndef PROFILE                                         
+                                                        uint16_t  dilation_x);
+#else
+                                                        uint16_t  dilation_x,
+                                                        uint32_t *im2col_cycles,
+                                                        uint32_t *hotloop_prep_cycles,
+                                                        uint32_t *hotloop_cycles,
+                                                        uint32_t *threshold_cycles,
+                                                        uint32_t *hotloop_leftover_cycles,
+                                                        uint32_t *matmul4x2_leftover_cycles);
+#endif
+#endif
diff --git a/rt_nn_tests/xptnn_conv/xpulp_tnn_conv1d_ternary_signed.h b/rt_nn_tests/xptnn_conv/xpulp_tnn_conv1d_ternary_signed.h
new file mode 100644
index 0000000..0f8ba08
--- /dev/null
+++ b/rt_nn_tests/xptnn_conv/xpulp_tnn_conv1d_ternary_signed.h
@@ -0,0 +1,35 @@
+#ifndef __XPULP_NN_CONV1D_TERNARY_SIGNED_H
+#define __XPULP_NN_CONV1D_TERNARY_SIGNED_H
+#include <stdio.h>
+#include "pulp_nn_utils.h"
+
+
+
+// TODO: review argument order
+void __attribute__((noinline)) xpulp_tnn_conv1d_ternary_signed(
+                                                        uint8_t  *pIn,
+                                                        uint8_t  *pIm2ColBuffer,
+                                                        int8_t   *pBias,
+                                                        uint8_t  *pOut,
+                                                        uint8_t  *pWeight,
+                                                        uint32_t *pThr,
+                                                        uint16_t  dim_in_x,
+                                                        uint16_t  ch_in,
+                                                        uint16_t  dim_out_x,
+                                                        uint16_t  ch_out,
+                                                        uint16_t  dim_kernel_x,
+                                                        uint16_t  padding_x_left,
+                                                        uint16_t  padding_x_right,
+                                                        uint16_t  stride_x,
+#ifndef PROFILE                                         
+                                                        uint16_t  dilation_x);
+#else
+                                                        uint16_t  dilation_x,
+                                                        uint32_t *im2col_cycles,
+                                                        uint32_t *hotloop_prep_cycles,
+                                                        uint32_t *hotloop_cycles,
+                                                        uint32_t *threshold_cycles,
+                                                        uint32_t *hotloop_leftover_cycles,
+                                                        uint32_t *matmul4x2_leftover_cycles);
+#endif
+#endif
diff --git a/rt_nn_tests/xptnn_conv/xpulp_tnn_conv_ternary.h b/rt_nn_tests/xptnn_conv/xpulp_tnn_conv_ternary.h
new file mode 100644
index 0000000..f4a8759
--- /dev/null
+++ b/rt_nn_tests/xptnn_conv/xpulp_tnn_conv_ternary.h
@@ -0,0 +1,40 @@
+#ifndef __XPULP_NN_CONV_TERNARY_H
+#define __XPULP_NN_CONV_TERNARY_H
+#include <stdio.h>
+#include "pulp_nn_utils.h"
+
+
+
+// TODO: review argument order
+void __attribute__((noinline)) xpulp_tnn_conv_ternary(
+                                                     uint8_t  *pIn,
+                                                     uint8_t  *pIm2ColBuffer,
+                                                     int8_t   *pBias,
+                                                     uint8_t  *pOut,
+                                                     uint8_t  *pWeight,
+                                                     uint32_t *pThr,
+                                                     uint16_t  dim_in_x,
+                                                     uint16_t  dim_in_y,
+                                                     uint16_t  ch_in,
+                                                     uint16_t  dim_out_x,
+                                                     uint16_t  dim_out_y,
+                                                     uint16_t  ch_out,
+                                                     uint16_t  dim_kernel_x,
+                                                     uint16_t  dim_kernel_y,
+                                                     uint16_t  padding_y_top,
+                                                     uint16_t  padding_y_bottom,
+                                                     uint16_t  padding_x_left,
+                                                     uint16_t  padding_x_right,
+                                                     uint16_t  stride_x,
+#ifndef PROFILE
+                                                     uint16_t  stride_y);
+#else
+                                                     uint16_t  stride_y,
+                                                     uint32_t *im2col_cycles,
+                                                     uint32_t *hotloop_prep_cycles,
+                                                     uint32_t *hotloop_cycles,
+                                                     uint32_t *threshold_cycles,
+                                                     uint32_t *hotloop_leftover_cycles,
+                                                     uint32_t *matmul4x2_leftover_cycles);
+#endif
+#endif
diff --git a/rt_nn_tests/xptnn_conv/xpulp_tnn_conv_ternary_signed.c b/rt_nn_tests/xptnn_conv/xpulp_tnn_conv_ternary_signed.c
new file mode 100644
index 0000000..92ff6cd
--- /dev/null
+++ b/rt_nn_tests/xptnn_conv/xpulp_tnn_conv_ternary_signed.c
@@ -0,0 +1,259 @@
+#include <stdio.h>
+#include "pmsis.h"
+#include "pulp_nn_utils.h"
+#include "xpulp_tnn_conv_ternary_signed.h"
+#include "xpulp_tnn_matmul_ternary_signed.h"
+#include "xpulp_tnn_matmul_ternary_signed_4x1.h"
+
+// TODO: review argument order
+void __attribute__((noinline)) xpulp_tnn_conv_ternary_signed(
+                        uint8_t  *pIn,
+                        uint8_t  *pIm2ColBuffer,
+                        int8_t   *pBias,
+                        uint8_t  *pOut,
+                        uint8_t  *pWeight,
+                        uint32_t *pThr,
+                        uint16_t  dim_in_x,
+                        uint16_t  dim_in_y,
+                        uint16_t  ch_in,
+                        uint16_t  dim_out_x,
+                        uint16_t  dim_out_y,
+                        uint16_t  ch_out,
+                        uint16_t  dim_kernel_x,
+                        uint16_t  dim_kernel_y,
+                        uint16_t  padding_y_top,
+                        uint16_t  padding_y_bottom,
+                        uint16_t  padding_x_left,
+                        uint16_t  padding_x_right,
+                        uint16_t  stride_x,
+#ifndef PROFILE
+                        uint16_t  stride_y)
+{
+#else
+                        uint16_t  stride_y,
+                        uint32_t *im2col_cycles,
+                        uint32_t *hotloop_prep_cycles,
+                        uint32_t *hotloop_cycles,
+                        uint32_t *threshold_cycles,
+                        uint32_t *hotloop_leftover_cycles,
+                        uint32_t *matmul4x2_leftover_cycles)
+{
+  uint32_t im2col_add;
+  uint32_t hotloop_prep_add;
+  uint32_t hotloop_add;
+  uint32_t threshold_add;
+  uint32_t hotloop_leftover_add;
+  uint32_t matmul4x2_leftover_add;
+#endif
+  //uint16_t ch_in_r = PACK_INT2_SIZE(ch_in);
+  uint16_t ch_in_r = ch_in/5;
+  uint16_t ch_out_r = ch_out/5;
+  uint16_t ch_in_c = ch_in_r * 4;
+  uint16_t ch_out_c = ch_out_r * 4;
+#ifdef FC_TEST
+  int core_id = 0;
+#else
+  int core_id = pi_core_id();
+#endif
+  uint8_t * pIm2ColBase = pIm2ColBuffer + (2 * core_id * ch_in_r * dim_kernel_x * dim_kernel_y);
+  int i_out_y, i_out_x, i_ker_y, i_ker_x;
+  int Log2Core;
+
+  uint8_t extra_chunk = ((dim_out_y & (NUM_CORES-1)) != 0);
+  uint8_t extra_chunk_r;
+  uint16_t dim_out_x_r;
+  uint8_t section;
+  int core_id_r;
+
+  if(extra_chunk && dim_out_x > 1)
+  {
+    Log2Core = log2(NUM_CORES >> 1);
+    core_id_r = (core_id >> 1);
+    dim_out_x_r = (dim_out_x >> 1);
+    section = (core_id & 0x1);
+    extra_chunk_r = ((dim_out_y & ((NUM_CORES >> 1) - 1)) != 0);
+  }
+  else
+  {
+    Log2Core = log2(NUM_CORES);
+    core_id_r = core_id;
+    dim_out_x_r = dim_out_x;
+    section = 0;
+    extra_chunk_r = extra_chunk;
+    extra_chunk = 0;
+  }
+
+  uint8_t flag_dim_out_x_odd = dim_out_x & 0x01;
+
+  int chunk = (dim_out_y >> Log2Core) + extra_chunk_r;
+
+  int start_pixel = min((chunk * core_id_r), dim_out_y);
+  int stop_pixel = min(start_pixel + chunk, dim_out_y);
+
+  uint8_t *pIm2Col = pIm2ColBase;
+  uint8_t *pOutBuffer = pOut + (start_pixel * ch_out_r * dim_out_x) + (section * ch_out_r * dim_out_x_r);
+  int thrc_res1 = 0, thrc_res2 = 0;
+
+  for (i_out_y = start_pixel; i_out_y < stop_pixel; i_out_y++)
+  {
+    for(i_out_x=(section * dim_out_x_r); i_out_x<(dim_out_x_r + (section * (dim_out_x_r + flag_dim_out_x_odd))); i_out_x++)
+    {
+      #ifdef PROFILE
+      //printf("Started im2col counter\n");
+      reset_cycle_counter();
+      start_cycle_counter();
+      #endif
+      if(i_out_y < padding_y_top)
+      {
+        for(i_ker_y=((i_out_y * stride_y) - padding_y_top); i_ker_y<((i_out_y * stride_y) - padding_y_top + dim_kernel_y); i_ker_y++)
+        {
+          for(i_ker_x=((i_out_x * stride_x) - padding_x_left); i_ker_x<((i_out_x * stride_x) - padding_x_left + dim_kernel_x); i_ker_x++)
+          {
+            if((i_ker_y < 0) || (i_ker_y >= dim_in_y) || (i_ker_x < 0) || (i_ker_x >= dim_in_x))
+            {
+              xpulp_tnn_zero_mem_ternary(pIm2Col, ch_in_c, 0);
+            }
+            else
+            {
+              xpulp_nn_im2col_u2_to_u2((uint8_t*) (pIn + ((i_ker_y * dim_in_x + i_ker_x) * ch_in_r)), pIm2Col, ch_in_c);
+            }
+            pIm2Col+=ch_in_r;
+          }
+        }
+      }
+      else if(i_out_y < dim_out_y - padding_y_bottom)
+      {
+        if(i_out_x < padding_x_left)
+        {
+          for(i_ker_y=((i_out_y * stride_y) - padding_y_top); i_ker_y<((i_out_y * stride_y) - padding_y_top + dim_kernel_y); i_ker_y++)
+          {
+            for(i_ker_x=((i_out_x * stride_x) - padding_x_left); i_ker_x<((i_out_x * stride_x) - padding_x_left + dim_kernel_x); i_ker_x++)
+            {
+              if((i_ker_x < 0) || (i_ker_x >= dim_in_x))
+              {
+                xpulp_tnn_zero_mem_ternary(pIm2Col, ch_in_c, 0);
+              }
+              else
+              {
+                xpulp_nn_im2col_u2_to_u2((uint8_t*) (pIn + ((i_ker_y * dim_in_x + i_ker_x) * ch_in_r)), pIm2Col, ch_in_c);
+              }
+              pIm2Col+=ch_in_r;
+            }
+          }
+        }
+        else if(i_out_x < (dim_out_x - padding_x_right))
+        {
+          for(i_ker_y=((i_out_y * stride_y) - padding_y_top); i_ker_y<((i_out_y * stride_y) - padding_y_top + dim_kernel_y); i_ker_y++)
+          {
+            xpulp_nn_im2col_u2_to_u2((uint8_t*) pIn + (i_ker_y * dim_in_x + i_out_x * stride_x - padding_x_left)*ch_in_r,pIm2Col,ch_in_c * dim_kernel_x);
+            pIm2Col+= ch_in_r * dim_kernel_x;
+          }
+        }
+        else
+        {
+          for(i_ker_y=((i_out_y * stride_y) - padding_y_top); i_ker_y<((i_out_y * stride_y) - padding_y_top + dim_kernel_y); i_ker_y++)
+          {
+            for(i_ker_x = i_out_x * stride_x - padding_x_left; i_ker_x < i_out_x * stride_x - padding_x_left + dim_kernel_x; i_ker_x++)
+            {
+              if((i_ker_x < 0) || (i_ker_x >= dim_in_x))
+              {
+                xpulp_tnn_zero_mem_ternary(pIm2Col, ch_in_c, 0);
+              }
+              else
+              {
+                xpulp_nn_im2col_u2_to_u2((uint8_t *)pIn + (i_ker_y*dim_in_x+i_ker_x)* ch_in_r, pIm2Col, ch_in_c);
+              }
+              pIm2Col += ch_in_r;
+            }
+          }
+        }
+      }
+      else
+      {
+        for(i_ker_y=((i_out_y * stride_y) - padding_y_top); i_ker_y<((i_out_y * stride_y) - padding_y_top + dim_kernel_y); i_ker_y++)
+        {
+          for(i_ker_x = i_out_x * stride_x - padding_x_left; i_ker_x < i_out_x * stride_x - padding_x_left + dim_kernel_x; i_ker_x++)
+          {
+            if(i_ker_y < 0 || (i_ker_y >= dim_in_y) || i_ker_x < 0 || i_ker_x >= dim_in_x)
+            {
+              xpulp_tnn_zero_mem_ternary(pIm2Col, ch_in_c, 0);
+            }
+            else
+            {
+              xpulp_nn_im2col_u2_to_u2((uint8_t *) pIn + (i_ker_y * dim_in_x + i_ker_x) * ch_in_r, pIm2Col, ch_in_c);
+            }
+            pIm2Col += ch_in_r;
+          }
+        }
+      }
+      #ifdef PROFILE
+      stop_cycle_counter();
+      //printf("Stopped im2col counter\n");
+      read_cycle_counter(im2col_add);
+      *im2col_cycles += im2col_add;
+      #endif
+      if(pIm2Col == (pIm2ColBase + ((ch_in_r * dim_kernel_x * dim_kernel_y) << 1)))
+      {
+        pOutBuffer = xpulp_tnn_matmul_ternary_signed(
+          pIm2ColBase,
+          pBias,
+          pThr,
+          pOutBuffer,
+          pOutBuffer + ch_out_r,
+          pWeight,
+          (ch_in * dim_kernel_x * dim_kernel_y),
+          ch_out,
+          &thrc_res1,
+          #ifndef PROFILE
+          &thrc_res2
+          #else
+          &thrc_res2,
+          hotloop_prep_cycles,
+          hotloop_cycles,
+          threshold_cycles,
+          hotloop_leftover_cycles
+          #endif
+        );
+        pIm2Col = pIm2ColBase;
+      }
+    }
+    thrc_res1 = 0;
+    if(pIm2Col != pIm2ColBase)
+    {
+      #ifdef PROFILE
+      //printf("Started matmul4x2_leftover counter\n");
+      reset_cycle_counter();
+      start_cycle_counter();
+      #endif
+      pOutBuffer = xpulp_tnn_matmul_ternary_signed_4x1(
+        pIm2ColBase,
+        pBias,
+        pThr,
+        pOutBuffer,
+        pWeight,
+        (ch_in * dim_kernel_x * dim_kernel_y),
+        ch_out,
+#ifndef PROFILE
+        &thrc_res1
+#else
+        &thrc_res1,
+        hotloop_prep_cycles,
+        hotloop_leftover_cycles, // everything done by the 4x1 hotloop is "hotloop leftover"!
+        threshold_cycles,
+        hotloop_leftover_cycles
+#endif
+      );
+      #ifdef PROFILE
+      stop_cycle_counter();
+      //printf("Stopped matmul4x2_leftover counter\n");
+      read_cycle_counter(matmul4x2_leftover_add);
+      *matmul4x2_leftover_cycles += matmul4x2_leftover_add;
+      #endif
+    }
+    pOutBuffer+=(extra_chunk * ((dim_out_x_r + ((1 - section) * flag_dim_out_x_odd)) * ch_out_r));
+    pIm2Col = pIm2ColBase;
+  }
+#ifndef FC_TEST
+  pi_cl_team_barrier(0);
+#endif
+}
diff --git a/rt_nn_tests/xptnn_conv/xpulp_tnn_conv_ternary_signed.h b/rt_nn_tests/xptnn_conv/xpulp_tnn_conv_ternary_signed.h
new file mode 100644
index 0000000..d05c2ab
--- /dev/null
+++ b/rt_nn_tests/xptnn_conv/xpulp_tnn_conv_ternary_signed.h
@@ -0,0 +1,40 @@
+#ifndef __XPULP_NN_CONV_TERNARY_SIGNED_H
+#define __XPULP_NN_CONV_TERNARY_SIGNED_H
+#include <stdio.h>
+#include "pulp_nn_utils.h"
+
+
+
+// TODO: review argument order
+void __attribute__((noinline)) xpulp_tnn_conv_ternary_signed(
+                                                     uint8_t  *pIn,
+                                                     uint8_t  *pIm2ColBuffer,
+                                                     int8_t   *pBias,
+                                                     uint8_t  *pOut,
+                                                     uint8_t  *pWeight,
+                                                     uint32_t *pThr,
+                                                     uint16_t  dim_in_x,
+                                                     uint16_t  dim_in_y,
+                                                     uint16_t  ch_in,
+                                                     uint16_t  dim_out_x,
+                                                     uint16_t  dim_out_y,
+                                                     uint16_t  ch_out,
+                                                     uint16_t  dim_kernel_x,
+                                                     uint16_t  dim_kernel_y,
+                                                     uint16_t  padding_y_top,
+                                                     uint16_t  padding_y_bottom,
+                                                     uint16_t  padding_x_left,
+                                                     uint16_t  padding_x_right,
+                                                     uint16_t  stride_x,
+#ifndef PROFILE
+                                                     uint16_t  stride_y);
+#else
+                                                     uint16_t  stride_y,
+                                                     uint32_t *im2col_cycles,
+                                                     uint32_t *hotloop_prep_cycles,
+                                                     uint32_t *hotloop_cycles,
+                                                     uint32_t *threshold_cycles,
+                                                     uint32_t *hotloop_leftover_cycles,
+                                                     uint32_t *matmul4x2_leftover_cycles);
+#endif
+#endif
diff --git a/rt_nn_tests/xptnn_conv/xpulp_tnn_linear_ternary_i32.h b/rt_nn_tests/xptnn_conv/xpulp_tnn_linear_ternary_i32.h
new file mode 100644
index 0000000..5c353d3
--- /dev/null
+++ b/rt_nn_tests/xptnn_conv/xpulp_tnn_linear_ternary_i32.h
@@ -0,0 +1,12 @@
+#ifndef __XPULP_TNN_LINEAR_TERNARY_I32_H
+#define __XPULP_TNN_LINEAR_TERNARY_I32_H
+
+void __attribute__((noinline)) xpulp_tnn_linear_ternary_i32(
+                   uint8_t *pIn,
+                   int8_t *pBias,
+                   int8_t *pOut,
+                   int8_t *pWeight,
+                   uint16_t dim_vec,
+                   uint16_t num_o_neurons);
+
+#endif
\ No newline at end of file
diff --git a/rt_nn_tests/xptnn_conv/xpulp_tnn_linear_ternary_i32_signed.h b/rt_nn_tests/xptnn_conv/xpulp_tnn_linear_ternary_i32_signed.h
new file mode 100644
index 0000000..218c5c4
--- /dev/null
+++ b/rt_nn_tests/xptnn_conv/xpulp_tnn_linear_ternary_i32_signed.h
@@ -0,0 +1,12 @@
+#ifndef __XPULP_TNN_LINEAR_TERNARY_I32_SIGNED_H
+#define __XPULP_TNN_LINEAR_TERNARY_I32_SIGNED_H
+
+void __attribute__((noinline)) xpulp_tnn_linear_ternary_i32_signed(
+                   int8_t *pIn,
+                   int8_t *pBias,
+                   int8_t *pOut,
+                   int8_t *pWeight,
+                   uint16_t dim_vec,
+                   uint16_t num_o_neurons);
+
+#endif
\ No newline at end of file
diff --git a/rt_nn_tests/xptnn_conv/xpulp_tnn_matmul_ternary.h b/rt_nn_tests/xptnn_conv/xpulp_tnn_matmul_ternary.h
new file mode 100644
index 0000000..914daee
--- /dev/null
+++ b/rt_nn_tests/xptnn_conv/xpulp_tnn_matmul_ternary.h
@@ -0,0 +1,113 @@
+#ifndef __XPULP_TNN_MATMUL_TERNARY_H
+#define __XPULP_TNN_MATMUL_TERNARY_H
+
+#include "pulp_nn_utils.h"
+
+uint8_t * __attribute__((noinline)) xpulp_tnn_matmul_ternary(
+                        uint8_t  *pIn,
+                        int8_t   *pBias,
+                        uint32_t *pThr,
+                        uint8_t  *pOut,
+                        uint8_t  *pOut2,
+                        uint8_t  *pWeight,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+#ifndef PROFILE
+                        uint32_t *thrc_res1,
+                        uint32_t *thrc_res2);
+#else
+                        uint32_t *thrc_res1,
+                        uint32_t *thrc_res2,
+                        uint32_t *hotloop_prep_cycles,
+                        uint32_t *hotloop_cycles,
+                        uint32_t *threshold_cycles,
+                        uint32_t *hotloop_leftover_cycles);
+#endif
+
+inline void __attribute__((aligned(4))) hotloop_4x2(
+  int *sum, int *sum2, int *sum3, int *sum4, int *sum5, int *sum6, int *sum7, int *sum8,
+  uint32_t *ptrA, uint32_t *ptrA2, uint32_t *ptrA3, uint32_t *ptrA4,
+  uint32_t *ptrB, uint32_t *ptrB2, int num_col_im2col_words
+){
+  for (int j=0; j<num_col_im2col_words; j++)
+  {
+    *sum = MacLoad20(0, 0, 0, 0, ptrA, *sum);
+    *sum2 = MacLoad20(0, 0, 1, 0, ptrA2, *sum2);
+    *sum3 = MacLoad20(0, 0, 2, 0, ptrA3, *sum3);
+    *sum4 = MacLoad20(0, 1, 3, 0, ptrB, *sum4);
+    ptrB = MacLoadUpdate(ptrB);
+
+    *sum5 = MacLoad20(1, 0, 0, 1, ptrA, *sum5);
+    ptrA = MacLoadUpdate(ptrA);
+    *sum6 = MacLoad20(1, 0, 1, 1, ptrA2, *sum6);
+    ptrA2 = MacLoadUpdate(ptrA2);
+    *sum7 = MacLoad20(1, 0, 2, 1, ptrA3, *sum7);
+    ptrA3 = MacLoadUpdate(ptrA3);
+    *sum8 = MacLoad20(1, 0, 3, 1, ptrA4, *sum8);
+    ptrA4 = MacLoadUpdate(ptrA4);
+
+    ptrB2 = MacLoadInit(0, 1, 0, 1, ptrB2);
+  }
+}
+
+
+inline void __attribute__((aligned(4))) hotloop_3x2(
+  int *sum, int *sum2, int *sum3, int *sum4, int *sum5, int *sum6,
+  uint32_t *ptrA, uint32_t *ptrA2, uint32_t *ptrA3, uint32_t *ptrB, uint32_t *ptrB2, int num_col_im2col_words
+){
+  for (int j=0; j<num_col_im2col_words; j++)
+  {
+    *sum = MacLoad20(0, 0, 0, 0, ptrA, *sum);
+    *sum2 = MacLoad20(0, 0, 1, 0, ptrA2, *sum2);
+    *sum3 = MacLoad20(0, 1, 2, 0, ptrB, *sum3);
+    ptrB = MacLoadUpdate(ptrB);
+
+    *sum4 = MacLoad20(1, 0, 0, 1, ptrA, *sum4);
+    ptrA = MacLoadUpdate(ptrA);
+    *sum5 = MacLoad20(1, 0, 1, 1, ptrA2, *sum5);
+    ptrA2 = MacLoadUpdate(ptrA2);
+    *sum6 = MacLoad20(1, 0, 2, 1, ptrA3, *sum6);
+    ptrA3 = MacLoadUpdate(ptrA3);
+
+    ptrB2 = MacLoadInit(0, 1, 0, 1, ptrB2);
+  }
+}
+
+
+inline void __attribute__((aligned(4))) hotloop_2x2(
+  int *sum, int *sum2, int *sum3, int *sum4,
+  uint32_t *ptrA, uint32_t *ptrA2, uint32_t *ptrB, uint32_t *ptrB2, int num_col_im2col_words
+){
+  for (int j=0; j<num_col_im2col_words; j++)
+  {
+    *sum = MacLoad20(0, 0, 0, 0, ptrA, *sum);
+    *sum2 = MacLoad20(0, 1, 1, 0, ptrB, *sum2);
+    ptrB = MacLoadUpdate(ptrB);
+
+    *sum3 = MacLoad20(1, 0, 0, 1, ptrA, *sum3);
+    ptrA = MacLoadUpdate(ptrA);
+    *sum4 = MacLoad20(1, 0, 1, 1, ptrA2, *sum4);
+    ptrA2 = MacLoadUpdate(ptrA2);
+
+    ptrB2 = MacLoadInit(0, 1, 0, 1, ptrB2);
+  }
+}
+
+
+inline void __attribute__((aligned(4))) hotloop_1x2(
+  int *sum, int *sum2, uint32_t *ptrA, uint32_t *ptrB, uint32_t *ptrB2, int num_col_im2col_words
+){
+  for (int j=0; j<num_col_im2col_words; j++)
+  {
+    *sum = MacLoad20(0, 1, 0, 0, ptrB, *sum);
+    ptrB = MacLoadUpdate(ptrB);
+    *sum2 = MacLoad20(1, 0, 0, 1, ptrA, *sum2);
+    ptrA = MacLoadUpdate(ptrA);
+
+    ptrB2 = MacLoadInit(0, 1, 0, 1, ptrB2);
+  }
+}
+
+
+
+#endif
\ No newline at end of file
diff --git a/rt_nn_tests/xptnn_conv/xpulp_tnn_matmul_ternary_4x1.h b/rt_nn_tests/xptnn_conv/xpulp_tnn_matmul_ternary_4x1.h
new file mode 100644
index 0000000..fd06e93
--- /dev/null
+++ b/rt_nn_tests/xptnn_conv/xpulp_tnn_matmul_ternary_4x1.h
@@ -0,0 +1,96 @@
+#ifndef __XPULP_TNN_MATMUL_TERNARY_4X1_H
+#define __XPULP_TNN_MATMUL_TERNARY_4X1_H
+
+#include "pulp_nn_utils.h"
+
+uint8_t * __attribute__((noinline)) xpulp_tnn_matmul_ternary_4x1(
+                        uint8_t  *pIn,
+                        int8_t   *pBias,
+                        uint32_t *pThr,
+                        uint8_t  *pOut,
+                        uint8_t  *pWeight,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+#ifndef PROFILE
+                        uint32_t *thrc_res);
+#else
+                        uint32_t *thrc_res,
+                        uint32_t *hotloop_prep_cycles,
+                        uint32_t *hotloop_cycles,
+                        uint32_t *threshold_cycles,
+                        uint32_t *hotloop_leftover_cycles);
+#endif
+
+inline void __attribute__((aligned(4))) hotloop_4x1(
+  int *sum, int *sum2, int *sum3, int *sum4,
+  uint32_t *ptrA, uint32_t *ptrA2, uint32_t *ptrA3, uint32_t *ptrA4,
+  uint32_t *ptrB, int num_col_im2col_words
+){
+  for (int j=0; j<num_col_im2col_words; j++)
+  {
+    *sum = MacLoad20(1, 0, 0, 0, ptrA, *sum);
+    ptrA = MacLoadUpdate(ptrA);
+    *sum2 = MacLoad20(1, 0, 1, 0, ptrA2, *sum2);
+    ptrA2 = MacLoadUpdate(ptrA2);
+    *sum3 = MacLoad20(1, 0, 2, 0, ptrA3, *sum3);
+    ptrA3 = MacLoadUpdate(ptrA3);
+    *sum4 = MacLoad20(1, 0, 3, 0, ptrA4, *sum4);
+    ptrA4 = MacLoadUpdate(ptrA4);
+
+    ptrB = MacLoadInit(0, 1, 0, 0, ptrB);
+    asm volatile ("nop;");
+  }
+}
+
+
+inline void __attribute__((aligned(4))) hotloop_3x1(
+  int *sum, int *sum2, int *sum3,
+  uint32_t *ptrA, uint32_t *ptrA2, uint32_t *ptrA3, uint32_t *ptrB, int num_col_im2col_words
+){
+  for (int j=0; j<num_col_im2col_words; j++)
+  {
+    *sum = MacLoad20(1, 0, 0, 0, ptrA, *sum);
+    ptrA = MacLoadUpdate(ptrA);
+    *sum2 = MacLoad20(1, 0, 1, 0, ptrA2, *sum2);
+    ptrA2 = MacLoadUpdate(ptrA2);
+    *sum3 = MacLoad20(1, 0, 2, 0, ptrA3, *sum3);
+    ptrA3 = MacLoadUpdate(ptrA3);
+
+    ptrB = MacLoadInit(0, 1, 0, 0, ptrB);
+    asm volatile ("nop;");
+  }
+}
+
+
+inline void __attribute__((aligned(4))) hotloop_2x1(
+  int *sum, int *sum2, uint32_t *ptrA, uint32_t *ptrA2, uint32_t *ptrB, int num_col_im2col_words
+){
+  for (int j=0; j<num_col_im2col_words; j++)
+  {
+    *sum = MacLoad20(1, 0, 0, 0, ptrA, *sum);
+    ptrA = MacLoadUpdate(ptrA);
+    *sum2 = MacLoad20(1, 0, 1, 0, ptrA2, *sum2);
+    ptrA2 = MacLoadUpdate(ptrA2);
+
+    ptrB = MacLoadInit(0, 1, 0, 0, ptrB);
+    asm volatile ("nop;");
+  }
+}
+
+
+inline void __attribute__((aligned(4))) hotloop_1x1(
+  int *sum, uint32_t *ptrA, uint32_t *ptrB, int num_col_im2col_words
+){
+  for (int j=0; j<num_col_im2col_words; j++)
+  {
+    *sum = MacLoad20(1, 0, 0, 0, ptrA, *sum);
+    ptrA = MacLoadUpdate(ptrA);
+
+    ptrB = MacLoadInit(0, 1, 0, 0, ptrB);
+    asm volatile ("nop;");
+  }
+}
+
+
+
+#endif
\ No newline at end of file
diff --git a/rt_nn_tests/xptnn_conv/xpulp_tnn_matmul_ternary_signed.c b/rt_nn_tests/xptnn_conv/xpulp_tnn_matmul_ternary_signed.c
new file mode 100644
index 0000000..fb179f5
--- /dev/null
+++ b/rt_nn_tests/xptnn_conv/xpulp_tnn_matmul_ternary_signed.c
@@ -0,0 +1,767 @@
+#include <stdio.h>
+#include <math.h>
+#include "pulp_nn_utils.h"
+#include "xpulp_tnn_matmul_ternary_signed.h"
+
+
+
+// TODO: review argument order
+uint8_t * __attribute__((noinline)) xpulp_tnn_matmul_ternary_signed(
+                        uint8_t  *pIn,
+                        int8_t   *pBias,
+                        uint32_t *pThr,
+                        uint8_t  *pOut,
+                        uint8_t  *pOut2,
+                        uint8_t  *pWeight,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+#ifndef PROFILE
+                        uint32_t *thrc_res1,
+                        uint32_t *thrc_res2)
+{
+#else
+                        uint32_t *thrc_res1,
+                        uint32_t *thrc_res2,
+                        uint32_t *hotloop_prep_cycles,
+                        uint32_t *hotloop_cycles,
+                        uint32_t *threshold_cycles,
+                        uint32_t *hotloop_leftover_cycles)
+{
+  uint32_t hotloop_prep_add;
+  uint32_t hotloop_add;
+  uint32_t threshold_add;
+  uint32_t hotloop_leftover_add;
+#endif
+  uint16_t ch_out_r = ch_out / 5;
+
+  uint16_t num_col_im2col_w = num_col_im2col / 5;
+  uint16_t num_col_im2col_a = num_col_im2col_w;
+  uint16_t num_col_im2col_c = num_col_im2col_w << 2;
+  // used for hotloop calls
+  uint16_t num_col_im2col_words = num_col_im2col_w >> 2;
+
+  uint8_t *pA = pWeight;
+  v2s *currThr = (v2s *) pThr;
+
+  int res1, res2, incr_val;
+  res1 = *thrc_res1;
+  res2 = *thrc_res2;
+  incr_val = 0;
+
+  for(int i=0; i < ch_out >> 2; i++)
+  {
+    #ifdef PROFILE
+    //printf("Started hotloop_prep counter\n");
+    reset_cycle_counter();
+    start_cycle_counter();
+    #endif
+    uint8_t *pB = pIn;
+    uint8_t *pB2 = (pB + num_col_im2col_a);
+
+    uint32_t *ptrB  = (uint32_t *) pB;
+    uint32_t *ptrB2 = (uint32_t *) pB2;
+
+    uint8_t *pA2 = (pA + num_col_im2col_w);
+    uint8_t *pA3 = (pA2 + num_col_im2col_w);
+    uint8_t *pA4 = (pA3 + num_col_im2col_w);
+
+    uint32_t *ptrA  = (uint32_t *) pA ;
+    uint32_t *ptrA2 = (uint32_t *) pA2;
+    uint32_t *ptrA3 = (uint32_t *) pA3;
+    uint32_t *ptrA4 = (uint32_t *) pA4;
+
+    ptrA = MacLoadInit(1, 0, 0, 0, ptrA);
+    ptrA2 = MacLoadInit(1, 0, 1, 0, ptrA2);
+    ptrA3 = MacLoadInit(1, 0, 2, 0, ptrA3);
+    ptrA4 = MacLoadInit(1, 0, 3, 0, ptrA4);
+
+  
+    ptrB = MacLoadInit(0, 1, 0, 0, ptrB);
+    ptrB2 = MacLoadInit(0, 1, 0, 1, ptrB2);
+  
+    int sum = 0;
+    int sum2 = 0;
+    int sum3 = 0;
+    int sum4 = 0;
+    int sum5 = 0;
+    int sum6 = 0;
+    int sum7 = 0;
+    int sum8 = 0;
+
+    if (pBias != NULL)
+    {
+      sum = ((int) (*pBias++));
+      sum2 = ((int) (*pBias++));
+      sum3 = ((int) (*pBias++));
+      sum4 = ((int) (*pBias++));
+
+      sum5 = sum;
+      sum6 = sum2;
+      sum7 = sum3;
+      sum8 = sum4;
+    }
+    #ifdef PROFILE
+    stop_cycle_counter();
+    //printf("Stopped hotloop_prep counter\n");
+    read_cycle_counter(hotloop_prep_add);
+    *hotloop_prep_cycles += hotloop_prep_add;
+    #endif
+
+    #ifdef PROFILE
+    //printf("Started hotloop counter\n");
+    reset_cycle_counter();
+    start_cycle_counter();
+    #endif
+    
+    hotloop_4x2(
+      &sum, &sum2, &sum3, &sum4, &sum5, &sum6, &sum7, &sum8,
+      ptrA, ptrA2, ptrA3, ptrA4, ptrB, ptrB2, num_col_im2col_words
+    );
+
+    #ifdef PROFILE
+    stop_cycle_counter();
+    //printf("Stopped hotloop counter\n");
+    read_cycle_counter(hotloop_add);
+    *hotloop_cycles += hotloop_add;
+    #endif
+
+    #ifdef PROFILE
+    //printf("Started hotloop_leftover counter\n");
+    reset_cycle_counter();
+    start_cycle_counter();
+    #endif
+    int col_cnt_im2col = num_col_im2col_c & 0xf;
+
+    if (col_cnt_im2col)
+    {
+      uint16_t loop_cnt_im2col_w = (num_col_im2col_words) << 2;
+      pA+=loop_cnt_im2col_w;
+      pA2+=loop_cnt_im2col_w;
+      pA3+=loop_cnt_im2col_w;
+      pA4+=loop_cnt_im2col_w;
+
+      uint16_t loop_cnt_im2col_a = (num_col_im2col_words) << 2;
+      pB+=loop_cnt_im2col_a;
+      pB2+=loop_cnt_im2col_a;
+
+      volatile uint32_t valA, valA2, valA3, valA4, valB, valB2;
+      // pack the remaining weights and activations into 32-bit vectors
+      // padding with 0xd9 because ternary_decoder(0xd9) = 0000000000
+      if (col_cnt_im2col == 4)
+      {
+        valA = 0xd9 << 24 | 0xd9 << 16 | 0xd9 << 8 | *pA;
+        valA2 = 0xd9 << 24 | 0xd9 << 16 | 0xd9 << 8 | *pA2;
+        valA3 = 0xd9 << 24 | 0xd9 << 16 | 0xd9 << 8 | *pA3;
+        valA4 = 0xd9 << 24 | 0xd9 << 16 | 0xd9 << 8 | *pA4;
+
+        valB = 0xd9 << 24 | 0xd9 << 16 | 0xd9 << 8 | *pB;
+        valB2 = 0xd9 << 24 | 0xd9 << 16 | 0xd9 << 8 | *pB2;
+      }
+      else if (col_cnt_im2col == 8)
+      {
+        valA = 0xd9 << 24 | 0xd9 << 16 | *(pA + 1) << 8 | *pA;
+        valA2 = 0xd9 << 24 | 0xd9 << 16 | *(pA2 + 1) << 8 | *pA2;
+        valA3 = 0xd9 << 24 | 0xd9 << 16 | *(pA3 + 1) << 8 | *pA3;
+        valA4 = 0xd9 << 24 | 0xd9 << 16 | *(pA4 + 1) << 8 | *pA4;
+
+        valB = 0xd9 << 24 | 0xd9 << 16 | *(pB + 1) << 8 | *pB;
+        valB2 = 0xd9 << 24 | 0xd9 << 16 | *(pB2 + 1) << 8 | *pB2;
+      }
+      else // col_cnt_im2col == 12
+      {
+        valA = 0xd9 << 24 | *(pA + 2) << 16 | *(pA + 1) << 8 | *pA;
+        valA2 = 0xd9 << 24 | *(pA2 + 2) << 16 | *(pA2 + 1) << 8 | *pA2;
+        valA3 = 0xd9 << 24 | *(pA3 + 2) << 16 | *(pA3 + 1) << 8 | *pA3;
+        valA4 = 0xd9 << 24 | *(pA4 + 2) << 16 | *(pA4 + 1) << 8 | *pA4;
+
+        valB = 0xd9 << 24 | *(pB + 2) << 16 | *(pB + 1) << 8 | *pB;
+        valB2 = 0xd9 << 24 | *(pB2 + 2) << 16 | *(pB2 + 1) << 8 | *pB2;
+      }
+
+      uint32_t *pA_p = &valA;
+      uint32_t *pA2_p = &valA2;
+      uint32_t *pA3_p = &valA3;
+      uint32_t *pA4_p = &valA4;
+
+      uint32_t *pB_p = &valB;
+      uint32_t *pB2_p = &valB2;
+
+      pA_p = MacLoadInit(1, 0, 0, 0, pA_p);
+      pA2_p = MacLoadInit(1, 0, 1, 0, pA2_p);
+      pA3_p = MacLoadInit(1, 0, 2, 0, pA3_p);
+      pA4_p = MacLoadInit(1, 0, 3, 0, pA4_p);
+      pB_p = MacLoadInit(0, 1, 0, 0, pB_p);
+      pB2_p = MacLoadInit(0, 1, 0, 1, pB2_p);
+  
+      pA += PACK_INT2_SIZE(col_cnt_im2col);
+
+      sum = MacLoads20(0, 0, 0, 0, pA_p, sum);
+      sum2 = MacLoads20(0, 0, 1, 0, pA2_p, sum2);
+      sum3 = MacLoads20(0, 0, 2, 0, pA3_p, sum3);
+      sum4 = MacLoads20(0, 0, 3, 0, pA4_p, sum4);
+
+      sum5 = MacLoads20(0, 0, 0, 1, pA_p, sum5);
+      sum6 = MacLoads20(0, 0, 1, 1, pA2_p, sum6);
+      sum7 = MacLoads20(0, 0, 2, 1, pA3_p, sum7);
+      sum8 = MacLoads20(0, 0, 3, 1, pA4_p, sum8);
+    }
+    #ifdef PROFILE
+    stop_cycle_counter();
+    //printf("Stopped hotloop_leftover counter\n");
+    read_cycle_counter(hotloop_leftover_add);
+    *hotloop_leftover_cycles += hotloop_leftover_add;
+    #endif
+
+    #ifdef PROFILE
+    //printf("Started threshold counter\n");
+    reset_cycle_counter();
+    start_cycle_counter();
+    #endif
+    res1 = thr_cmp(res1, sum, *currThr);
+    check_store(res1, pOut);
+    res2 = thr_cmp(res2, sum5, *currThr++);
+    check_store(res2, pOut2);
+    reset_currThr();
+
+    res1 = thr_cmp(res1, sum2, *currThr);
+    check_store(res1, pOut);
+    res2 = thr_cmp(res2, sum6, *currThr++);
+    check_store(res2, pOut2);
+    reset_currThr();
+
+    res1 = thr_cmp(res1, sum3, *currThr);
+    check_store(res1, pOut);
+    res2 = thr_cmp(res2, sum7, *currThr++);
+    check_store(res2, pOut2);
+    reset_currThr();
+
+    res1 = thr_cmp(res1, sum4, *currThr);
+    check_store(res1, pOut);
+    res2 = thr_cmp(res2, sum8, *currThr++);
+    check_store(res2, pOut2);
+    reset_currThr();
+
+    #ifdef PROFILE
+    stop_cycle_counter();
+    //printf("Stopped threshold counter\n");
+    read_cycle_counter(threshold_add);
+    *threshold_cycles += threshold_add;
+    #endif
+
+    if (!col_cnt_im2col)
+    {
+      pA+=(4*num_col_im2col_w);
+    }
+    else
+    {
+      pA+=(3*num_col_im2col_w);
+    }
+  }
+
+  // leftover part : the hotloop above produces 4N output channels. If out_ch not divisible
+  // by 4, the remaining output channels are computed below
+  int out_ch_left = ch_out & 0x3;
+
+  if (out_ch_left == 1)
+  {
+    #ifdef PROFILE
+    //printf("Started hotloop_prep counter\n");
+    reset_cycle_counter();
+    start_cycle_counter();
+    #endif
+    uint8_t *pB = pIn;
+    uint8_t *pB2 = (pB + num_col_im2col_a);
+
+    uint32_t *ptrB  = (uint32_t *) pB;
+    uint32_t *ptrB2 = (uint32_t *) pB2;
+
+    uint32_t *ptrA  = (uint32_t *) pA ;
+
+    ptrA = MacLoadInit(1, 0, 0, 0, ptrA);
+
+  
+    ptrB = MacLoadInit(0, 1, 0, 0, ptrB);
+    ptrB2 = MacLoadInit(0, 1, 0, 1, ptrB2);
+  
+    int sum = 0;
+    int sum2 = 0;
+
+    if (pBias != NULL)
+    {
+      sum = ((int) (*pBias++));
+      sum2 = sum;
+    }
+    #ifdef PROFILE
+    stop_cycle_counter();
+    //printf("Stopped hotloop_prep counter\n");
+    read_cycle_counter(hotloop_prep_add);
+    *hotloop_prep_cycles += hotloop_prep_add;
+    #endif
+
+    #ifdef PROFILE
+    //printf("Started hotloop counter\n");
+    reset_cycle_counter();
+    start_cycle_counter();
+    #endif
+    
+    hotloop_1x2(
+      &sum, &sum2, ptrA, ptrB, ptrB2, num_col_im2col_words
+    );
+
+    #ifdef PROFILE
+    stop_cycle_counter();
+    //printf("Stopped hotloop counter\n");
+    read_cycle_counter(hotloop_add);
+    *hotloop_cycles += hotloop_add;
+    #endif
+
+    #ifdef PROFILE
+    //printf("Started hotloop_leftover counter\n");
+    reset_cycle_counter();
+    start_cycle_counter();
+    #endif
+    int col_cnt_im2col = num_col_im2col_c & 0xf;
+
+    if (col_cnt_im2col)
+    {
+      uint16_t loop_cnt_im2col_w = (num_col_im2col_words) << 2;
+      pA+=loop_cnt_im2col_w;
+
+      uint16_t loop_cnt_im2col_a = (num_col_im2col_words) << 2;
+      pB+=loop_cnt_im2col_a;
+      pB2+=loop_cnt_im2col_a;
+
+      volatile uint32_t valA, valB, valB2;
+      // pack the remaining weights and activations into 32-bit vectors
+      // padding with 0xd9 because ternary_decoder(0xd9) = 0000000000
+      if (col_cnt_im2col == 4)
+      {
+        valA = 0xd9 << 24 | 0xd9 << 16 | 0xd9 << 8 | *pA;
+
+        valB = 0xd9 << 24 | 0xd9 << 16 | 0xd9 << 8 | *pB;
+        valB2 = 0xd9 << 24 | 0xd9 << 16 | 0xd9 << 8 | *pB2;
+      }
+      else if (col_cnt_im2col == 8)
+      {
+        valA = 0xd9 << 24 | 0xd9 << 16 | *(pA + 1) << 8 | *pA;
+
+        valB = 0xd9 << 24 | 0xd9 << 16 | *(pB + 1) << 8 | *pB;
+        valB2 = 0xd9 << 24 | 0xd9 << 16 | *(pB2 + 1) << 8 | *pB2;
+      }
+      else // col_cnt_im2col == 12
+      {
+        valA = 0xd9 << 24 | *(pA + 2) << 16 | *(pA + 1) << 8 | *pA;
+
+        valB = 0xd9 << 24 | *(pB + 2) << 16 | *(pB + 1) << 8 | *pB;
+        valB2 = 0xd9 << 24 | *(pB2 + 2) << 16 | *(pB2 + 1) << 8 | *pB2;
+      }
+
+      uint32_t *pA_p = &valA;
+
+      uint32_t *pB_p = &valB;
+      uint32_t *pB2_p = &valB2;
+
+      pA_p = MacLoadInit(1, 0, 0, 0, pA_p);
+      pB_p = MacLoadInit(0, 1, 0, 0, pB_p);
+      pB2_p = MacLoadInit(0, 1, 0, 1, pB2_p);
+  
+      pA += PACK_INT2_SIZE(col_cnt_im2col);
+
+      sum = MacLoads20(0, 0, 0, 0, pA_p, sum);
+
+      sum2 = MacLoads20(0, 0, 0, 1, pA_p, sum2);
+    }
+    #ifdef PROFILE
+    stop_cycle_counter();
+    //printf("Stopped hotloop_leftover counter\n");
+    read_cycle_counter(hotloop_leftover_add);
+    *hotloop_leftover_cycles += hotloop_leftover_add;
+    #endif
+
+    #ifdef PROFILE
+    //printf("Started threshold counter\n");
+    reset_cycle_counter();
+    start_cycle_counter();
+    #endif
+    res1 = thr_cmp(res1, sum, *currThr);
+    check_store(res1, pOut);
+    res2 = thr_cmp(res2, sum2, *currThr++);
+    check_store(res2, pOut2);
+    reset_currThr();
+
+    #ifdef PROFILE
+    stop_cycle_counter();
+    //printf("Stopped threshold counter\n");
+    read_cycle_counter(threshold_add);
+    *threshold_cycles += threshold_add;
+    #endif
+
+    if (!col_cnt_im2col)
+    {
+      pA+=num_col_im2col_w;
+    }
+  }
+  else if (out_ch_left == 2)
+  {
+    #ifdef PROFILE
+    //printf("Started hotloop_prep counter\n");
+    reset_cycle_counter();
+    start_cycle_counter();
+    #endif
+    uint8_t *pB = pIn;
+    uint8_t *pB2 = (pB + num_col_im2col_a);
+
+    uint32_t *ptrB  = (uint32_t *) pB;
+    uint32_t *ptrB2 = (uint32_t *) pB2;
+
+    uint8_t *pA2 = (pA + num_col_im2col_w);
+
+    uint32_t *ptrA  = (uint32_t *) pA ;
+    uint32_t *ptrA2 = (uint32_t *) pA2;
+
+    ptrA = MacLoadInit(1, 0, 0, 0, ptrA);
+    ptrA2 = MacLoadInit(1, 0, 1, 0, ptrA2);
+
+  
+    ptrB = MacLoadInit(0, 1, 0, 0, ptrB);
+    ptrB2 = MacLoadInit(0, 1, 0, 1, ptrB2);
+  
+    int sum = 0;
+    int sum2 = 0;
+    int sum3 = 0;
+    int sum4 = 0;
+
+    if (pBias != NULL)
+    {
+      sum = ((int) (*pBias++));
+      sum2 = ((int) (*pBias++));
+
+      sum3 = sum;
+      sum4 = sum2;
+    }
+    #ifdef PROFILE
+    stop_cycle_counter();
+    //printf("Stopped hotloop_prep counter\n");
+    read_cycle_counter(hotloop_prep_add);
+    *hotloop_prep_cycles += hotloop_prep_add;
+    #endif
+
+    #ifdef PROFILE
+    //printf("Started hotloop counter\n");
+    reset_cycle_counter();
+    start_cycle_counter();
+    #endif
+    
+    hotloop_2x2(
+      &sum, &sum2, &sum3, &sum4, ptrA, ptrA2, ptrB, ptrB2, num_col_im2col_words
+    );
+
+    #ifdef PROFILE
+    stop_cycle_counter();
+    //printf("Stopped hotloop counter\n");
+    read_cycle_counter(hotloop_add);
+    *hotloop_cycles += hotloop_add;
+    #endif
+
+    #ifdef PROFILE
+    //printf("Started hotloop_leftover counter\n");
+    reset_cycle_counter();
+    start_cycle_counter();
+    #endif
+    int col_cnt_im2col = num_col_im2col_c & 0xf;
+
+    if (col_cnt_im2col)
+    {
+      uint16_t loop_cnt_im2col_w = (num_col_im2col_words) << 2;
+      pA+=loop_cnt_im2col_w;
+      pA2+=loop_cnt_im2col_w;
+
+      uint16_t loop_cnt_im2col_a = (num_col_im2col_words) << 2;
+      pB+=loop_cnt_im2col_a;
+      pB2+=loop_cnt_im2col_a;
+
+      volatile uint32_t valA, valA2, valB, valB2;
+      // pack the remaining weights and activations into 32-bit vectors
+      // padding with 0xd9 because ternary_decoder(0xd9) = 0000000000
+      if (col_cnt_im2col == 4)
+      {
+        valA = 0xd9 << 24 | 0xd9 << 16 | 0xd9 << 8 | *pA;
+        valA2 = 0xd9 << 24 | 0xd9 << 16 | 0xd9 << 8 | *pA2;
+
+        valB = 0xd9 << 24 | 0xd9 << 16 | 0xd9 << 8 | *pB;
+        valB2 = 0xd9 << 24 | 0xd9 << 16 | 0xd9 << 8 | *pB2;
+      }
+      else if (col_cnt_im2col == 8)
+      {
+        valA = 0xd9 << 24 | 0xd9 << 16 | *(pA + 1) << 8 | *pA;
+        valA2 = 0xd9 << 24 | 0xd9 << 16 | *(pA2 + 1) << 8 | *pA2;
+
+        valB = 0xd9 << 24 | 0xd9 << 16 | *(pB + 1) << 8 | *pB;
+        valB2 = 0xd9 << 24 | 0xd9 << 16 | *(pB2 + 1) << 8 | *pB2;
+      }
+      else // col_cnt_im2col == 12
+      {
+        valA = 0xd9 << 24 | *(pA + 2) << 16 | *(pA + 1) << 8 | *pA;
+        valA2 = 0xd9 << 24 | *(pA2 + 2) << 16 | *(pA2 + 1) << 8 | *pA2;
+
+        valB = 0xd9 << 24 | *(pB + 2) << 16 | *(pB + 1) << 8 | *pB;
+        valB2 = 0xd9 << 24 | *(pB2 + 2) << 16 | *(pB2 + 1) << 8 | *pB2;
+      }
+
+      uint32_t *pA_p = &valA;
+      uint32_t *pA2_p = &valA2;
+
+      uint32_t *pB_p = &valB;
+      uint32_t *pB2_p = &valB2;
+
+      pA_p = MacLoadInit(1, 0, 0, 0, pA_p);
+      pA2_p = MacLoadInit(1, 0, 1, 0, pA2_p);
+      pB_p = MacLoadInit(0, 1, 0, 0, pB_p);
+      pB2_p = MacLoadInit(0, 1, 0, 1, pB2_p);
+  
+      pA += PACK_INT2_SIZE(col_cnt_im2col);
+
+      sum = MacLoads20(0, 0, 0, 0, pA_p, sum);
+      sum2 = MacLoads20(0, 0, 1, 0, pA2_p, sum2);
+
+      sum3 = MacLoads20(0, 0, 0, 1, pA_p, sum3);
+      sum4 = MacLoads20(0, 0, 1, 1, pA2_p, sum4);
+    }
+    #ifdef PROFILE
+    stop_cycle_counter();
+    //printf("Stopped hotloop_leftover counter\n");
+    read_cycle_counter(hotloop_leftover_add);
+    *hotloop_leftover_cycles += hotloop_leftover_add;
+    #endif
+
+    #ifdef PROFILE
+    //printf("Started threshold counter\n");
+    reset_cycle_counter();
+    start_cycle_counter();
+    #endif
+    res1 = thr_cmp(res1, sum, *currThr);
+    check_store(res1, pOut);
+    res2 = thr_cmp(res2, sum3, *currThr++);
+    check_store(res2, pOut2);
+    reset_currThr();
+
+    res1 = thr_cmp(res1, sum2, *currThr);
+    check_store(res1, pOut);
+    res2 = thr_cmp(res2, sum4, *currThr++);
+    check_store(res2, pOut2);
+    reset_currThr();
+
+    #ifdef PROFILE
+    stop_cycle_counter();
+    //printf("Stopped threshold counter\n");
+    read_cycle_counter(threshold_add);
+    *threshold_cycles += threshold_add;
+    #endif
+
+    if (!col_cnt_im2col)
+    {
+      pA+=(2*num_col_im2col_w);
+    }
+    else
+    {
+      pA+=num_col_im2col_w;
+    }
+  }
+  else if (out_ch_left == 3)
+  {
+    #ifdef PROFILE
+    //printf("Started hotloop_prep counter\n");
+    reset_cycle_counter();
+    start_cycle_counter();
+    #endif
+    uint8_t *pB = pIn;
+    uint8_t *pB2 = (pB + num_col_im2col_a);
+
+    uint32_t *ptrB  = (uint32_t *) pB;
+    uint32_t *ptrB2 = (uint32_t *) pB2;
+
+    uint8_t *pA2 = (pA + num_col_im2col_w);
+    uint8_t *pA3 = (pA2 + num_col_im2col_w);
+
+    uint32_t *ptrA  = (uint32_t *) pA ;
+    uint32_t *ptrA2 = (uint32_t *) pA2;
+    uint32_t *ptrA3 = (uint32_t *) pA3;
+    ptrA = MacLoadInit(1, 0, 0, 0, ptrA);
+    ptrA2 = MacLoadInit(1, 0, 1, 0, ptrA2);
+    ptrA3 = MacLoadInit(1, 0, 2, 0, ptrA3);
+
+  
+    ptrB = MacLoadInit(0, 1, 0, 0, ptrB);
+    ptrB2 = MacLoadInit(0, 1, 0, 1, ptrB2);
+  
+    int sum = 0;
+    int sum2 = 0;
+    int sum3 = 0;
+    int sum4 = 0;
+    int sum5 = 0;
+    int sum6 = 0;
+
+    if (pBias != NULL)
+    {
+      sum = ((int) (*pBias++));
+      sum2 = ((int) (*pBias++));
+      sum3 = ((int) (*pBias++));
+
+      sum4 = sum;
+      sum5 = sum2;
+      sum6 = sum3;
+    }
+    #ifdef PROFILE
+    stop_cycle_counter();
+    //printf("Stopped hotloop_prep counter\n");
+    read_cycle_counter(hotloop_prep_add);
+    *hotloop_prep_cycles += hotloop_prep_add;
+    #endif
+
+    #ifdef PROFILE
+    //printf("Started hotloop counter\n");
+    reset_cycle_counter();
+    start_cycle_counter();
+    #endif
+    
+    hotloop_3x2(
+      &sum, &sum2, &sum3, &sum4, &sum5, &sum6,
+      ptrA, ptrA2, ptrA3, ptrB, ptrB2, num_col_im2col_words
+    );
+
+    #ifdef PROFILE
+    stop_cycle_counter();
+    //printf("Stopped hotloop counter\n");
+    read_cycle_counter(hotloop_add);
+    *hotloop_cycles += hotloop_add;
+    #endif
+
+    #ifdef PROFILE
+    //printf("Started hotloop_leftover counter\n");
+    reset_cycle_counter();
+    start_cycle_counter();
+    #endif
+    int col_cnt_im2col = num_col_im2col_c & 0xf;
+
+    if (col_cnt_im2col)
+    {
+      uint16_t loop_cnt_im2col_w = (num_col_im2col_words) << 2;
+      pA+=loop_cnt_im2col_w;
+      pA2+=loop_cnt_im2col_w;
+      pA3+=loop_cnt_im2col_w;
+
+      uint16_t loop_cnt_im2col_a = (num_col_im2col_words) << 2;
+      pB+=loop_cnt_im2col_a;
+      pB2+=loop_cnt_im2col_a;
+
+      volatile uint32_t valA, valA2, valA3, valB, valB2;
+      // pack the remaining weights and activations into 32-bit vectors
+      // padding with 0xd9 because ternary_decoder(0xd9) = 0000000000
+      if (col_cnt_im2col == 4)
+      {
+        valA = 0xd9 << 24 | 0xd9 << 16 | 0xd9 << 8 | *pA;
+        valA2 = 0xd9 << 24 | 0xd9 << 16 | 0xd9 << 8 | *pA2;
+        valA3 = 0xd9 << 24 | 0xd9 << 16 | 0xd9 << 8 | *pA3;
+
+        valB = 0xd9 << 24 | 0xd9 << 16 | 0xd9 << 8 | *pB;
+        valB2 = 0xd9 << 24 | 0xd9 << 16 | 0xd9 << 8 | *pB2;
+      }
+      else if (col_cnt_im2col == 8)
+      {
+        valA = 0xd9 << 24 | 0xd9 << 16 | *(pA + 1) << 8 | *pA;
+        valA2 = 0xd9 << 24 | 0xd9 << 16 | *(pA2 + 1) << 8 | *pA2;
+        valA3 = 0xd9 << 24 | 0xd9 << 16 | *(pA3 + 1) << 8 | *pA3;
+
+        valB = 0xd9 << 24 | 0xd9 << 16 | *(pB + 1) << 8 | *pB;
+        valB2 = 0xd9 << 24 | 0xd9 << 16 | *(pB2 + 1) << 8 | *pB2;
+      }
+      else // col_cnt_im2col == 12
+      {
+        valA = 0xd9 << 24 | *(pA + 2) << 16 | *(pA + 1) << 8 | *pA;
+        valA2 = 0xd9 << 24 | *(pA2 + 2) << 16 | *(pA2 + 1) << 8 | *pA2;
+        valA3 = 0xd9 << 24 | *(pA3 + 2) << 16 | *(pA3 + 1) << 8 | *pA3;
+
+        valB = 0xd9 << 24 | *(pB + 2) << 16 | *(pB + 1) << 8 | *pB;
+        valB2 = 0xd9 << 24 | *(pB2 + 2) << 16 | *(pB2 + 1) << 8 | *pB2;
+      }
+
+      uint32_t *pA_p = &valA;
+      uint32_t *pA2_p = &valA2;
+      uint32_t *pA3_p = &valA3;
+
+      uint32_t *pB_p = &valB;
+      uint32_t *pB2_p = &valB2;
+
+      pA_p = MacLoadInit(1, 0, 0, 0, pA_p);
+      pA2_p = MacLoadInit(1, 0, 1, 0, pA2_p);
+      pA3_p = MacLoadInit(1, 0, 2, 0, pA3_p);
+      pB_p = MacLoadInit(0, 1, 0, 0, pB_p);
+      pB2_p = MacLoadInit(0, 1, 0, 1, pB2_p);
+  
+      pA += PACK_INT2_SIZE(col_cnt_im2col);
+
+      sum = MacLoads20(0, 0, 0, 0, pA_p, sum);
+      sum2 = MacLoads20(0, 0, 1, 0, pA2_p, sum2);
+      sum3 = MacLoads20(0, 0, 2, 0, pA3_p, sum3);
+
+      sum4 = MacLoads20(0, 0, 0, 1, pA_p, sum4);
+      sum5 = MacLoads20(0, 0, 1, 1, pA2_p, sum5);
+      sum6 = MacLoads20(0, 0, 2, 1, pA3_p, sum6);
+    }
+    #ifdef PROFILE
+    stop_cycle_counter();
+    //printf("Stopped hotloop_leftover counter\n");
+    read_cycle_counter(hotloop_leftover_add);
+    *hotloop_leftover_cycles += hotloop_leftover_add;
+    #endif
+
+    #ifdef PROFILE
+    //printf("Started threshold counter\n");
+    reset_cycle_counter();
+    start_cycle_counter();
+    #endif
+    res1 = thr_cmp(res1, sum, *currThr);
+    check_store(res1, pOut);
+    res2 = thr_cmp(res2, sum4, *currThr++);
+    check_store(res2, pOut2);
+    reset_currThr();
+
+    res1 = thr_cmp(res1, sum2, *currThr);
+    check_store(res1, pOut);
+    res2 = thr_cmp(res2, sum5, *currThr++);
+    check_store(res2, pOut2);
+    reset_currThr();
+
+    res1 = thr_cmp(res1, sum3, *currThr);
+    check_store(res1, pOut);
+    res2 = thr_cmp(res2, sum6, *currThr++);
+    check_store(res2, pOut2);
+    reset_currThr();
+
+    #ifdef PROFILE
+    stop_cycle_counter();
+    //printf("Stopped threshold counter\n");
+    read_cycle_counter(threshold_add);
+    *threshold_cycles += threshold_add;
+    #endif
+
+    if (!col_cnt_im2col)
+    {
+      pA+=(3*num_col_im2col_w);
+    }
+    else
+    {
+      pA+=(2*num_col_im2col_w);
+    }
+  }
+
+  *thrc_res1 = res1;
+  *thrc_res2 = res2;
+
+  pOut+=incr_val; // ch_out_r if a store was performed, else 0
+  return pOut;
+}
+
+
+
+
diff --git a/rt_nn_tests/xptnn_conv/xpulp_tnn_matmul_ternary_signed.h b/rt_nn_tests/xptnn_conv/xpulp_tnn_matmul_ternary_signed.h
new file mode 100644
index 0000000..1493b5d
--- /dev/null
+++ b/rt_nn_tests/xptnn_conv/xpulp_tnn_matmul_ternary_signed.h
@@ -0,0 +1,113 @@
+#ifndef __XPULP_TNN_MATMUL_TERNARY_SIGNED_H
+#define __XPULP_TNN_MATMUL_TERNARY_SIGNED_H
+
+#include "pulp_nn_utils.h"
+
+uint8_t * __attribute__((noinline)) xpulp_tnn_matmul_ternary_signed(
+                        uint8_t  *pIn,
+                        int8_t   *pBias,
+                        uint32_t *pThr,
+                        uint8_t  *pOut,
+                        uint8_t  *pOut2,
+                        uint8_t  *pWeight,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+#ifndef PROFILE
+                        uint32_t *thrc_res1,
+                        uint32_t *thrc_res2);
+#else
+                        uint32_t *thrc_res1,
+                        uint32_t *thrc_res2,
+                        uint32_t *hotloop_prep_cycles,
+                        uint32_t *hotloop_cycles,
+                        uint32_t *threshold_cycles,
+                        uint32_t *hotloop_leftover_cycles);
+#endif
+
+inline void __attribute__((aligned(4))) hotloop_4x2(
+  int *sum, int *sum2, int *sum3, int *sum4, int *sum5, int *sum6, int *sum7, int *sum8,
+  uint32_t *ptrA, uint32_t *ptrA2, uint32_t *ptrA3, uint32_t *ptrA4,
+  uint32_t *ptrB, uint32_t *ptrB2, int num_col_im2col_words
+){
+  for (int j=0; j<num_col_im2col_words; j++)
+  {
+    *sum = MacLoads20(0, 0, 0, 0, ptrA, *sum);
+    *sum2 = MacLoads20(0, 0, 1, 0, ptrA2, *sum2);
+    *sum3 = MacLoads20(0, 0, 2, 0, ptrA3, *sum3);
+    *sum4 = MacLoads20(0, 1, 3, 0, ptrB, *sum4);
+    ptrB = MacLoadUpdate(ptrB);
+
+    *sum5 = MacLoads20(1, 0, 0, 1, ptrA, *sum5);
+    ptrA = MacLoadUpdate(ptrA);
+    *sum6 = MacLoads20(1, 0, 1, 1, ptrA2, *sum6);
+    ptrA2 = MacLoadUpdate(ptrA2);
+    *sum7 = MacLoads20(1, 0, 2, 1, ptrA3, *sum7);
+    ptrA3 = MacLoadUpdate(ptrA3);
+    *sum8 = MacLoads20(1, 0, 3, 1, ptrA4, *sum8);
+    ptrA4 = MacLoadUpdate(ptrA4);
+
+    ptrB2 = MacLoadInit(0, 1, 0, 1, ptrB2);
+  }
+}
+
+
+inline void __attribute__((aligned(4))) hotloop_3x2(
+  int *sum, int *sum2, int *sum3, int *sum4, int *sum5, int *sum6,
+  uint32_t *ptrA, uint32_t *ptrA2, uint32_t *ptrA3, uint32_t *ptrB, uint32_t *ptrB2, int num_col_im2col_words
+){
+  for (int j=0; j<num_col_im2col_words; j++)
+  {
+    *sum = MacLoads20(0, 0, 0, 0, ptrA, *sum);
+    *sum2 = MacLoads20(0, 0, 1, 0, ptrA2, *sum2);
+    *sum3 = MacLoads20(0, 1, 2, 0, ptrB, *sum3);
+    ptrB = MacLoadUpdate(ptrB);
+
+    *sum4 = MacLoads20(1, 0, 0, 1, ptrA, *sum4);
+    ptrA = MacLoadUpdate(ptrA);
+    *sum5 = MacLoads20(1, 0, 1, 1, ptrA2, *sum5);
+    ptrA2 = MacLoadUpdate(ptrA2);
+    *sum6 = MacLoads20(1, 0, 2, 1, ptrA3, *sum6);
+    ptrA3 = MacLoadUpdate(ptrA3);
+
+    ptrB2 = MacLoadInit(0, 1, 0, 1, ptrB2);
+  }
+}
+
+
+inline void __attribute__((aligned(4))) hotloop_2x2(
+  int *sum, int *sum2, int *sum3, int *sum4,
+  uint32_t *ptrA, uint32_t *ptrA2, uint32_t *ptrB, uint32_t *ptrB2, int num_col_im2col_words
+){
+  for (int j=0; j<num_col_im2col_words; j++)
+  {
+    *sum = MacLoads20(0, 0, 0, 0, ptrA, *sum);
+    *sum2 = MacLoads20(0, 1, 1, 0, ptrB, *sum2);
+    ptrB = MacLoadUpdate(ptrB);
+
+    *sum3 = MacLoads20(1, 0, 0, 1, ptrA, *sum3);
+    ptrA = MacLoadUpdate(ptrA);
+    *sum4 = MacLoads20(1, 0, 1, 1, ptrA2, *sum4);
+    ptrA2 = MacLoadUpdate(ptrA2);
+
+    ptrB2 = MacLoadInit(0, 1, 0, 1, ptrB2);
+  }
+}
+
+
+inline void __attribute__((aligned(4))) hotloop_1x2(
+  int *sum, int *sum2, uint32_t *ptrA, uint32_t *ptrB, uint32_t *ptrB2, int num_col_im2col_words
+){
+  for (int j=0; j<num_col_im2col_words; j++)
+  {
+    *sum = MacLoads20(0, 1, 0, 0, ptrB, *sum);
+    ptrB = MacLoadUpdate(ptrB);
+    *sum2 = MacLoads20(1, 0, 0, 1, ptrA, *sum2);
+    ptrA = MacLoadUpdate(ptrA);
+
+    ptrB2 = MacLoadInit(0, 1, 0, 1, ptrB2);
+  }
+}
+
+
+
+#endif
\ No newline at end of file
diff --git a/rt_nn_tests/xptnn_conv/xpulp_tnn_matmul_ternary_signed_4x1.c b/rt_nn_tests/xptnn_conv/xpulp_tnn_matmul_ternary_signed_4x1.c
new file mode 100644
index 0000000..d088b68
--- /dev/null
+++ b/rt_nn_tests/xptnn_conv/xpulp_tnn_matmul_ternary_signed_4x1.c
@@ -0,0 +1,666 @@
+#include <stdio.h>
+#include <math.h>
+#include "pulp_nn_utils.h"
+#include "xpulp_tnn_matmul_ternary_signed_4x1.h"
+
+
+
+// TODO: review argument order
+uint8_t * __attribute__((noinline)) xpulp_tnn_matmul_ternary_signed_4x1(
+                        uint8_t  *pIn,
+                        int8_t   *pBias,
+                        uint32_t *pThr,
+                        uint8_t  *pOut,
+                        uint8_t  *pWeight,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+#ifndef PROFILE
+                        uint32_t *thrc_res)
+{
+#else
+                        uint32_t *thrc_res,
+                        uint32_t *hotloop_prep_cycles,
+                        uint32_t *hotloop_cycles,
+                        uint32_t *threshold_cycles,
+                        uint32_t *hotloop_leftover_cycles)
+{
+  uint32_t hotloop_prep_add;
+  uint32_t hotloop_add;
+  uint32_t threshold_add;
+  uint32_t hotloop_leftover_add;
+#endif
+  uint16_t ch_out_r = ch_out / 5;
+
+  uint16_t num_col_im2col_w = num_col_im2col / 5;
+  uint16_t num_col_im2col_a = num_col_im2col_w;
+  uint16_t num_col_im2col_c = num_col_im2col_w << 2;
+  // used for hotloop calls
+  uint16_t num_col_im2col_words = num_col_im2col_w >> 2;
+
+  uint8_t *pA = pWeight;
+  v2s *currThr = (v2s *) pThr;
+
+  int res;
+  res = *thrc_res;
+
+  for(int i=0; i < ch_out >> 2; i++)
+  {
+    #ifdef PROFILE
+    //printf("Started hotloop_prep counter\n");
+    reset_cycle_counter();
+    start_cycle_counter();
+    #endif
+    uint8_t *pB = pIn;
+
+    uint32_t *ptrB  = (uint32_t *) pB;
+
+    uint8_t *pA2 = (pA + num_col_im2col_w);
+    uint8_t *pA3 = (pA2 + num_col_im2col_w);
+    uint8_t *pA4 = (pA3 + num_col_im2col_w);
+
+    uint32_t *ptrA  = (uint32_t *) pA ;
+    uint32_t *ptrA2 = (uint32_t *) pA2;
+    uint32_t *ptrA3 = (uint32_t *) pA3;
+    uint32_t *ptrA4 = (uint32_t *) pA4;
+
+    ptrA = MacLoadInit(1, 0, 0, 0, ptrA);
+    ptrA2 = MacLoadInit(1, 0, 1, 0, ptrA2);
+    ptrA3 = MacLoadInit(1, 0, 2, 0, ptrA3);
+    ptrA4 = MacLoadInit(1, 0, 3, 0, ptrA4);
+
+  
+    ptrB = MacLoadInit(0, 1, 0, 0, ptrB);
+  
+    int sum = 0;
+    int sum2 = 0;
+    int sum3 = 0;
+    int sum4 = 0;
+
+    if (pBias != NULL)
+    {
+      sum = ((int) (*pBias++));
+      sum2 = ((int) (*pBias++));
+      sum3 = ((int) (*pBias++));
+      sum4 = ((int) (*pBias++));
+    }
+    #ifdef PROFILE
+    stop_cycle_counter();
+    //printf("Stopped hotloop_prep counter\n");
+    read_cycle_counter(hotloop_prep_add);
+    *hotloop_prep_cycles += hotloop_prep_add;
+    #endif
+
+    #ifdef PROFILE
+    //printf("Started hotloop counter\n");
+    reset_cycle_counter();
+    start_cycle_counter();
+    #endif
+    
+    hotloop_4x1(
+      &sum, &sum2, &sum3, &sum4, ptrA, ptrA2, ptrA3, ptrA4, ptrB, num_col_im2col_words
+    );
+
+    #ifdef PROFILE
+    stop_cycle_counter();
+    //printf("Stopped hotloop counter\n");
+    read_cycle_counter(hotloop_add);
+    *hotloop_cycles += hotloop_add;
+    #endif
+
+    #ifdef PROFILE
+    //printf("Started hotloop_leftover counter\n");
+    reset_cycle_counter();
+    start_cycle_counter();
+    #endif
+    int col_cnt_im2col = num_col_im2col_c & 0xf;
+
+    if (col_cnt_im2col)
+    {
+      uint16_t loop_cnt_im2col_w = (num_col_im2col_words) << 2;
+      pA+=loop_cnt_im2col_w;
+      pA2+=loop_cnt_im2col_w;
+      pA3+=loop_cnt_im2col_w;
+      pA4+=loop_cnt_im2col_w;
+
+      uint16_t loop_cnt_im2col_a = (num_col_im2col_words) << 2;
+      pB+=loop_cnt_im2col_a;
+
+      volatile uint32_t valA, valA2, valA3, valA4, valB;
+      // pack the remaining weights and activations into 32-bit vectors
+      // padding with 0xd9 because ternary_decoder(0xd9) = 0000000000
+      if (col_cnt_im2col == 4)
+      {
+        valA = 0xd9 << 24 | 0xd9 << 16 | 0xd9 << 8 | *pA;
+        valA2 = 0xd9 << 24 | 0xd9 << 16 | 0xd9 << 8 | *pA2;
+        valA3 = 0xd9 << 24 | 0xd9 << 16 | 0xd9 << 8 | *pA3;
+        valA4 = 0xd9 << 24 | 0xd9 << 16 | 0xd9 << 8 | *pA4;
+
+        valB = 0xd9 << 24 | 0xd9 << 16 | 0xd9 << 8 | *pB;
+      }
+      else if (col_cnt_im2col == 8)
+      {
+        valA = 0xd9 << 24 | 0xd9 << 16 | *(pA + 1) << 8 | *pA;
+        valA2 = 0xd9 << 24 | 0xd9 << 16 | *(pA2 + 1) << 8 | *pA2;
+        valA3 = 0xd9 << 24 | 0xd9 << 16 | *(pA3 + 1) << 8 | *pA3;
+        valA4 = 0xd9 << 24 | 0xd9 << 16 | *(pA4 + 1) << 8 | *pA4;
+
+        valB = 0xd9 << 24 | 0xd9 << 16 | *(pB + 1) << 8 | *pB;
+      }
+      else // col_cnt_im2col == 12
+      {
+        valA = 0xd9 << 24 | *(pA + 2) << 16 | *(pA + 1) << 8 | *pA;
+        valA2 = 0xd9 << 24 | *(pA2 + 2) << 16 | *(pA2 + 1) << 8 | *pA2;
+        valA3 = 0xd9 << 24 | *(pA3 + 2) << 16 | *(pA3 + 1) << 8 | *pA3;
+        valA4 = 0xd9 << 24 | *(pA4 + 2) << 16 | *(pA4 + 1) << 8 | *pA4;
+
+        valB = 0xd9 << 24 | *(pB + 2) << 16 | *(pB + 1) << 8 | *pB;
+      }
+
+      uint32_t *pA_p = &valA;
+      uint32_t *pA2_p = &valA2;
+      uint32_t *pA3_p = &valA3;
+      uint32_t *pA4_p = &valA4;
+
+      uint32_t *pB_p = &valB;
+
+      pA_p = MacLoadInit(1, 0, 0, 0, pA_p);
+      pA2_p = MacLoadInit(1, 0, 1, 0, pA2_p);
+      pA3_p = MacLoadInit(1, 0, 2, 0, pA3_p);
+      pA4_p = MacLoadInit(1, 0, 3, 0, pA4_p);
+      pB_p = MacLoadInit(0, 1, 0, 0, pB_p);
+  
+      pA += PACK_INT2_SIZE(col_cnt_im2col);
+
+      sum = MacLoads20(0, 0, 0, 0, pA_p, sum);
+      sum2 = MacLoads20(0, 0, 1, 0, pA2_p, sum2);
+      sum3 = MacLoads20(0, 0, 2, 0, pA3_p, sum3);
+      sum4 = MacLoads20(0, 0, 3, 0, pA4_p, sum4);
+    }
+    #ifdef PROFILE
+    stop_cycle_counter();
+    //printf("Stopped hotloop_leftover counter\n");
+    read_cycle_counter(hotloop_leftover_add);
+    *hotloop_leftover_cycles += hotloop_leftover_add;
+    #endif
+
+    #ifdef PROFILE
+    //printf("Started threshold counter\n");
+    reset_cycle_counter();
+    start_cycle_counter();
+    #endif
+    res = thr_cmp(res, sum, *currThr++);
+    check_store_4x1(res, pOut);
+    reset_currThr();
+
+    res = thr_cmp(res, sum2, *currThr++);
+    check_store_4x1(res, pOut);
+    reset_currThr();
+
+    res = thr_cmp(res, sum3, *currThr++);
+    check_store_4x1(res, pOut);
+    reset_currThr();
+
+    res = thr_cmp(res, sum4, *currThr++);
+    check_store_4x1(res, pOut);
+    reset_currThr();
+
+    #ifdef PROFILE
+    stop_cycle_counter();
+    //printf("Stopped threshold counter\n");
+    read_cycle_counter(threshold_add);
+    *threshold_cycles += threshold_add;
+    #endif
+
+    if (!col_cnt_im2col)
+    {
+      pA+=(4*num_col_im2col_w);
+    }
+    else
+    {
+      pA+=(3*num_col_im2col_w);
+    }
+  }
+
+  // leftover part : the hotloop above produces 4N output channels. If out_ch not divisible
+  // by 4, the remaining output channels are computed below
+  int out_ch_left = ch_out & 0x3;
+
+  if (out_ch_left == 1)
+  {
+    #ifdef PROFILE
+    //printf("Started hotloop_prep counter\n");
+    reset_cycle_counter();
+    start_cycle_counter();
+    #endif
+    uint8_t *pB = pIn;
+
+    uint32_t *ptrB  = (uint32_t *) pB;
+
+    uint32_t *ptrA  = (uint32_t *) pA ;
+
+    ptrA = MacLoadInit(1, 0, 0, 0, ptrA);
+
+  
+    ptrB = MacLoadInit(0, 1, 0, 0, ptrB);
+  
+    int sum = 0;
+
+    if (pBias != NULL)
+    {
+      sum = ((int) (*pBias++));
+    }
+    #ifdef PROFILE
+    stop_cycle_counter();
+    //printf("Stopped hotloop_prep counter\n");
+    read_cycle_counter(hotloop_prep_add);
+    *hotloop_prep_cycles += hotloop_prep_add;
+    #endif
+
+    #ifdef PROFILE
+    //printf("Started hotloop counter\n");
+    reset_cycle_counter();
+    start_cycle_counter();
+    #endif
+    
+    hotloop_1x1(
+      &sum, ptrA, ptrB, num_col_im2col_words
+    );
+
+    #ifdef PROFILE
+    stop_cycle_counter();
+    //printf("Stopped hotloop counter\n");
+    read_cycle_counter(hotloop_add);
+    *hotloop_cycles += hotloop_add;
+    #endif
+
+    #ifdef PROFILE
+    //printf("Started hotloop_leftover counter\n");
+    reset_cycle_counter();
+    start_cycle_counter();
+    #endif
+    int col_cnt_im2col = num_col_im2col_c & 0xf;
+
+    if (col_cnt_im2col)
+    {
+      uint16_t loop_cnt_im2col_w = (num_col_im2col_words) << 2;
+      pA+=loop_cnt_im2col_w;
+
+      uint16_t loop_cnt_im2col_a = (num_col_im2col_words) << 2;
+      pB+=loop_cnt_im2col_a;
+
+      volatile uint32_t valA, valB;
+      // pack the remaining weights and activations into 32-bit vectors
+      // padding with 0xd9 because ternary_decoder(0xd9) = 0000000000
+      if (col_cnt_im2col == 4)
+      {
+        valA = 0xd9 << 24 | 0xd9 << 16 | 0xd9 << 8 | *pA;
+
+        valB = 0xd9 << 24 | 0xd9 << 16 | 0xd9 << 8 | *pB;
+      }
+      else if (col_cnt_im2col == 8)
+      {
+        valA = 0xd9 << 24 | 0xd9 << 16 | *(pA + 1) << 8 | *pA;
+
+        valB = 0xd9 << 24 | 0xd9 << 16 | *(pB + 1) << 8 | *pB;
+      }
+      else // col_cnt_im2col == 12
+      {
+        valA = 0xd9 << 24 | *(pA + 2) << 16 | *(pA + 1) << 8 | *pA;
+
+        valB = 0xd9 << 24 | *(pB + 2) << 16 | *(pB + 1) << 8 | *pB;
+      }
+
+      uint32_t *pA_p = &valA;
+
+      uint32_t *pB_p = &valB;
+
+      pA_p = MacLoadInit(1, 0, 0, 0, pA_p);
+      pB_p = MacLoadInit(0, 1, 0, 0, pB_p);
+  
+      pA += PACK_INT2_SIZE(col_cnt_im2col);
+
+      sum = MacLoads20(0, 0, 0, 0, pA_p, sum);
+    }
+    #ifdef PROFILE
+    stop_cycle_counter();
+    //printf("Stopped hotloop_leftover counter\n");
+    read_cycle_counter(hotloop_leftover_add);
+    *hotloop_leftover_cycles += hotloop_leftover_add;
+    #endif
+
+    #ifdef PROFILE
+    //printf("Started threshold counter\n");
+    reset_cycle_counter();
+    start_cycle_counter();
+    #endif
+    res = thr_cmp(res, sum, *currThr++);
+    check_store_4x1(res, pOut);
+    reset_currThr();
+
+    #ifdef PROFILE
+    stop_cycle_counter();
+    //printf("Stopped threshold counter\n");
+    read_cycle_counter(threshold_add);
+    *threshold_cycles += threshold_add;
+    #endif
+
+    if (!col_cnt_im2col)
+    {
+      pA+=num_col_im2col_w;
+    }
+  }
+  else if (out_ch_left == 2)
+  {
+    #ifdef PROFILE
+    //printf("Started hotloop_prep counter\n");
+    reset_cycle_counter();
+    start_cycle_counter();
+    #endif
+    uint8_t *pB = pIn;
+
+    uint32_t *ptrB  = (uint32_t *) pB;
+
+    uint8_t *pA2 = (pA + num_col_im2col_w);
+
+    uint32_t *ptrA  = (uint32_t *) pA ;
+    uint32_t *ptrA2 = (uint32_t *) pA2;
+
+    ptrA = MacLoadInit(1, 0, 0, 0, ptrA);
+    ptrA2 = MacLoadInit(1, 0, 1, 0, ptrA2);
+
+  
+    ptrB = MacLoadInit(0, 1, 0, 0, ptrB);
+  
+    int sum = 0;
+    int sum2 = 0;
+
+    if (pBias != NULL)
+    {
+      sum = ((int) (*pBias++));
+      sum2 = ((int) (*pBias++));
+    }
+    #ifdef PROFILE
+    stop_cycle_counter();
+    //printf("Stopped hotloop_prep counter\n");
+    read_cycle_counter(hotloop_prep_add);
+    *hotloop_prep_cycles += hotloop_prep_add;
+    #endif
+
+    #ifdef PROFILE
+    //printf("Started hotloop counter\n");
+    reset_cycle_counter();
+    start_cycle_counter();
+    #endif
+    
+    hotloop_2x1(
+      &sum, &sum2, ptrA, ptrA2, ptrB, num_col_im2col_words
+    );
+
+    #ifdef PROFILE
+    stop_cycle_counter();
+    //printf("Stopped hotloop counter\n");
+    read_cycle_counter(hotloop_add);
+    *hotloop_cycles += hotloop_add;
+    #endif
+
+    #ifdef PROFILE
+    //printf("Started hotloop_leftover counter\n");
+    reset_cycle_counter();
+    start_cycle_counter();
+    #endif
+    int col_cnt_im2col = num_col_im2col_c & 0xf;
+
+    if (col_cnt_im2col)
+    {
+      uint16_t loop_cnt_im2col_w = (num_col_im2col_words) << 2;
+      pA+=loop_cnt_im2col_w;
+      pA2+=loop_cnt_im2col_w;
+
+      uint16_t loop_cnt_im2col_a = (num_col_im2col_words) << 2;
+      pB+=loop_cnt_im2col_a;
+
+      volatile uint32_t valA, valA2, valB;
+      // pack the remaining weights and activations into 32-bit vectors
+      // padding with 0xd9 because ternary_decoder(0xd9) = 0000000000
+      if (col_cnt_im2col == 4)
+      {
+        valA = 0xd9 << 24 | 0xd9 << 16 | 0xd9 << 8 | *pA;
+        valA2 = 0xd9 << 24 | 0xd9 << 16 | 0xd9 << 8 | *pA2;
+
+        valB = 0xd9 << 24 | 0xd9 << 16 | 0xd9 << 8 | *pB;
+      }
+      else if (col_cnt_im2col == 8)
+      {
+        valA = 0xd9 << 24 | 0xd9 << 16 | *(pA + 1) << 8 | *pA;
+        valA2 = 0xd9 << 24 | 0xd9 << 16 | *(pA2 + 1) << 8 | *pA2;
+
+        valB = 0xd9 << 24 | 0xd9 << 16 | *(pB + 1) << 8 | *pB;
+      }
+      else // col_cnt_im2col == 12
+      {
+        valA = 0xd9 << 24 | *(pA + 2) << 16 | *(pA + 1) << 8 | *pA;
+        valA2 = 0xd9 << 24 | *(pA2 + 2) << 16 | *(pA2 + 1) << 8 | *pA2;
+
+        valB = 0xd9 << 24 | *(pB + 2) << 16 | *(pB + 1) << 8 | *pB;
+      }
+
+      uint32_t *pA_p = &valA;
+      uint32_t *pA2_p = &valA2;
+
+      uint32_t *pB_p = &valB;
+
+      pA_p = MacLoadInit(1, 0, 0, 0, pA_p);
+      pA2_p = MacLoadInit(1, 0, 1, 0, pA2_p);
+      pB_p = MacLoadInit(0, 1, 0, 0, pB_p);
+  
+      pA += PACK_INT2_SIZE(col_cnt_im2col);
+
+      sum = MacLoads20(0, 0, 0, 0, pA_p, sum);
+      sum2 = MacLoads20(0, 0, 1, 0, pA2_p, sum2);
+    }
+    #ifdef PROFILE
+    stop_cycle_counter();
+    //printf("Stopped hotloop_leftover counter\n");
+    read_cycle_counter(hotloop_leftover_add);
+    *hotloop_leftover_cycles += hotloop_leftover_add;
+    #endif
+
+    #ifdef PROFILE
+    //printf("Started threshold counter\n");
+    reset_cycle_counter();
+    start_cycle_counter();
+    #endif
+    res = thr_cmp(res, sum, *currThr++);
+    check_store_4x1(res, pOut);
+    reset_currThr();
+
+    res = thr_cmp(res, sum2, *currThr++);
+    check_store_4x1(res, pOut);
+    reset_currThr();
+
+    #ifdef PROFILE
+    stop_cycle_counter();
+    //printf("Stopped threshold counter\n");
+    read_cycle_counter(threshold_add);
+    *threshold_cycles += threshold_add;
+    #endif
+
+    if (!col_cnt_im2col)
+    {
+      pA+=(2*num_col_im2col_w);
+    }
+    else
+    {
+      pA+=num_col_im2col_w;
+    }
+  }
+  else if (out_ch_left == 3)
+  {
+    #ifdef PROFILE
+    //printf("Started hotloop_prep counter\n");
+    reset_cycle_counter();
+    start_cycle_counter();
+    #endif
+    uint8_t *pB = pIn;
+
+    uint32_t *ptrB  = (uint32_t *) pB;
+
+    uint8_t *pA2 = (pA + num_col_im2col_w);
+    uint8_t *pA3 = (pA2 + num_col_im2col_w);
+
+    uint32_t *ptrA  = (uint32_t *) pA ;
+    uint32_t *ptrA2 = (uint32_t *) pA2;
+    uint32_t *ptrA3 = (uint32_t *) pA3;
+    ptrA = MacLoadInit(1, 0, 0, 0, ptrA);
+    ptrA2 = MacLoadInit(1, 0, 1, 0, ptrA2);
+    ptrA3 = MacLoadInit(1, 0, 2, 0, ptrA3);
+
+  
+    ptrB = MacLoadInit(0, 1, 0, 0, ptrB);
+  
+    int sum = 0;
+    int sum2 = 0;
+    int sum3 = 0;
+
+    if (pBias != NULL)
+    {
+      sum = ((int) (*pBias++));
+      sum2 = ((int) (*pBias++));
+      sum3 = ((int) (*pBias++));
+    }
+    #ifdef PROFILE
+    stop_cycle_counter();
+    //printf("Stopped hotloop_prep counter\n");
+    read_cycle_counter(hotloop_prep_add);
+    *hotloop_prep_cycles += hotloop_prep_add;
+    #endif
+
+    #ifdef PROFILE
+    //printf("Started hotloop counter\n");
+    reset_cycle_counter();
+    start_cycle_counter();
+    #endif
+    
+    hotloop_3x1(
+      &sum, &sum2, &sum3,
+      ptrA, ptrA2, ptrA3, ptrB, num_col_im2col_words
+    );
+
+    #ifdef PROFILE
+    stop_cycle_counter();
+    //printf("Stopped hotloop counter\n");
+    read_cycle_counter(hotloop_add);
+    *hotloop_cycles += hotloop_add;
+    #endif
+
+    #ifdef PROFILE
+    //printf("Started hotloop_leftover counter\n");
+    reset_cycle_counter();
+    start_cycle_counter();
+    #endif
+    int col_cnt_im2col = num_col_im2col_c & 0xf;
+
+    if (col_cnt_im2col)
+    {
+      uint16_t loop_cnt_im2col_w = (num_col_im2col_words) << 2;
+      pA+=loop_cnt_im2col_w;
+      pA2+=loop_cnt_im2col_w;
+      pA3+=loop_cnt_im2col_w;
+
+      uint16_t loop_cnt_im2col_a = (num_col_im2col_words) << 2;
+      pB+=loop_cnt_im2col_a;
+
+      volatile uint32_t valA, valA2, valA3, valB;
+      // pack the remaining weights and activations into 32-bit vectors
+      // padding with 0xd9 because ternary_decoder(0xd9) = 0000000000
+      if (col_cnt_im2col == 4)
+      {
+        valA = 0xd9 << 24 | 0xd9 << 16 | 0xd9 << 8 | *pA;
+        valA2 = 0xd9 << 24 | 0xd9 << 16 | 0xd9 << 8 | *pA2;
+        valA3 = 0xd9 << 24 | 0xd9 << 16 | 0xd9 << 8 | *pA3;
+
+        valB = 0xd9 << 24 | 0xd9 << 16 | 0xd9 << 8 | *pB;
+      }
+      else if (col_cnt_im2col == 8)
+      {
+        valA = 0xd9 << 24 | 0xd9 << 16 | *(pA + 1) << 8 | *pA;
+        valA2 = 0xd9 << 24 | 0xd9 << 16 | *(pA2 + 1) << 8 | *pA2;
+        valA3 = 0xd9 << 24 | 0xd9 << 16 | *(pA3 + 1) << 8 | *pA3;
+
+        valB = 0xd9 << 24 | 0xd9 << 16 | *(pB + 1) << 8 | *pB;
+      }
+      else // col_cnt_im2col == 12
+      {
+        valA = 0xd9 << 24 | *(pA + 2) << 16 | *(pA + 1) << 8 | *pA;
+        valA2 = 0xd9 << 24 | *(pA2 + 2) << 16 | *(pA2 + 1) << 8 | *pA2;
+        valA3 = 0xd9 << 24 | *(pA3 + 2) << 16 | *(pA3 + 1) << 8 | *pA3;
+
+        valB = 0xd9 << 24 | *(pB + 2) << 16 | *(pB + 1) << 8 | *pB;
+      }
+
+      uint32_t *pA_p = &valA;
+      uint32_t *pA2_p = &valA2;
+      uint32_t *pA3_p = &valA3;
+
+      uint32_t *pB_p = &valB;
+
+      pA_p = MacLoadInit(1, 0, 0, 0, pA_p);
+      pA2_p = MacLoadInit(1, 0, 1, 0, pA2_p);
+      pA3_p = MacLoadInit(1, 0, 2, 0, pA3_p);
+      pB_p = MacLoadInit(0, 1, 0, 0, pB_p);
+  
+      pA += PACK_INT2_SIZE(col_cnt_im2col);
+
+      sum = MacLoads20(0, 0, 0, 0, pA_p, sum);
+      sum2 = MacLoads20(0, 0, 1, 0, pA2_p, sum2);
+      sum3 = MacLoads20(0, 0, 2, 0, pA3_p, sum3);
+    }
+    #ifdef PROFILE
+    stop_cycle_counter();
+    //printf("Stopped hotloop_leftover counter\n");
+    read_cycle_counter(hotloop_leftover_add);
+    *hotloop_leftover_cycles += hotloop_leftover_add;
+    #endif
+
+    #ifdef PROFILE
+    //printf("Started threshold counter\n");
+    reset_cycle_counter();
+    start_cycle_counter();
+    #endif
+    res = thr_cmp(res, sum, *currThr++);
+    check_store_4x1(res, pOut);
+    reset_currThr();
+
+    res = thr_cmp(res, sum2, *currThr++);
+    check_store_4x1(res, pOut);
+    reset_currThr();
+
+    res = thr_cmp(res, sum3, *currThr++);
+    check_store_4x1(res, pOut);
+    reset_currThr();
+
+    #ifdef PROFILE
+    stop_cycle_counter();
+    //printf("Stopped threshold counter\n");
+    read_cycle_counter(threshold_add);
+    *threshold_cycles += threshold_add;
+    #endif
+
+    if (!col_cnt_im2col)
+    {
+      pA+=(3*num_col_im2col_w);
+    }
+    else
+    {
+      pA+=(2*num_col_im2col_w);
+    }
+  }
+
+  *thrc_res = res;
+
+  return pOut;
+}
+
+
+
+
diff --git a/rt_nn_tests/xptnn_conv/xpulp_tnn_matmul_ternary_signed_4x1.h b/rt_nn_tests/xptnn_conv/xpulp_tnn_matmul_ternary_signed_4x1.h
new file mode 100644
index 0000000..a847bfb
--- /dev/null
+++ b/rt_nn_tests/xptnn_conv/xpulp_tnn_matmul_ternary_signed_4x1.h
@@ -0,0 +1,96 @@
+#ifndef __XPULP_TNN_MATMUL_TERNARY_SIGNED_4X1_H
+#define __XPULP_TNN_MATMUL_TERNARY_SIGNED_4X1_H
+
+#include "pulp_nn_utils.h"
+
+uint8_t * __attribute__((noinline)) xpulp_tnn_matmul_ternary_signed_4x1(
+                        uint8_t  *pIn,
+                        int8_t   *pBias,
+                        uint32_t *pThr,
+                        uint8_t  *pOut,
+                        uint8_t  *pWeight,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+#ifndef PROFILE
+                        uint32_t *thrc_res);
+#else
+                        uint32_t *thrc_res,
+                        uint32_t *hotloop_prep_cycles,
+                        uint32_t *hotloop_cycles,
+                        uint32_t *threshold_cycles,
+                        uint32_t *hotloop_leftover_cycles);
+#endif
+
+inline void __attribute__((aligned(4))) hotloop_4x1(
+  int *sum, int *sum2, int *sum3, int *sum4,
+  uint32_t *ptrA, uint32_t *ptrA2, uint32_t *ptrA3, uint32_t *ptrA4,
+  uint32_t *ptrB, int num_col_im2col_words
+){
+  for (int j=0; j<num_col_im2col_words; j++)
+  {
+    *sum = MacLoads20(1, 0, 0, 0, ptrA, *sum);
+    ptrA = MacLoadUpdate(ptrA);
+    *sum2 = MacLoads20(1, 0, 1, 0, ptrA2, *sum2);
+    ptrA2 = MacLoadUpdate(ptrA2);
+    *sum3 = MacLoads20(1, 0, 2, 0, ptrA3, *sum3);
+    ptrA3 = MacLoadUpdate(ptrA3);
+    *sum4 = MacLoads20(1, 0, 3, 0, ptrA4, *sum4);
+    ptrA4 = MacLoadUpdate(ptrA4);
+
+    ptrB = MacLoadInit(0, 1, 0, 0, ptrB);
+    asm volatile ("nop;");
+  }
+}
+
+
+inline void __attribute__((aligned(4))) hotloop_3x1(
+  int *sum, int *sum2, int *sum3,
+  uint32_t *ptrA, uint32_t *ptrA2, uint32_t *ptrA3, uint32_t *ptrB, int num_col_im2col_words
+){
+  for (int j=0; j<num_col_im2col_words; j++)
+  {
+    *sum = MacLoads20(1, 0, 0, 0, ptrA, *sum);
+    ptrA = MacLoadUpdate(ptrA);
+    *sum2 = MacLoads20(1, 0, 1, 0, ptrA2, *sum2);
+    ptrA2 = MacLoadUpdate(ptrA2);
+    *sum3 = MacLoads20(1, 0, 2, 0, ptrA3, *sum3);
+    ptrA3 = MacLoadUpdate(ptrA3);
+
+    ptrB = MacLoadInit(0, 1, 0, 0, ptrB);
+    asm volatile ("nop;");
+  }
+}
+
+
+inline void __attribute__((aligned(4))) hotloop_2x1(
+  int *sum, int *sum2, uint32_t *ptrA, uint32_t *ptrA2, uint32_t *ptrB, int num_col_im2col_words
+){
+  for (int j=0; j<num_col_im2col_words; j++)
+  {
+    *sum = MacLoads20(1, 0, 0, 0, ptrA, *sum);
+    ptrA = MacLoadUpdate(ptrA);
+    *sum2 = MacLoads20(1, 0, 1, 0, ptrA2, *sum2);
+    ptrA2 = MacLoadUpdate(ptrA2);
+
+    ptrB = MacLoadInit(0, 1, 0, 0, ptrB);
+    asm volatile ("nop;");
+  }
+}
+
+
+inline void __attribute__((aligned(4))) hotloop_1x1(
+  int *sum, uint32_t *ptrA, uint32_t *ptrB, int num_col_im2col_words
+){
+  for (int j=0; j<num_col_im2col_words; j++)
+  {
+    *sum = MacLoads20(1, 0, 0, 0, ptrA, *sum);
+    ptrA = MacLoadUpdate(ptrA);
+
+    ptrB = MacLoadInit(0, 1, 0, 0, ptrB);
+    asm volatile ("nop;");
+  }
+}
+
+
+
+#endif
\ No newline at end of file
diff --git a/rt_nn_tests/xptnn_conv/xpulp_tnn_maxpool_ternary.h b/rt_nn_tests/xptnn_conv/xpulp_tnn_maxpool_ternary.h
new file mode 100644
index 0000000..2b65fca
--- /dev/null
+++ b/rt_nn_tests/xptnn_conv/xpulp_tnn_maxpool_ternary.h
@@ -0,0 +1,20 @@
+#ifndef __XPULP_TNN_MAXPOOL_TERNARY_H
+#define __XPULP_TNN_MAXPOOL_TERNARY_H
+void __attribute__ ((noinline)) xpulp_tnn_maxpool_ternary(
+                                                          int8_t * pIn,
+                                                          int8_t * pOut,
+                                                          uint16_t  dim_im_in_x,
+                                                          uint16_t  dim_im_in_y,
+                                                          // ch_im_in: UNCOMPRESSED number of input/output channels
+                                                          uint16_t  ch_im_in,
+                                                          uint16_t  dim_im_out_x,
+                                                          uint16_t  dim_im_out_y,
+                                                          uint16_t  dim_kernel_x,
+                                                          uint16_t  dim_kernel_y,
+                                                          uint16_t  padding_t,
+                                                          uint16_t  padding_b,
+                                                          uint16_t  padding_l,
+                                                          uint16_t  padding_r,
+                                                          uint16_t  stride_x,
+                                                          uint16_t  stride_y);
+#endif
diff --git a/rt_nn_tests/xptnn_linear/Makefile b/rt_nn_tests/xptnn_linear/Makefile
new file mode 100644
index 0000000..3bcbb3d
--- /dev/null
+++ b/rt_nn_tests/xptnn_linear/Makefile
@@ -0,0 +1,18 @@
+APP = test
+PULP_APP = test
+PULP_APP_SRCS = test.c
+PULP_APP_SRCS += xpulp_tnn_linear_ternary_i32_signed.c
+
+CORE=8
+
+
+PULP_CFLAGS += -DNUM_CORES=$(CORE) -I. -O3
+PULP_LDFLAGS += -lc -lm -lgcc -Wl,-print-memory-usage 
+
+PULP_APP_CFLAGS += -DNUM_CORES=$(CORE) -I. -O3
+PULP_APP_LDFLAGS += -lc -lm -lgcc -Wl,-print-memory-usage 
+
+PULP_CFLAGS += -DPULP_RUNTIME -DSINGLE_CORE_DMA -DALWAYS_BLOCK_DMA_TRANSFERS
+
+
+include $(PULP_SDK_HOME)/install/rules/pulp.mk
diff --git a/rt_nn_tests/xptnn_linear/data_statstest.h b/rt_nn_tests/xptnn_linear/data_statstest.h
new file mode 100644
index 0000000..bf0d1a8
--- /dev/null
+++ b/rt_nn_tests/xptnn_linear/data_statstest.h
@@ -0,0 +1,2153 @@
+// this file is generated automatically by the python script `generate_statstest.py`
+
+#ifndef _DATA_H
+#define _DATA_H
+
+
+
+#define IM2COL_DIM (1 * NUM_CORES * 2) // for NUM_CORES cores
+#define OUTPUT_DIM 1
+#define OUTPUT_DIM_FP (100)
+uint8_t pWeight_0 [] = {
+0x95,
+0xd9,
+0x0d,
+0xcd,
+0xba,
+0xc2,
+0xc3,
+0x26,
+0xa9,
+0x73,
+0x92,
+0xa5,
+0x32,
+0xcb,
+0x91,
+0x27,
+0xda,
+0x61,
+0xe2,
+0xfc,
+0x8c,
+0x66,
+0xb7,
+0xa8,
+0xb6,
+0x85,
+0xf6,
+0x07,
+0x55,
+0xdb,
+0x88,
+0xdb,
+0xd1,
+0x5e,
+0xd5,
+0xe6,
+0x61,
+0xe7,
+0x4b,
+0x81,
+0x2e,
+0x36,
+0xeb,
+0x5a,
+0x34,
+0x92,
+0x8c,
+0x28,
+0x68,
+0x14,
+0x77,
+0xde,
+0x97,
+0x86,
+0x46,
+0x2d,
+0x06,
+0x42,
+0x73,
+0xa5,
+0x28,
+0x17,
+0x44,
+0x8f,
+0x44,
+0xd0,
+0x7a,
+0xaf,
+0xec,
+0xb8,
+0xa7,
+0xcf,
+0xc5,
+0x78,
+0x4a,
+0xa0,
+0x83,
+0x77,
+0x22,
+0x48,
+0x01,
+0xe3,
+0x21,
+0x51,
+0x27,
+0x43,
+0xc5,
+0xf6,
+0x76,
+0x9d,
+0x6f,
+0x8b,
+0xb8,
+0x94,
+0x1f,
+0xfe,
+0xa6,
+0x46,
+0x64,
+0x4c,
+0x76,
+0x8b,
+0xe6,
+0x08,
+0x27,
+0xa5,
+0xc0,
+0x33,
+0xa9,
+0x9d,
+0x80,
+0x8d,
+0x61,
+0x81,
+0x0e,
+0x3b,
+0x17,
+0x8f,
+0xbf,
+0x7e,
+0x8b,
+0x9d,
+0xe9,
+0x05,
+0x75,
+0x03,
+0x48,
+0x9b,
+0x1c,
+0x4d,
+0x28,
+0xc4,
+0x40,
+0xea,
+0xf1,
+0x54,
+0x44,
+0xcd,
+0xa7,
+0x13,
+0x43,
+0x1f,
+0x69,
+0x1c,
+0x20,
+0x0c,
+0xa2,
+0x92,
+0x10,
+0x6d,
+0xe0,
+0x92,
+0x73,
+0xbc,
+0xff,
+0xfa,
+0xa5,
+0xff,
+0x1c,
+0xa2,
+0xcd,
+0x8a,
+0x2d,
+0x40,
+0x05,
+0x14,
+0x2b,
+0x97,
+0xf6,
+0x57,
+0xe4,
+0x76,
+0xe3,
+0x5b,
+0x69,
+0xb6,
+0xef,
+0x03,
+0x56,
+0xb2,
+0x2f,
+0xc1,
+0xb8,
+0xf1,
+0x02,
+0xb4,
+0xcf,
+0xca,
+0x7b,
+0x4b,
+0xcc,
+0x94,
+0x56,
+0x97,
+0x2a,
+0x69,
+0x7c,
+0x94,
+0x74,
+0x37,
+0x0c,
+0x1d,
+0x1b,
+0xe7,
+0x63,
+0xd1,
+0xa5,
+0xd1,
+0xfc,
+0x29,
+0x1f,
+0x55,
+0xd7,
+0x2b,
+0x62,
+0x62,
+0x9f,
+0xf5,
+0x96,
+0xf3,
+0x73,
+0x06,
+0x42,
+0x95,
+0x42,
+0x58,
+0xf0,
+0xb2,
+0x42,
+0xc7,
+0x11,
+0xaa,
+0x58,
+0x81,
+0x5a,
+0x09,
+0x5d,
+0x21,
+0x0e,
+0xe5,
+0x63,
+0x90,
+0xe9,
+0x53,
+0x6d,
+0xcd,
+0xf3,
+0x0d,
+0xde,
+0x51,
+0x56,
+0x3f,
+0xfd,
+0xb4,
+0xcc,
+0x12,
+0x0d,
+0x64,
+0xcb,
+0x73,
+0x5b,
+0x05,
+0xfc,
+0xb5,
+0x98,
+0xfd,
+0xd7,
+0xc7,
+0x0c,
+0xc6,
+0x1f,
+0x65,
+0x5f,
+0x55,
+0x31,
+0x3c,
+0x94,
+0x0b,
+0x0b,
+0x84,
+0x7d,
+0x4e,
+0xd6,
+0x10,
+0xc3,
+0xa8,
+0xa2,
+0x7c,
+0x3b,
+0xd8,
+0x9f,
+0x35,
+0x31,
+0x97,
+0x05,
+0xb2,
+0x04,
+0xba,
+0xec,
+0x5a,
+0xc7,
+0x60,
+0x32,
+0x41,
+0x45,
+0x57,
+0x2c,
+0x8c,
+0x9a,
+0xb7,
+0x84,
+0x68,
+0x71,
+0xcd,
+0xe6,
+0x0d,
+0xe9,
+0xd4,
+0x05,
+0xfc,
+0x07,
+0x1c,
+0xe2,
+0x42,
+0xde,
+0x6a,
+0x22,
+0xe1,
+0xc4,
+0x20,
+0x43,
+0xa0,
+0xfa,
+0x2c,
+0x77,
+0xca,
+0xd8,
+0x1d,
+0x7a,
+0x2c,
+0x9d,
+0x0a,
+0x4c,
+0x46,
+0x5f,
+0xde,
+0x1a,
+0x7f,
+0xd9,
+0x93,
+0x37,
+0x3b,
+0xd1,
+0xdc,
+0xa0,
+0x15,
+0xf1,
+0xca,
+0x1d,
+0x26,
+0x7d,
+0xd4,
+0x38,
+0x75,
+0x49,
+0xb1,
+0xa2,
+0x24,
+0x8c,
+0x70,
+0x61,
+0x73,
+0xcb,
+0x3d,
+0x35,
+0xc7,
+0x39,
+0x78,
+0xd7,
+0x43,
+0x50,
+0xdf,
+0x1e,
+0x4c,
+0xdd,
+0x41,
+0x41,
+0x37,
+0x0b,
+0x7f,
+0x21,
+0x18,
+0xbc,
+0x84,
+0x0f,
+0x4d,
+0x62,
+0xbb,
+0x37,
+0x64,
+0x4d,
+0x7c,
+0x1a,
+0x8d,
+0x77,
+0xd7,
+0x57,
+0x55,
+0x1e,
+0xc3,
+0x7b,
+0xa4,
+0xc5,
+0x9c,
+0x4b,
+0xb1,
+0x4d,
+0xf0,
+0x18,
+0x09,
+0xc2,
+0x11,
+0x5e,
+0xab,
+0xbc,
+0x51,
+0x6d,
+0x7e,
+0x56,
+0x51,
+0x4a,
+0x05,
+0xc7,
+0x18,
+0x16,
+0x14,
+0x17,
+0x01,
+0x61,
+0xef,
+0x73,
+0x0e,
+0xed,
+0x27,
+0x11,
+0xbb,
+0x98,
+0xc9,
+0x01,
+0x36,
+0x40,
+0x27,
+0x50,
+0xbf,
+0xd1,
+0x7e,
+0xf3,
+0xb9,
+0x81,
+0x06,
+0xb6,
+0x50,
+0x02,
+0xde,
+0xe1,
+0xc6,
+0xda,
+0x13,
+0x31,
+0xd9,
+0xbd,
+0x69,
+0x6b,
+0xa3,
+0x51,
+0xbf,
+0x67,
+0xff,
+0x6f,
+0x57,
+0xed,
+0xaf,
+0x88,
+0xac,
+0x4d,
+0x03,
+0x4f,
+0x86,
+0x37,
+0xd5,
+0x2b,
+0x56,
+0x47,
+0x41,
+0x26,
+0x31,
+0x4c,
+0xce,
+0x20,
+0x1d,
+0x0a,
+0xdb,
+0xf6,
+0xbf,
+0x8c,
+0x6c,
+0x54,
+0xfc,
+0x23,
+0xfb,
+0x61,
+0x74,
+0xc1,
+0x1a,
+0x8f,
+0x2c,
+0x53,
+0xb0,
+0x1d,
+0x6d,
+0xc5,
+0x17,
+0xb8,
+0xd8,
+0xaf,
+0x96,
+0x03,
+0xab,
+0x01,
+0xde,
+0x85,
+0xd4,
+0x68,
+0x05,
+0x6e,
+0xe6,
+0xea,
+0x38,
+0x40,
+0x77,
+0x1b,
+0xe6,
+0x93,
+0xc3,
+0xb8,
+0x5c,
+0x3f,
+0xe9,
+0xec,
+0x61,
+0x67,
+0xa6,
+0xa4,
+0x5d,
+0xb9,
+0x5c,
+0x06,
+0x01,
+0xbc,
+0xb1,
+0xc0,
+0x46,
+0x2b,
+0x69,
+0x92,
+0x37,
+0x35,
+0x3f,
+0x80,
+0xcb,
+0xe1,
+0x22,
+0x2a,
+0xab,
+0x4c,
+0xf4,
+0x60,
+0xa1,
+0xb8,
+0x22,
+0x49,
+0x4d,
+0x98,
+0x90,
+0xf3,
+0x2d,
+0x71,
+0xf7,
+0xfb,
+0xf3,
+0xeb,
+0x04,
+0x3b,
+0x16,
+0x4d,
+0x84,
+0x3c,
+0xab,
+0xf3,
+0xd1,
+0x30,
+0x45,
+0x46,
+0x44,
+0xc0,
+0xec,
+0x95,
+0x40,
+0x92,
+0x5a,
+0x95,
+0x76,
+0xc2,
+0x0c,
+0x84,
+0xd3,
+0xda,
+0x23,
+0x4c,
+0xa6,
+0xfc,
+0xa8,
+0x5c,
+0xd4,
+0x88,
+0x63,
+0xaa,
+0x81,
+0x7c,
+0xee,
+0x70,
+0xc3,
+0xc4,
+0x25,
+0x96,
+0x1a,
+0x2a,
+0x57,
+0x6f,
+0x23,
+0x27,
+0x05,
+0x37,
+0xdb,
+0x1c,
+0xfa,
+0x2a,
+0x76,
+0xa8,
+0xe8,
+0xec,
+0x55,
+0x94,
+0xaf,
+0xec,
+0x03,
+0x41,
+0x5e,
+0xbd,
+0x56,
+0x04,
+0xb0,
+0x6a,
+0x9b,
+0x05,
+0x89,
+0xd4,
+0x73,
+0x55,
+0x6d,
+0xf1,
+0xec,
+0x2d,
+0xee,
+0x8f,
+0x89,
+0xed,
+0x88,
+0xf3,
+0x98,
+0x18,
+0xbb,
+0x80,
+0x65,
+0x21,
+0x1b,
+0x3e,
+0x53,
+0xd8,
+0x6e,
+0xc6,
+0x69,
+0xc5,
+0x42,
+0xaf,
+0xe0,
+0xa6,
+0x0f,
+0xbb,
+0xa9,
+0x6c,
+0xcb,
+0x1b,
+0x8b,
+0x9a,
+0xa4,
+0xb9,
+0x81,
+0x5f,
+0x29,
+0x9f,
+0x3c,
+0x24,
+0xe1,
+0x4f,
+0xea,
+0x39,
+0x7d,
+0xbd,
+0xd9,
+0x94,
+0x8b,
+0x71,
+0xe1,
+0xb8,
+0xf4,
+0xcc,
+0x8a,
+0x44,
+0xdb,
+0xb4,
+0x11,
+0x65,
+0x67,
+0x51,
+0xe5,
+0x38,
+0x09,
+0xff,
+0x10,
+0x24,
+0xc3,
+0x23,
+0x8a,
+0xe7,
+0x35,
+0x34,
+0xc7,
+0xe8,
+0xa5,
+0x05,
+0xeb,
+0x06,
+0x4e,
+0xd1,
+0xa8,
+0x46,
+0xa9,
+0xa1,
+0x9c,
+0x2b,
+0x96,
+0x7b,
+0x56,
+0x55,
+0x4c,
+0x44,
+0x94,
+0x74,
+0xdd,
+0xb3,
+0x0c,
+0x57,
+0x4e,
+0xbc,
+0xe8,
+0xf1,
+0xc8,
+0x85,
+0x34,
+0x7a,
+0x24,
+0x4e,
+0xb6,
+0x3a,
+0x26,
+0xbc,
+0xc7,
+0xfa,
+0xe5,
+0x0a,
+0x91,
+0xd6,
+0xd3,
+0x2d,
+0xdb,
+0x3e,
+0x1b,
+0x92,
+0xcf,
+0x48,
+0x6b,
+0xb6,
+0x82,
+0xbc,
+0x6f,
+0x9b,
+0x2b,
+0x92,
+0x05,
+0xdf,
+0xd6,
+0x23,
+0x90,
+0xe8,
+0x26,
+0x7f,
+0xee,
+0xe7,
+0xec,
+0xec,
+0xbb,
+0x9c,
+0xde,
+0x29,
+0x28,
+0x7a,
+0xef,
+0xce,
+0x18,
+0x0a,
+0x1c,
+0x1a,
+0x87,
+0x77,
+0xcd,
+0x6f,
+0x8a,
+0x98,
+0xc6,
+0x11,
+0xe4,
+0xcc,
+0xfb,
+0x68,
+0xc0,
+0xb2,
+0x29,
+0x43,
+0x3d,
+0x9f,
+0x92,
+0x1f,
+0xb6,
+0x68,
+0x49,
+0x2c,
+0xe4,
+0xcd,
+0xed,
+0x82,
+0x62,
+0xa3,
+0xba,
+0xca,
+0x0a,
+0x92,
+0xfa,
+0x5e,
+0x1e,
+0x27,
+0xde,
+0xe8,
+0x77,
+0xe2,
+0x97,
+0x64,
+0x34,
+0xb4,
+0x05,
+0xaa,
+0xc9,
+0x41,
+0xba,
+0x06,
+0x96,
+0x26,
+0x61,
+0x09,
+0xd7,
+0x9b,
+0x7a,
+0xc4,
+0x22,
+0x87,
+0xc7,
+0x48,
+0xb5,
+0xb1,
+0x90,
+0x3e,
+0x34,
+0xb5,
+0xa4,
+0x81,
+0xd3,
+0x54,
+0x07,
+0xe6,
+0xf5,
+0x63,
+0x87,
+0x91,
+0x89,
+0x1a,
+0xa8,
+0xa2,
+0x9d,
+0x92,
+0xd7,
+0xb2,
+0x2d,
+0xa5,
+0x45,
+0x85,
+0x42,
+0x74,
+0x80,
+0x42,
+0xf1,
+0x48,
+0x98,
+0x98,
+0x9b,
+0xb9,
+0x92,
+0x32,
+0x44,
+0xd9,
+0x3e,
+0xb8,
+0x23,
+0xfe,
+0x69,
+0xa9,
+0x4c,
+0x84,
+0xb1,
+0x71,
+0x4e,
+0x4c,
+0x01,
+0x06,
+0xc7,
+0xa9,
+0xf6,
+0xd3,
+0x10,
+0x0a,
+0x82,
+0x48,
+0x48,
+0xb9,
+0xf5,
+0xad,
+0xb1,
+0xff,
+0xb3,
+0x00,
+0x60,
+0x31,
+0x2d,
+0x70,
+0x05,
+0xf6,
+0xcf,
+0xdb,
+0xb6,
+0x47,
+0xa9,
+0x42,
+0xed,
+0x6c,
+0xe9,
+0x95,
+0xea,
+0x29,
+0x1a,
+0x1e,
+0x89,
+0x3c,
+0x5b,
+0x3f,
+0x66,
+0x25,
+0xb2,
+0x66,
+0xec,
+0x61,
+0xf5,
+0x04,
+0xad,
+0xac,
+0xed,
+0x3c,
+0xb1,
+0x9f,
+0x03,
+0x6f,
+0x12,
+0x48,
+0x7d,
+0x35,
+0xa8,
+0x7a,
+0x28,
+0x48,
+0xb7,
+0x28,
+0xdf,
+0x3a,
+0x92,
+0x03,
+0x75,
+0xc8,
+0x16,
+0xaf,
+0x34,
+0xd7,
+0x2b,
+0xe8,
+0x0e,
+0x61,
+0x02,
+0x64,
+0x81,
+0xe8,
+0x6d,
+0x5f,
+0x5a,
+0x17,
+0x68,
+0xcf,
+0xf6,
+0x84,
+0xf1,
+0xa5,
+0xf7,
+0x47,
+0xad,
+0x50,
+0x65,
+0x39,
+0x74,
+0xf5,
+0x4c,
+0xca,
+0x97,
+0xd1,
+0x5d,
+0x70,
+0x7d,
+0xa9,
+0xaf,
+0xe2,
+0x9a,
+0x33,
+0x5e,
+0xdb,
+0xb1,
+0xaa,
+0x17,
+0xea,
+0x84,
+0x7c,
+0x48,
+0x62,
+0x8c,
+0x86,
+0x00,
+0x92,
+0x3d,
+0xcc,
+0xfd,
+0x61,
+0x68,
+0x3f,
+0xe7,
+0x0a,
+0xfb,
+0x1a,
+0x49,
+0x40,
+0x09,
+0x27,
+0x08,
+0xd4,
+0x04,
+0x03,
+0xb1,
+0x80,
+0x27,
+0xa4,
+0x89,
+0x63,
+0x4c,
+0x22,
+0xfb,
+0x66,
+0x61,
+0x5d,
+0x6a,
+0xf0,
+0xed,
+0x6f,
+0x17,
+0xbb,
+0xa1,
+0x05,
+0xe9,
+0x15,
+0xda,
+0xb7,
+0xa4,
+0x36,
+0xe2,
+0xb1,
+0xc2,
+0x7f,
+0x1a,
+0xf7,
+0xca,
+0xbc,
+0x5b,
+0xfc,
+0x01,
+0x65,
+0x14,
+0xb7,
+0xfd,
+0xd3,
+0x56,
+0xdc,
+0x92,
+0xda,
+0x16,
+0x55,
+0x1e,
+0xa1,
+0xa1,
+0xcf,
+0x15,
+0x60,
+0x87,
+0xe9,
+0x04,
+0xed,
+0xac,
+0x1b,
+0xa1,
+0xf6,
+0x57,
+0xf6,
+0xe7,
+0xd4,
+0x1d,
+0x95,
+0x7f,
+0xee,
+0x25,
+0x61,
+0x0f,
+0x00,
+0x85,
+0x5d,
+0xd9,
+0x22,
+0x25,
+0x15,
+0x18,
+0xb0,
+0x20,
+0x9b,
+0xb2,
+0x45,
+0x91,
+0x87,
+0xf1,
+0x8c,
+0xfc,
+0x50,
+0xb3,
+0x57,
+0x22,
+0x7b,
+0x64,
+0xd7,
+0x36,
+0x43,
+0x1d,
+0xe1,
+0x4f,
+0x8a,
+0x7f,
+0xee,
+0x4e,
+0x64,
+0xb0,
+0x5b,
+0x70,
+0x2d,
+0xfa,
+0x75,
+0xec,
+0x17,
+0x45,
+0x3d,
+0xa7,
+0x7a,
+0x6c,
+0x2b,
+0xa7,
+0xc8,
+0xb4,
+0xcf,
+0xea,
+0xa0,
+0x1d,
+0x16,
+0x7c,
+0xde,
+0x06,
+0xd7,
+0x67,
+0x1c,
+0xfd,
+0x82,
+0xa4,
+0xb9,
+0x98,
+0x0b,
+0x3e,
+0x47,
+0x5f,
+0xb6,
+0x48,
+0x76,
+0x5e,
+0x67,
+0x2f,
+0xcc,
+0x64,
+0xe3,
+0x8f,
+0x09,
+0xbf,
+0xa6,
+0xaf,
+0x89,
+0xe9,
+0x4e,
+0x66,
+0xed,
+0x0f,
+0xeb,
+0x2d,
+0x58,
+0xb6,
+0x7f,
+0x23,
+0xe7,
+0x1b,
+0xe9,
+0xfa,
+0xe5,
+0x2e,
+0xb7,
+0xc1,
+0x23,
+0xa3,
+0x60,
+0xd6,
+0x4c,
+0x8c,
+0xb0,
+0x78,
+0x61,
+0xf5,
+0xe6,
+0x5e,
+0xb6,
+0x97,
+0xca,
+0xd9,
+0xe8,
+0xe4,
+0x37,
+0xef,
+0x93,
+0x48,
+0x82,
+0xa0,
+0xb0,
+0xe1,
+0x33,
+0x5d,
+0x6f,
+0x0f,
+0xcc,
+0x47,
+0x06,
+0x82,
+0x49,
+0x09,
+0xdd,
+0x7e,
+0xc2,
+0x41,
+0x07,
+0x6b,
+0xab,
+0x95,
+0x0e,
+0x04,
+0x1b,
+0x60,
+0xc8,
+0x8d,
+0x5b,
+0x90,
+0xa4,
+0x9f,
+0x61,
+0x9b,
+0xe9,
+0xfc,
+0x2a,
+0xe7,
+0xfe,
+0x03,
+0x5d,
+0x39,
+0x09,
+0xf3,
+0x11,
+0xdd,
+0xc7,
+0x8f,
+0xda,
+0x2d,
+0x24,
+0xfc,
+0x7c,
+0xb4,
+0x61,
+0x0e,
+0xed,
+0xd3,
+0x6d,
+0x9a,
+0x0e,
+0x17,
+0x25,
+0xfb,
+0xac,
+0xc4,
+0xf0,
+0x35,
+0x81,
+0x2c,
+0x40,
+0x16,
+0xdd,
+0x91,
+0x36,
+0x82,
+0x2c,
+0x23,
+0xaa,
+0x14,
+0x46,
+0xc9,
+0x1b,
+0x1a,
+0x18,
+0xc5,
+0x08,
+0xee,
+0x93,
+0xaf,
+0x37,
+0x56,
+0x02,
+0x8f,
+0x54,
+0x71,
+0x9f,
+0x05,
+0xe2,
+0x9b,
+0xc1,
+0x69,
+0xdf,
+0xdc,
+0x8a,
+0x82,
+0x73,
+0x50,
+0x1b,
+0x28,
+0xcc,
+0x42,
+0xeb,
+0x1b,
+0x9a,
+0xa2,
+0x94,
+0x9b,
+0xc4,
+0x50,
+0xd5,
+0xf4,
+0x06,
+0x16,
+0x2a,
+0xfa,
+0x05,
+0xcb,
+0x4c,
+0x08,
+0x23,
+0x80,
+0x74,
+0x5a,
+0x18,
+0xba,
+0x49,
+0x68,
+0x32,
+0x6c,
+0xeb,
+0xb3,
+0x4f,
+0xca,
+0xd6,
+0x84,
+0x87,
+0x84,
+0x87,
+0x3d,
+0xdb,
+0xdb,
+0x1b,
+0xa3,
+0xd1,
+0xe7,
+0x3e,
+0xea,
+0xee,
+0x6a,
+0xe9,
+0x97,
+0xe5,
+0x5f,
+0xde,
+0x29,
+0x43,
+0xab,
+0x46,
+0xda,
+0x14,
+0x7d,
+0x6d,
+0x96,
+0x7b,
+0x25,
+0x53,
+0xa1,
+0x80,
+0x4d,
+0xdd,
+0xca,
+0x2e,
+0xcd,
+0x25,
+0x40,
+0xe1,
+0x7f,
+0xa6,
+0xd9,
+0x1d,
+0x9f,
+0x60,
+0xdf,
+0xb3,
+0xd6,
+0x95,
+0x29,
+0x98,
+0x15,
+0x98,
+0x68,
+0xa5,
+0xcd,
+0xe9,
+0xf6,
+0xf4,
+0x36,
+0x0f,
+0xa0,
+0xe4,
+0xff,
+0x8b,
+0xad,
+0x49,
+0x66,
+0xb0,
+0xc9,
+0x0a,
+0x7f,
+0x84,
+0xd4,
+0xd3,
+0x6e,
+0x17,
+0x91,
+0x95,
+0xd8,
+0xa3,
+0xb8,
+0x58,
+0xbb,
+0x02,
+0x1d,
+0x2e,
+0x70,
+0x5c,
+0x65,
+0x22,
+0x22,
+0x61,
+0x98,
+0xa3,
+0x54,
+0xb4,
+0x4c,
+0xfe,
+0x97,
+0xe6,
+0xf3,
+0x15,
+0x96,
+0x1c,
+0x3a,
+0x18,
+0xc6,
+0x74,
+0x41,
+0x3f,
+0x4e,
+0x60,
+0x10,
+0x7c,
+0xe1,
+0xa3,
+0x4c,
+0xec,
+0x96,
+0xbb,
+0x9f,
+0xd0,
+0x24,
+0x93,
+0xdd,
+0xbc,
+0x43,
+0xe9,
+0xa9,
+0xa9,
+0xe1,
+0x5a,
+0x77,
+0xfa,
+0x8b,
+0x7a,
+0x95,
+0xd4,
+0xec,
+0x0a,
+0x33,
+0x7d,
+0x61,
+0xe8,
+0x77,
+0xa5,
+0x04,
+0x17,
+0x8a,
+0xf3,
+0x02,
+0xa1,
+0x67,
+0x12,
+0x8d,
+0xdc,
+0x07,
+0x69,
+0x7e,
+0xd1,
+0x09,
+0x33,
+0xb0,
+0x91,
+0x4d,
+0x1e,
+0xde,
+0xa5,
+0x1b,
+0xe4,
+0x25,
+0x27,
+0x83,
+0xe1,
+0xd8,
+0xe4,
+0x4d,
+0x2c,
+0xac,
+0xec,
+0x30,
+0x5d,
+0x1b,
+0x50,
+0x22,
+0x83,
+0x8d,
+0x45,
+0x0a,
+0x46,
+0xda,
+0x83,
+0x08,
+0xa9,
+0xfe,
+0xba,
+0x1d,
+0x2e,
+0xed,
+0x33,
+0x24,
+0x40,
+0xbb,
+0x6b,
+0xcb,
+0x37,
+0x94,
+0xa7,
+0x48,
+0x7c,
+0x89,
+0x6c,
+0xb9,
+0x58,
+0x94,
+0xab,
+0xec,
+0xc9,
+0x81,
+0x5e,
+0x12,
+0xf4,
+0xa3,
+0x64,
+0x47,
+0x06,
+0xb8,
+0x21,
+0x2b,
+0x18,
+0x63,
+0x33,
+0x69,
+0xfb,
+0xa2,
+0x3d,
+0xdb,
+0xff,
+0xc4,
+0x73,
+0xf7,
+0xc5,
+0x54,
+0xa6,
+0xfe,
+0xd0,
+0x69,
+0xb9,
+0xb8,
+0x82,
+0x6e,
+0xfe,
+0x26,
+0x4d,
+0xec,
+0xa0,
+0xf5,
+0x5a,
+0xa6,
+0xfe,
+0x9d,
+0xc5,
+0x32,
+0x21,
+0x98,
+0x98,
+0xfc,
+0x9c,
+0x98,
+0x34,
+0x71,
+0xbd,
+0x4e,
+0x51,
+0x68,
+0xf4,
+0xa8,
+0x3d,
+0x2f,
+0x02,
+0x25,
+0x45,
+0xf4,
+0xcb,
+0xf4,
+0xe5,
+0x4f,
+0xac,
+0xa7,
+0x8b,
+0x01,
+0x69,
+0xd3,
+0xed,
+0x1e,
+0xbb,
+0x84,
+0xa9,
+0xe1,
+0xee,
+0xef,
+0xe3,
+0x1d,
+0xd4,
+0x9d,
+0x8c,
+0x7a,
+0xe5,
+0x5e,
+0x3c,
+0x18,
+0x6c,
+0x23,
+0x92,
+0x6b,
+0x25,
+0xbd,
+0x1d,
+0x87,
+0xbd,
+0xf0,
+0x58,
+0xe5,
+0x61,
+0x3d,
+0x66,
+0x60,
+0xaf,
+0xbf,
+0x4b,
+0x6b,
+0xe3,
+0xe0,
+0x69,
+0x9f,
+0xed,
+0xf0,
+0xb3,
+0xf0,
+0x54,
+0x4c,
+0x47,
+0x9c,
+0x60,
+0xc3,
+0x32,
+0x17,
+0x68,
+0x73,
+0xce,
+0x50,
+0x14,
+0x38,
+0x4c,
+0x4a,
+0x8c,
+0x69,
+0x30,
+0xb3,
+0x70,
+0xc3,
+0xa7,
+0x1c,
+0x18,
+0x10,
+0x09,
+0x39,
+0xb5,
+0x36,
+0xf7,
+0xea,
+0x18,
+0x9d,
+0x08,
+0x03,
+0x6b,
+0x45,
+0x26,
+0x86,
+0xfb,
+0x9b,
+0x90,
+0xdf,
+0x63,
+0xec,
+0xe2,
+0x02,
+0xb7,
+0x97,
+0xaf,
+0x68,
+0x7b,
+0x87,
+0xa9,
+0x3b,
+0xad,
+0x01,
+0x1e,
+0xfe,
+0x0a,
+0x1f,
+0x77,
+0x56,
+0xee,
+0x9a,
+0xb1,
+0xfb,
+0x20,
+0x28,
+0x4a,
+0x05,
+0xaf,
+0x1b,
+0x3f,
+0x32,
+0x8b,
+0x36,
+0xd0,
+0x4d,
+0x42,
+0x45,
+0xec,
+0xe1,
+0x9a,
+0x38,
+0x9b,
+0xb1,
+0x65,
+0xb3,
+0x51,
+0x58,
+0x66,
+0x4a,
+0x5b,
+0xa0,
+0x94,
+0x44,
+0x43,
+0x28,
+0x0b,
+0xb0,
+0x98,
+0x4e,
+0x45,
+0x45,
+0xd8,
+0xf4,
+0x18,
+0xc1,
+0xc4,
+0x95,
+0x71,
+0xd0,
+0x0c,
+0x4b,
+0x38,
+0x9a,
+0x8a,
+0xf3,
+0x5f,
+0xa4,
+0x7e,
+0x27,
+0x0c,
+0x62,
+0x9a,
+0xbb,
+0x3a,
+0x58,
+0x29,
+0xdd,
+0x5e,
+0x03,
+0x35,
+0x30,
+0x95,
+0xbf,
+0x23,
+0xbb,
+0x71,
+0x0b,
+0xe7,
+0xeb,
+0x12,
+0x69,
+0xab,
+0x53,
+0xde,
+0xe3,
+0x54,
+0x76,
+0x86,
+0xc8,
+0x66,
+0xbd,
+0x0f,
+0x89,
+0xb6,
+0x1f,
+0x26,
+0x3c,
+0xd1,
+0xa6,
+0x18,
+0x39,
+0xf7,
+0xa2,
+0x8f,
+0xb8,
+0x9f,
+0xb2,
+0x73,
+0xc3,
+0x81,
+0x5f,
+0x3a,
+0x2e,
+0xaf,
+0x13,
+0xe2,
+0xd6,
+0x28,
+0x95,
+0xa2,
+0xb0,
+0x09,
+0x18,
+0xdd,
+0xc3,
+0x10,
+0xa0,
+0xa4,
+0x74,
+0xe2,
+0xe3,
+0x44,
+0xa2,
+0x03,
+0x5a,
+0x2b,
+0x8c,
+0xf0,
+0x77,
+0x02,
+0x55,
+0x00,
+0x6b,
+0x55,
+0x00,
+0x30,
+0xc8
+};
+
+uint8_t pIn_0 [] = {
+0xe0,
+0x8c,
+0x04,
+0x21,
+0x36,
+0x8c,
+0x07,
+0x51,
+0x42,
+0x68,
+0x85,
+0xc3,
+0x00,
+0xe1,
+0x8c,
+0x7b,
+0xb7,
+0x01,
+0x38,
+0x57
+};
+
+
+
+/* -----------EXPECTED OUTPUTS----------- */
+int32_t exp_outp_0[] = {
+0x00000006,
+0x00000001,
+0xfffffff9,
+0x00000003,
+0x00000000,
+0x0000000d,
+0x0000000e,
+0x00000002,
+0x00000000,
+0xfffffffc,
+0xfffffffe,
+0xfffffffb,
+0xfffffffa,
+0xffffffef,
+0xfffffff7,
+0x00000002,
+0xfffffff7,
+0x00000007,
+0xfffffffa,
+0xffffffff,
+0xffffffff,
+0xfffffff8,
+0x00000007,
+0x00000002,
+0x00000001,
+0xfffffff5,
+0x00000006,
+0xfffffffd,
+0x0000000d,
+0xfffffff8,
+0xfffffffe,
+0xfffffff9,
+0x00000006,
+0x00000009,
+0x0000000a,
+0x00000009,
+0xffffffff,
+0x00000005,
+0x0000000a,
+0xffffffff,
+0xffffffff,
+0xfffffffd,
+0x00000007,
+0x00000008,
+0x00000008,
+0xfffffffc,
+0x00000005,
+0x00000006,
+0xfffffff1,
+0xfffffff8,
+0x00000006,
+0x00000005,
+0xfffffff5,
+0xfffffffb,
+0x00000000,
+0x00000003,
+0xfffffff8,
+0x00000001,
+0x00000006,
+0xfffffffb,
+0x00000001,
+0x00000003,
+0x00000001,
+0xffffffff,
+0x0000000d,
+0x00000001,
+0x00000006,
+0xfffffffc,
+0xfffffffb,
+0x00000006,
+0xfffffff6,
+0x0000000a,
+0x00000005,
+0xfffffffb,
+0x00000006,
+0x00000006,
+0x00000002,
+0x00000004,
+0xfffffff3,
+0x00000000,
+0x00000005,
+0xfffffffb,
+0x00000007,
+0xfffffffe,
+0xfffffffb,
+0x00000001,
+0x0000000c,
+0x00000009,
+0xfffffffe,
+0x00000001,
+0xfffffffa,
+0x00000009,
+0x00000006,
+0xfffffff6,
+0xfffffff8,
+0x00000011,
+0x00000004,
+0x00000002,
+0x00000001,
+0xfffffffd
+};
+
+
+
+L1_DATA uint32_t threshs_l1 [100] = {0};
+L1_DATA int32_t kappa_l1 [100] = {0};
+L1_DATA int32_t lambda_l1 [100] = {0};
+L1_DATA uint8_t inp_l1 [20] = {0};
+L1_DATA uint8_t outp_l1 [400] = {0};
+L1_DATA int8_t wt_l1 [2000] = {0};
+L1_DATA uint8_t im2col_l1 [IM2COL_DIM] = {0};
+
+#endif
diff --git a/rt_nn_tests/xptnn_linear/pulp_nn_kernels.h b/rt_nn_tests/xptnn_linear/pulp_nn_kernels.h
new file mode 100644
index 0000000..f4ec977
--- /dev/null
+++ b/rt_nn_tests/xptnn_linear/pulp_nn_kernels.h
@@ -0,0 +1,26528 @@
+/*
+ * pulp_nn_kernels.h
+ * Nazareno Bruschi <nazareno.bruschi@unibo.it>
+ *
+ * Copyright (C) 2019-2020 University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef __XPULPNN_KERNELS__
+#define __XPULPNN_KERNELS__
+
+void xpulp_nn_conv_u8_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+#ifndef PROFILE
+                        uint8_t flag_batch_norm);
+#else
+  uint8_t flag_batch_norm,
+    uint32_t *im2col_cycles,
+    uint32_t *hotloop_prep_cycles,
+    uint32_t *hotloop_cycles,
+    uint32_t *requant_cycles,
+    uint32_t *hotloop_leftover_cycles,
+    uint32_t *matmul4x2_leftover_cycles);
+#endif
+                        
+
+void xpulp_nn_conv_i2_i2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_i8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_i8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_i8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_i4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_i4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_i4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_i2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_i2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_i2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_i8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_i8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_i8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_i4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_i4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_i4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_i2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_i2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_i2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_i8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_i8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_i8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_i4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_i4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_i4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_i2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_i2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_i2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u8_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i8_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u8_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i8_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u8_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i8_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u4_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i4_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u4_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i4_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u4_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i4_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u2_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i2_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u2_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i2_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u2_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i2_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u8_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i8_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u8_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i8_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u8_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i8_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u4_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i4_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u4_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i4_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u4_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i4_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u2_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i2_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u2_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i2_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u2_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i2_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u8_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i8_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u8_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i8_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u8_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i8_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u4_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i4_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u4_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i4_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u4_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i4_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u2_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i2_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u2_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i2_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+#ifndef PROFILE
+                        uint8_t flag_batch_norm);
+#else
+  uint8_t flag_batch_norm,
+    uint32_t *hotloop_prep_cycles,
+    uint32_t *hotloop_cycles,
+    uint32_t *requant_cycles,
+    uint32_t *hotloop_leftover_cycles);
+#endif
+
+uint8_t *xpulp_nn_matmul_u2_u2_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i2_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_i8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_i8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_i8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_i4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_i4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_i4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_i2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_i2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_i2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_i8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_i8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_i8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_i4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_i4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_i4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_i2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_i2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_i2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_i8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_i8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_i8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_i4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_i4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_i4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_i2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_i2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_i2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_i32_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_i8_i32_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_u8_i32_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_i8_i32_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_u8_i32_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_i8_i32_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_u4_i32_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_i4_i32_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_u4_i32_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_i4_i32_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_u4_i32_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_i4_i32_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_u2_i32_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_i2_i32_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_u2_i32_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_i2_i32_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_u2_i32_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_i2_i32_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_u8_u8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_i8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_u8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_i8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_u8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_i8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_u4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_i4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_u4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_i4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_u4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_i4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_u2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_i2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_u2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_i2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_u2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_i2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_u8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_i8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_u8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_i8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_u8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_i8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_u4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_i4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_u4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_i4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_u4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_i4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_u2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_i2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_u2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_i2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_u2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_i2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_u8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_i8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_u8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_i8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_u8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_i8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_u4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_i4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_u4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_i4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_u4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_i4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_u2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_i2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_u2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_i2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_u2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_i2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_maxpool_u8(
+                        uint8_t * pIn,
+                        uint8_t * pOut,
+                        uint16_t  dim_im_in_x,
+                        uint16_t  dim_im_in_y,
+                        uint16_t  ch_im_in,
+                        uint16_t  dim_im_out_x,
+                        uint16_t  dim_im_out_y,
+                        uint16_t  dim_kernel_x,
+                        uint16_t  dim_kernel_y,
+                        uint16_t  padding_t,
+                        uint16_t  padding_b,
+                        uint16_t  padding_l,
+                        uint16_t  padding_r,
+                        uint16_t  stride_x,
+                        uint16_t  stride_y);
+
+void xpulp_nn_maxpool_i8(
+                        int8_t * pIn,
+                        int8_t * pOut,
+                        uint16_t  dim_im_in_x,
+                        uint16_t  dim_im_in_y,
+                        uint16_t  ch_im_in,
+                        uint16_t  dim_im_out_x,
+                        uint16_t  dim_im_out_y,
+                        uint16_t  dim_kernel_x,
+                        uint16_t  dim_kernel_y,
+                        uint16_t  padding_t,
+                        uint16_t  padding_b,
+                        uint16_t  padding_l,
+                        uint16_t  padding_r,
+                        uint16_t  stride_x,
+                        uint16_t  stride_y);
+
+void xpulp_nn_maxpool_u4(
+                        uint8_t * pIn,
+                        uint8_t * pOut,
+                        uint16_t  dim_im_in_x,
+                        uint16_t  dim_im_in_y,
+                        uint16_t  ch_im_in,
+                        uint16_t  dim_im_out_x,
+                        uint16_t  dim_im_out_y,
+                        uint16_t  dim_kernel_x,
+                        uint16_t  dim_kernel_y,
+                        uint16_t  padding_t,
+                        uint16_t  padding_b,
+                        uint16_t  padding_l,
+                        uint16_t  padding_r,
+                        uint16_t  stride_x,
+                        uint16_t  stride_y);
+
+void xpulp_nn_maxpool_i4(
+                        int8_t * pIn,
+                        int8_t * pOut,
+                        uint16_t  dim_im_in_x,
+                        uint16_t  dim_im_in_y,
+                        uint16_t  ch_im_in,
+                        uint16_t  dim_im_out_x,
+                        uint16_t  dim_im_out_y,
+                        uint16_t  dim_kernel_x,
+                        uint16_t  dim_kernel_y,
+                        uint16_t  padding_t,
+                        uint16_t  padding_b,
+                        uint16_t  padding_l,
+                        uint16_t  padding_r,
+                        uint16_t  stride_x,
+                        uint16_t  stride_y);
+
+void xpulp_nn_maxpool_u2(
+                        uint8_t * pIn,
+                        uint8_t * pOut,
+                        uint16_t  dim_im_in_x,
+                        uint16_t  dim_im_in_y,
+                        uint16_t  ch_im_in,
+                        uint16_t  dim_im_out_x,
+                        uint16_t  dim_im_out_y,
+                        uint16_t  dim_kernel_x,
+                        uint16_t  dim_kernel_y,
+                        uint16_t  padding_t,
+                        uint16_t  padding_b,
+                        uint16_t  padding_l,
+                        uint16_t  padding_r,
+                        uint16_t  stride_x,
+                        uint16_t  stride_y);
+
+void xpulp_nn_maxpool_i2(
+                        int8_t * pIn,
+                        int8_t * pOut,
+                        uint16_t  dim_im_in_x,
+                        uint16_t  dim_im_in_y,
+                        uint16_t  ch_im_in,
+                        uint16_t  dim_im_out_x,
+                        uint16_t  dim_im_out_y,
+                        uint16_t  dim_kernel_x,
+                        uint16_t  dim_kernel_y,
+                        uint16_t  padding_t,
+                        uint16_t  padding_b,
+                        uint16_t  padding_l,
+                        uint16_t  padding_r,
+                        uint16_t  stride_x,
+                        uint16_t  stride_y);
+
+void xpulp_nn_avgpool_u8_u8(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u8_i8(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i8_u8(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i8_i8(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u8_u4(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u8_i4(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i8_u4(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i8_i4(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u8_u2(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u8_i2(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i8_u2(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i8_i2(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u4_u8(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u4_i8(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i4_u8(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i4_i8(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u4_u4(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u4_i4(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i4_u4(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i4_i4(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u4_u2(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u4_i2(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i4_u2(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i4_i2(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u2_u8(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u2_i8(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i2_u8(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i2_i8(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u2_u4(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u2_i4(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i2_u4(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i2_i4(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u2_u2(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u2_i2(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i2_u2(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i2_i2(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_add_u8_u8(
+                        uint8_t * pIn1,
+                        uint8_t * pIn2,
+                        uint8_t * pOut,
+                        uint16_t out_mult1,
+                        uint16_t out_mult2,
+                        uint16_t out_shift,
+                        uint16_t dim_im_in_x,
+                        uint16_t dim_im_in_y,
+                        uint16_t ch_im_in);
+
+void xpulp_nn_add_u8_u8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u4(
+                        uint8_t * pIn1,
+                        uint8_t * pIn2,
+                        uint8_t * pOut,
+                        uint16_t out_mult1,
+                        uint16_t out_mult2,
+                        uint16_t out_shift,
+                        uint16_t dim_im_in_x,
+                        uint16_t dim_im_in_y,
+                        uint16_t ch_im_in);
+
+void xpulp_nn_add_u8_u4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u2(
+                        uint8_t * pIn1,
+                        uint8_t * pIn2,
+                        uint8_t * pOut,
+                        uint16_t out_mult1,
+                        uint16_t out_mult2,
+                        uint16_t out_shift,
+                        uint16_t dim_im_in_x,
+                        uint16_t dim_im_in_y,
+                        uint16_t ch_im_in);
+
+void xpulp_nn_add_u8_u2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u4(
+                        uint8_t * pIn1,
+                        uint8_t * pIn2,
+                        uint8_t * pOut,
+                        uint16_t out_mult1,
+                        uint16_t out_mult2,
+                        uint16_t out_shift,
+                        uint16_t dim_im_in_x,
+                        uint16_t dim_im_in_y,
+                        uint16_t ch_im_in);
+
+void xpulp_nn_add_u4_u4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u2(
+                        uint8_t * pIn1,
+                        uint8_t * pIn2,
+                        uint8_t * pOut,
+                        uint16_t out_mult1,
+                        uint16_t out_mult2,
+                        uint16_t out_shift,
+                        uint16_t dim_im_in_x,
+                        uint16_t dim_im_in_y,
+                        uint16_t ch_im_in);
+
+void xpulp_nn_add_u4_u2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u2(
+                        uint8_t * pIn1,
+                        uint8_t * pIn2,
+                        uint8_t * pOut,
+                        uint16_t out_mult1,
+                        uint16_t out_mult2,
+                        uint16_t out_shift,
+                        uint16_t dim_im_in_x,
+                        uint16_t dim_im_in_y,
+                        uint16_t ch_im_in);
+
+void xpulp_nn_add_u2_u2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+
+
+#endif
diff --git a/rt_nn_tests/xptnn_linear/pulp_nn_mix_kernels.h b/rt_nn_tests/xptnn_linear/pulp_nn_mix_kernels.h
new file mode 100644
index 0000000..8b2a3c6
--- /dev/null
+++ b/rt_nn_tests/xptnn_linear/pulp_nn_mix_kernels.h
@@ -0,0 +1,7093 @@
+/*
+ * pulp_nn_kernels.h
+ * Nazareno Bruschi <nazareno.bruschi@unibo.it>
+ *
+ * Copyright (C) 2019-2020 University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef __XPULPNN-MIXED_KERNELS__
+#define __XPULPNN-MIXED_KERNELS__
+
+void xpulp_nn_mix_conv_u8_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u8_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u8_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u8_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u4_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u4_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u4_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u2_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u2_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u2_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u8_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u8_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u8_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u4_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u4_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u4_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u2_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u2_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u2_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u8_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u8_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u8_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u4_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u4_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u4_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u2_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u2_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u2_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_avgpool_u8_u8(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u8_i8(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i8_u8(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i8_i8(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u8_u4(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u8_i4(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i8_u4(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i8_i4(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u8_u2(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u8_i2(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i8_u2(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i8_i2(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u4_u8(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u4_i8(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i4_u8(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i4_i8(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u4_u4(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u4_i4(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i4_u4(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i4_i4(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u4_u2(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u4_i2(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i4_u2(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i4_i2(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u2_u8(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u2_i8(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i2_u8(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i2_i8(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u2_u4(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u2_i4(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i2_u4(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i2_i4(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u2_u2(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u2_i2(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i2_u2(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i2_i2(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_add_u8_u8(
+                        uint8_t * pIn1,
+                        uint8_t * pIn2,
+                        uint8_t * pOut,
+                        uint16_t out_mult1,
+                        uint16_t out_mult2,
+                        uint16_t out_shift,
+                        uint16_t dim_im_in_x,
+                        uint16_t dim_im_in_y,
+                        uint16_t ch_im_in);
+
+void xpulp_nn_add_u8_u8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u4(
+                        uint8_t * pIn1,
+                        uint8_t * pIn2,
+                        uint8_t * pOut,
+                        uint16_t out_mult1,
+                        uint16_t out_mult2,
+                        uint16_t out_shift,
+                        uint16_t dim_im_in_x,
+                        uint16_t dim_im_in_y,
+                        uint16_t ch_im_in);
+
+void xpulp_nn_add_u8_u4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u2(
+                        uint8_t * pIn1,
+                        uint8_t * pIn2,
+                        uint8_t * pOut,
+                        uint16_t out_mult1,
+                        uint16_t out_mult2,
+                        uint16_t out_shift,
+                        uint16_t dim_im_in_x,
+                        uint16_t dim_im_in_y,
+                        uint16_t ch_im_in);
+
+void xpulp_nn_add_u8_u2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u4(
+                        uint8_t * pIn1,
+                        uint8_t * pIn2,
+                        uint8_t * pOut,
+                        uint16_t out_mult1,
+                        uint16_t out_mult2,
+                        uint16_t out_shift,
+                        uint16_t dim_im_in_x,
+                        uint16_t dim_im_in_y,
+                        uint16_t ch_im_in);
+
+void xpulp_nn_add_u4_u4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u2(
+                        uint8_t * pIn1,
+                        uint8_t * pIn2,
+                        uint8_t * pOut,
+                        uint16_t out_mult1,
+                        uint16_t out_mult2,
+                        uint16_t out_shift,
+                        uint16_t dim_im_in_x,
+                        uint16_t dim_im_in_y,
+                        uint16_t ch_im_in);
+
+void xpulp_nn_add_u4_u2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u2(
+                        uint8_t * pIn1,
+                        uint8_t * pIn2,
+                        uint8_t * pOut,
+                        uint16_t out_mult1,
+                        uint16_t out_mult2,
+                        uint16_t out_shift,
+                        uint16_t dim_im_in_x,
+                        uint16_t dim_im_in_y,
+                        uint16_t ch_im_in);
+
+void xpulp_nn_add_u2_u2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+
+
+#endif
\ No newline at end of file
diff --git a/rt_nn_tests/xptnn_linear/pulp_nn_utils.h b/rt_nn_tests/xptnn_linear/pulp_nn_utils.h
new file mode 100644
index 0000000..44d2251
--- /dev/null
+++ b/rt_nn_tests/xptnn_linear/pulp_nn_utils.h
@@ -0,0 +1,2079 @@
+/*
+ * pulp_nn_utils.h
+ * Nazareno   Bruschi  <nazareno.bruschi@unibo.it>
+ * Alessandro Nadalini <alessandro.nadalini3@unibo.it>
+ * Georg Rutishauser   <georgr@iis.ee.ethz.ch>
+ *
+ * Copyright (C) 2019-2020 ETH Zurich & University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef __PULPNN_UTILS__
+#define __PULPNN_UTILS__
+
+#include <stdint.h>
+
+typedef   signed short v2s __attribute__((vector_size (4)));
+
+
+
+#define min(a,b)                                             ((a)<(b)?(a):(b))
+#define log2(x)                                              __builtin_pulp_fl1(x)
+#define PACK_INT2_SIZE(x)                                    ((x) >> 2)
+
+#define MemoryFence()                                        asm volatile("":::"memory")
+
+#define CHANS_DECOMPR(x)                                     (5*x >> 2) // equivalent to division by 0.8
+
+/* Functions for Compressed MAC */
+#define CompressedMAC(sum, ptr, config) asm volatile(                \
+    "pv.smlsdotsp.t %[shum], %[phtr], %[chonfig];"                    \
+    : [shum] "+r" (sum), [phtr] "+r" (ptr): [chonfig] "I" (config))
+
+#define CompressedMACUnsigned(sum, ptr, config) asm volatile(                   \
+    "pv.smlsdotsup.t %[shum], %[phtr], %[chonfig];" \
+    : [shum] "+r" (sum), [phtr] "+r" (ptr): [chonfig] "I" (config))
+
+#define InitNNRF(ptr, config) asm volatile(        \
+    "pv.smlsdotsp.t x0, %[phtr], %[chonfig];"       \
+    : [phtr] "+r" (ptr) : [chonfig] "I" (config))
+
+#define ThresholdCompress(res, val, thrs) asm volatile(                                                \
+    "pv.thrc %[rhes], %[vhal], %[thhrs];" : [rhes] "+r" (res) : [vhal] "r" (val), [thhrs] "r" (thrs))
+
+#define GetConfig(a_update, b_update, a_reg, b_reg) a_update << 4 | b_update << 3 | a_reg << 1 | b_reg
+
+/* Functions for threshold&compress */
+#define check_store(res, pOut)            \
+  if ((res & 0xe0000000) == 0x00000000) { \
+    *pOut = res & 0xff;                   \
+    pOut++;                               \
+    incr_val=ch_out_r; }
+
+#define check_store_4x1(res, pOut)        \
+  if ((res & 0xe0000000) == 0x00000000) { \
+    *pOut = res & 0xff;                   \
+    pOut++; }
+
+#define reset_currThr()                           \
+  if ((uint32_t *) currThr == (uint32_t *) (pThr + ch_out)) { \
+    currThr = (v2s *) pThr;                                    \
+  }
+
+#define MacLoads20(a_update, b_update, a_reg, b_reg, ptr, sum)  __builtin_pulp_mlsdotsp20_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+
+#define MacLoad20(a_update, b_update, a_reg, b_reg, ptr, sum)  __builtin_pulp_mlsdotsup20_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+
+/* Functions for compressed min/max */
+#define CompressedMax(res, in1, in2) asm volatile( \
+                                                   "pv.max.t %[res_], %[in1_], %[in2_];" : [res_] "=r" (res) : [in1_] "r" (in1), [in2_] "r" (in2))
+
+#define CompressedMin(res, in1, in2) asm volatile(                      \
+                                                   "pv.min.t %[res_], %[in1_], %[in2_];" : [res_] "=r" (res) : [in1_] "r" (in1), [in2_] "r" (in2))
+
+#define Max16(res, in1, in2) asm volatile(                      \
+                                                   "pv.max.c %[res_], %[in1_], %[in2_];" : [res_] "=r" (res) : [in1_] "r" (in1), [in2_] "r" (in2))
+
+#define Min16(res, in1, in2) asm volatile(                      \
+                                                                        "pv.min.c %[res_], %[in1_], %[in2_];" : [res_] "=r" (res) : [in1_] "r" (in1), [in2_] "r" (in2))
+
+#define thr_cmp(state, val, threshs)  __builtin_pulp_thresh_compr(state, val, threshs)
+
+typedef unsigned char  v4u __attribute__((vector_size (4)));
+typedef   signed char  v4s __attribute__((vector_size (4)));
+
+#define bitext(x,size,off)                                      __builtin_pulp_bextract(x,size,off)
+#define bitextu(x,size,off)                                     __builtin_pulp_bextractu(x,size,off)
+#ifdef __clang__
+#define bitins(dst,not_mask_imm,src,mask_imm,off)               __builtin_binsert(dst,not_mask_imm,src,mask_imm,off)
+#else
+#define bitins(dst,not_mask_imm,src,mask_imm,off)               __builtin_pulp_binsert(dst,not_mask_imm,src,mask_imm,off)
+#endif
+#define pack(x,y,z,t)                                           __builtin_pulp_pack4(x,y,z,t)
+#define max4(a,b)                                               __builtin_pulp_maxu4(a,b)
+#define maxs4(a, b)                                             __builtin_pulp_max4(a, b)
+#define max8(a, b)                                              __builtin_pulp_maxu8(a, b)
+#define maxs8(a, b)                                             __builtin_pulp_max8(a, b)
+#define max16(a, b)                                             __builtin_pulp_maxu16(a, b)
+#define maxs16(a, b)                                            __builtin_pulp_max16(a, b)
+#define maxs20(a, b)                                            __builtin_pulp_max20(a, b)
+#define max32(a,b)                                              __builtin_pulp_maxusi(a,b)
+#define maxs32(a,b)                                             __builtin_pulp_maxsi(a,b)
+#define min32(a,b)                                              __builtin_pulp_minusi(a,b)
+#define mins32(a,b)                                             __builtin_pulp_minsi(a,b)
+#define min4(a, b)                                              __builtin_pulp_minu4(a, b)
+#define mins4(a, b)                                             __builtin_pulp_min4(a, b)
+#define min8(a, b)                                              __builtin_pulp_minu8(a, b)
+#define mins8(a, b)                                             __builtin_pulp_min8(a, b)
+#define min16(a, b)                                             __builtin_pulp_minu16(a, b)
+#define mins16(a, b)                                            __builtin_pulp_min16(a, b)
+#define mins20(a, b)                                            __builtin_pulp_min20(a, b)
+#define avg4(a,b)                                               __builtin_pulp_avgu4(a,b)
+#define avg8(a,b)                                               __builtin_pulp_avgu8(a,b)
+#define avg16(a,b)                                              __builtin_pulp_avgu16(a,b)
+#define log2(x)                                                 __builtin_pulp_fl1(x)
+#define min(a,b)                                                ((a)<(b)?(a):(b))
+#define SumDotp4(a, b, c)                                       __builtin_pulp_sdotusp4(a, b, c)
+#define SumDotp8(a, b, c)                                       __builtin_pulp_sdotusp8(a, b, c)
+#define SumDotp16(a, b, c)                                      __builtin_pulp_sdotusp16(a, b, c)
+#define SumDotps4(a, b, c)                                      __builtin_pulp_sdotsp4(a, b, c)
+#define SumDotps8(a, b, c)                                      __builtin_pulp_sdotsp8(a, b, c)
+#define SumDotps16(a, b, c)                                     __builtin_pulp_sdotsp16(a, b, c)
+#define clip4(x)                                                __builtin_pulp_clipu_r(x, 15)
+#define clip2(x)                                                __builtin_pulp_clipu_r(x, 3)
+#define clip8(x)                                                __builtin_pulp_clipu_r(x, 255)
+
+#define clips4(x)                                               __builtin_pulp_clip_r(x, 7)
+#define clips2(x)                                               __builtin_pulp_clip_r(x, 1)
+#define clips8(x)                                               __builtin_pulp_clip_r(x, 127)
+#define MacLoadInit(a_update, b_update, a_reg, b_reg, ptr)      __builtin_pulp_mlinitspr_v3(a_update, b_update, a_reg, b_reg, ptr)
+#define MacLoadUpdate(ptr)                                      __builtin_pulp_mlupdatespr_v3(ptr)
+#define MacLoadAssign(ptr)                                      __builtin_pulp_mlassignspr_v3(ptr)
+#define MacLoad4(a_update, b_update, a_reg, b_reg, ptr, sum)    __builtin_pulp_mlsdotsup4_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+#define MacLoad8(a_update, b_update, a_reg, b_reg, ptr, sum)    __builtin_pulp_mlsdotsup8_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+#define MacLoad16(a_update, b_update, a_reg, b_reg, ptr, sum)   __builtin_pulp_mlsdotsup16_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+#define MacLoads4(a_update, b_update, a_reg, b_reg, ptr, sum)   __builtin_pulp_mlsdotsp4_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+#define MacLoads8(a_update, b_update, a_reg, b_reg, ptr, sum)   __builtin_pulp_mlsdotsp8_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+#define MacLoads16(a_update, b_update, a_reg, b_reg, ptr, sum)  __builtin_pulp_mlsdotsp16_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+
+#define PACK_INT8_SIZE(x)                                    (x)
+#define PACK_INT4_SIZE(x)                                    ((x) >> 1)
+#define PACK_INT2_SIZE(x)                                    ((x) >> 2)
+
+#define MemoryFence()                                        asm volatile("":::"memory")
+
+#define LEGACY_MODE(x)                                       asm volatile ("csrwi 0x010," x)
+#define IVEC_FMT(x)                                          asm volatile ("csrwi 0x00D," x)
+#define MIXED_SKIP(x)                                        asm volatile ("csrwi 0x00F," x)
+#define A_ADDRESS(x)                                         asm volatile ("csrw 0x100, %0" :: "r" (x))
+#define W_ADDRESS(x)                                         asm volatile ("csrw 0x101, %0" :: "r" (x))
+#define A_STRIDE(x)                                          asm volatile ("csrw 0x102, %0":: "r" (x))
+#define W_STRIDE(x)                                          asm volatile ("csrw 0x103, %0":: "r" (x))
+#define A_ROLLBACK(x)                                        asm volatile ("csrw 0x104, %0":: "r" (x))
+#define W_ROLLBACK(x)                                        asm volatile ("csrw 0x105, %0":: "r" (x))
+#define A_SKIP(x)                                            asm volatile ("csrwi 0x106," x)
+#define W_SKIP(x)                                            asm volatile ("csrwi 0x107," x)
+
+static uint8_t __attribute__((noinline)) pulp_nn_quant_u2 (
+  int32_t phi,
+  int16_t m,
+  int8_t d
+  ) {
+  int32_t x = (m * phi) >> d;
+  uint8_t res = clip2(x);
+  return res;
+}
+static uint8_t __attribute__((noinline)) pulp_nn_bn_quant_u2 (
+  int32_t phi,
+  int32_t k,
+  int32_t lambda,
+  int8_t d
+  ) {
+  int32_t integer_image_phi = (k * phi) + lambda;
+  int32_t x = (integer_image_phi) >> d;
+  uint8_t res = clip2(x);
+  return res;
+}
+static uint8_t __attribute__((noinline)) pulp_nn_add_quant_u2 (
+  uint8_t pix1,
+  uint8_t pix2,
+  int16_t m1,
+  int16_t m2,
+  int8_t d
+  ) {
+  uint32_t integer_image = pix1*m1 + pix2*m2;
+  uint32_t x = (integer_image) >> d;
+  uint8_t res = clip2(x);
+  return res;
+}
+
+static int8_t __attribute__((noinline)) pulp_nn_quant_i2 (
+  int32_t phi,
+  int16_t m,
+  int8_t d
+  ) {
+  int32_t x = (m * phi) >> d;
+  int8_t res = clips2(x);
+  return res;
+}
+static int8_t __attribute__((noinline)) pulp_nn_bn_quant_i2 (
+  int32_t phi,
+  int32_t k,
+  int32_t lambda,
+  int8_t d
+  ) {
+  int32_t integer_image_phi = (k * phi) + lambda;
+  int32_t x = (integer_image_phi) >> d;
+  int8_t res = clips2(x);
+  return res;
+}
+static int8_t __attribute__((noinline)) pulp_nn_add_quant_i2 (
+  int8_t pix1,
+  int8_t pix2,
+  int16_t m1,
+  int16_t m2,
+  int8_t d
+  ) {
+  int32_t integer_image = pix1*m1 + pix2*m2;
+  int32_t x = (integer_image) >> d;
+  int8_t res = clips2(x);
+  return res;
+}
+
+static uint8_t __attribute__((noinline)) pulp_nn_quant_u4 (
+  int32_t phi,
+  int16_t m,
+  int8_t d
+  ) {
+  int32_t x = (m * phi) >> d;
+  uint8_t res = clip4(x);
+  return res;
+}
+static uint8_t __attribute__((noinline)) pulp_nn_bn_quant_u4 (
+  int32_t phi,
+  int32_t k,
+  int32_t lambda,
+  int8_t d
+  ) {
+  int32_t integer_image_phi = (k * phi) + lambda;
+  int32_t x = (integer_image_phi) >> d;
+  uint8_t res = clip4(x);
+  return res;
+}
+static uint8_t __attribute__((noinline)) pulp_nn_add_quant_u4 (
+  uint8_t pix1,
+  uint8_t pix2,
+  int16_t m1,
+  int16_t m2,
+  int8_t d
+  ) {
+  uint32_t integer_image = pix1*m1 + pix2*m2;
+  uint32_t x = (integer_image) >> d;
+  uint8_t res = clip4(x);
+  return res;
+}
+
+static int8_t __attribute__((noinline)) pulp_nn_quant_i4 (
+  int32_t phi,
+  int16_t m,
+  int8_t d
+  ) {
+  int32_t x = (m * phi) >> d;
+  int8_t res = clips4(x);
+  return res;
+}
+static int8_t __attribute__((noinline)) pulp_nn_bn_quant_i4 (
+  int32_t phi,
+  int32_t k,
+  int32_t lambda,
+  int8_t d
+  ) {
+  int32_t integer_image_phi = (k * phi) + lambda;
+  int32_t x = (integer_image_phi) >> d;
+  int8_t res = clips4(x);
+  return res;
+}
+static int8_t __attribute__((noinline)) pulp_nn_add_quant_i4 (
+  int8_t pix1,
+  int8_t pix2,
+  int16_t m1,
+  int16_t m2,
+  int8_t d
+  ) {
+  int32_t integer_image = pix1*m1 + pix2*m2;
+  int32_t x = (integer_image) >> d;
+  int8_t res = clips4(x);
+  return res;
+}
+
+static uint8_t __attribute__((noinline)) pulp_nn_quant_u8 (
+  int32_t phi,
+  int16_t m,
+  int8_t d
+  ) {
+  int32_t x = (m * phi) >> d;
+  uint8_t res = clip8(x);
+  return res;
+}
+static uint8_t __attribute__((noinline)) pulp_nn_bn_quant_u8 (
+  int32_t phi,
+  int32_t k,
+  int32_t lambda,
+  int8_t d
+  ) {
+  int32_t integer_image_phi = (k * phi) + lambda;
+  int32_t x = (integer_image_phi) >> d;
+  uint8_t res = clip8(x);
+  return res;
+}
+static uint8_t __attribute__((noinline)) pulp_nn_add_quant_u8 (
+  uint8_t pix1,
+  uint8_t pix2,
+  int16_t m1,
+  int16_t m2,
+  int8_t d
+  ) {
+  uint32_t integer_image = pix1*m1 + pix2*m2;
+  uint32_t x = (integer_image) >> d;
+  uint8_t res = clip8(x);
+  return res;
+}
+
+static int8_t __attribute__((noinline)) pulp_nn_quant_i8 (
+  int32_t phi,
+  int16_t m,
+  int8_t d
+  ) {
+  int32_t x = (m * phi) >> d;
+  int8_t res = clips8(x);
+  return res;
+}
+static int8_t __attribute__((noinline)) pulp_nn_bn_quant_i8 (
+  int32_t phi,
+  int32_t k,
+  int32_t lambda,
+  int8_t d
+  ) {
+  int32_t integer_image_phi = (k * phi) + lambda;
+  int32_t x = (integer_image_phi) >> d;
+  int8_t res = clips8(x);
+  return res;
+}
+static int8_t __attribute__((noinline)) pulp_nn_add_quant_i8 (
+  int8_t pix1,
+  int8_t pix2,
+  int16_t m1,
+  int16_t m2,
+  int8_t d
+  ) {
+  int32_t integer_image = pix1*m1 + pix2*m2;
+  int32_t x = (integer_image) >> d;
+  int8_t res = clips8(x);
+  return res;
+}
+
+
+static uint8_t __attribute__((noinline)) pulp_nn_u4_quant(int input, int16_t * pThr)
+{
+  if(input <= pThr[7] )
+  {
+    if( input <= pThr[3])
+    {
+      if( input <= pThr[1])
+      {
+        if( input <= pThr[0])
+          return 0;
+        else
+          return 1;
+      }
+      else
+      {
+        if( input <= pThr[2])
+          return 2;
+        else
+          return 3;
+      }
+    }
+    else
+    {
+      if( input <= pThr[5])
+      {
+        if( input <= pThr[4])
+          return 4;
+        else
+          return 5;
+      }
+      else
+      {
+        if( input <= pThr[6])
+          return 6;
+        else
+          return 7;
+      }
+    }
+  }
+  else
+  {
+    if( input <= pThr[11])
+    {
+      if( input <= pThr[9])
+      {
+        if( input <= pThr[8])
+          return 8;
+        else
+          return 9;
+      }
+      else
+      {
+        if( input <= pThr[10])
+          return 10;
+        else
+          return 11;
+      }
+    }
+    else
+    {
+      if( input <= pThr[13])
+      {
+        if( input <= pThr[12])
+          return 12;
+        else
+          return 13;
+      }
+      else
+      {
+        if( input <= pThr[14])
+          return 14;
+        else
+          return 15;
+      }
+    }
+  }
+}
+
+static uint8_t __attribute__((noinline)) pulp_nn_u2_quant(int input, int16_t * pThr)
+{
+  if( input <= pThr[1])
+  {
+    if( input <= pThr[0])
+        {
+      return 0;
+    }
+        else
+        {
+      return 1;
+    }
+  }
+    else
+    {
+    if( input <= pThr[2])
+        {
+      return 2;
+    }
+        else
+        {
+      return 3;
+    }
+  }
+}
+
+/*
+ * Common
+ */
+
+
+static v4s __attribute__((noinline)) pulp_nn_i4_to_i8_r( int8_t *pSrc)
+{
+  v4s Src = *((v4s*) pSrc);
+  int8_t bext1, bext2, bext3, bext4;
+
+  bext1 = (int8_t) bitext((int) Src, 4, 0);
+  bext2 = (int8_t) bitext((int) Src, 4, 4);
+  bext3 = (int8_t) bitext((int) Src, 4, 8);
+  bext4 = (int8_t) bitext((int) Src, 4, 12);
+  v4s res = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+
+  return res;
+}
+
+static v4u __attribute__((noinline)) pulp_nn_u4_to_u8_r(uint8_t *pSrc)
+{
+  v4u Src = *((v4u*) pSrc);
+  uint8_t bext1, bext2, bext3, bext4;
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 4, 0);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 4, 4);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 4, 8);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 4, 12);
+  v4u res = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+
+  return res;
+}
+
+static v4s __attribute__((noinline)) pulp_nn_i2_to_i8_r( int8_t *pSrc)
+{
+  v4s Src = *((v4s*) pSrc);
+  int8_t bext1, bext2, bext3, bext4;
+
+  bext1 = (int8_t) bitext((int) Src, 2, 0);
+  bext2 = (int8_t) bitext((int) Src, 2, 2);
+  bext3 = (int8_t) bitext((int) Src, 2, 4);
+  bext4 = (int8_t) bitext((int) Src, 2, 6);
+  v4s res = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+
+  return res;
+}
+
+static v4u __attribute__((noinline)) pulp_nn_u2_to_u8_r(uint8_t *pSrc)
+{
+  v4u Src = *((v4u*) pSrc);
+  uint8_t bext1, bext2, bext3, bext4;
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 0);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 2);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 4);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 6);
+  v4u res = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+
+  return res;
+}
+
+static v4s __attribute__((noinline)) pulp_nn_i2_to_i4_r(int8_t *pSrc)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  v4s Src = *((v4s*) pSrc);
+  int8_t bext1, bext2, bext3, bext4;
+  int8_t out1, out2, out3, out4;
+
+  bext1 = (int8_t) bitextu((int) Src, 2, 0);
+  bext2 = (int8_t) bitextu((int) Src, 2, 2);
+  bext3 = (int8_t) bitextu((int) Src, 2, 4);
+  bext4 = (int8_t) bitextu((int) Src, 2, 6);
+
+  out1 = bitins(bext1, n_mask, bext2, mask, off);
+  out2 = bitins(bext3, n_mask, bext4, mask, off);
+
+  bext1 = (int8_t) bitextu((int) Src, 2, 8);
+  bext2 = (int8_t) bitextu((int) Src, 2, 10);
+  bext3 = (int8_t) bitextu((int) Src, 2, 12);
+  bext4 = (int8_t) bitextu((int) Src, 2, 14);
+
+  out3 = bitins(bext1, n_mask, bext2, mask, off);
+  out4 = bitins(bext3, n_mask, bext4, mask, off);
+
+  v4s res = pack((int8_t) out1, (int8_t) out2, (int8_t) out3, (int8_t) out4);
+
+  return res;
+}
+
+static v4u __attribute__((noinline)) pulp_nn_u2_to_u4_r(uint8_t *pSrc)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  v4u Src = *((v4u*) pSrc);
+  uint8_t bext1, bext2, bext3, bext4;
+  uint8_t out1, out2, out3, out4;
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 0);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 2);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 4);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 6);
+
+  out1 = bitins(bext1, n_mask, bext2, mask, off);
+  out2 = bitins(bext3, n_mask, bext4, mask, off);
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 8);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 10);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 12);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 14);
+
+  out3 = bitins(bext1, n_mask, bext2, mask, off);
+  out4 = bitins(bext3, n_mask, bext4, mask, off);
+
+  v4u res = pack((uint8_t) out1, (uint8_t) out2, (uint8_t) out3, (uint8_t) out4);
+
+  return res;
+}
+
+static int8_t *__attribute__((always_inline)) pulp_nn_i4_to_i8( int8_t *pSrc, int8_t *pDst)
+{
+  v4s Src = *((v4s*) pSrc);
+  int8_t bext1, bext2, bext3, bext4;
+
+  pSrc+=4;
+
+  bext1 = (int8_t) bitext((int) Src, 4, 0);
+  bext2 = (int8_t) bitext((int) Src, 4, 4);
+  bext3 = (int8_t) bitext((int) Src, 4, 8);
+  bext4 = (int8_t) bitext((int) Src, 4, 12);
+  *((v4s*)pDst) = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+   MemoryFence();
+  pDst+=4;
+  bext1 = (int8_t) bitext((int) Src, 4, 16);
+  bext2 = (int8_t) bitext((int) Src, 4, 20);
+  bext3 = (int8_t) bitext((int) Src, 4, 24);
+  bext4 = (int8_t) bitext((int) Src, 4, 28);
+  *((v4s*)pDst) = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+
+  return pSrc;
+}
+
+static uint8_t *__attribute__((always_inline)) pulp_nn_u4_to_u8(uint8_t *pSrc, uint8_t *pDst)
+{
+  v4u Src = *((v4u*) pSrc);
+  uint8_t bext1, bext2, bext3, bext4;
+
+  pSrc+=4;
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 4, 0);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 4, 4);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 4, 8);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 4, 12);
+  *((v4u*)pDst) = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 4, 16);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 4, 20);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 4, 24);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 4, 28);
+  *((v4u*)pDst) = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+
+  return pSrc;
+}
+
+static int8_t *__attribute__((always_inline)) pulp_nn_i2_to_i8( int8_t * pSrc, int8_t * pDst)
+{
+  v4s Src = *((v4s*) pSrc);
+  int8_t bext1, bext2, bext3, bext4;
+
+  pSrc+=4;
+
+  bext1 = (int8_t) bitext((int) Src, 2, 0);
+  bext2 = (int8_t) bitext((int) Src, 2, 2);
+  bext3 = (int8_t) bitext((int) Src, 2, 4);
+  bext4 = (int8_t) bitext((int) Src, 2, 6);
+  *((v4s*)pDst) = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (int8_t) bitext((int) Src, 2, 8);
+  bext2 = (int8_t) bitext((int) Src, 2, 10);
+  bext3 = (int8_t) bitext((int) Src, 2, 12);
+  bext4 = (int8_t) bitext((int) Src, 2, 14);
+  *((v4s*)pDst) = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (int8_t) bitext((int) Src, 2, 16);
+  bext2 = (int8_t) bitext((int) Src, 2, 18);
+  bext3 = (int8_t) bitext((int) Src, 2, 20);
+  bext4 = (int8_t) bitext((int) Src, 2, 22);
+  *((v4s*)pDst) = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (int8_t) bitext((int) Src, 2, 24);
+  bext2 = (int8_t) bitext((int) Src, 2, 26);
+  bext3 = (int8_t) bitext((int) Src, 2, 28);
+  bext4 = (int8_t) bitext((int) Src, 2, 30);
+  *((v4s*)pDst) = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+
+  return pSrc;
+}
+
+static uint8_t *__attribute__((always_inline)) pulp_nn_u2_to_u8(uint8_t * pSrc, uint8_t * pDst)
+{
+  v4u Src = *((v4u*) pSrc);
+  uint8_t bext1, bext2, bext3, bext4;
+
+  pSrc+=4;
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 0);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 2);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 4);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 6);
+  *((v4u*)pDst) = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 8);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 10);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 12);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 14);
+  *((v4u*)pDst) = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 16);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 18);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 20);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 22);
+  *((v4u*)pDst) = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 24);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 26);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 28);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 30);
+  *((v4u*)pDst) = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+
+  return pSrc;
+}
+
+static int8_t *__attribute__((always_inline)) pulp_nn_i2_to_i4( int8_t * pSrc, int8_t * pDst)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  v4s Src = *((v4s*) pSrc);
+  int8_t bext1, bext2, bext3, bext4;
+  int8_t out1, out2, out3, out4;
+
+  pSrc+=4;
+
+  bext1 = (int8_t) bitext((int) Src, 2, 0);
+  bext2 = (int8_t) bitext((int) Src, 2, 2);
+  bext3 = (int8_t) bitext((int) Src, 2, 4);
+  bext4 = (int8_t) bitext((int) Src, 2, 6);
+
+  out1 = bitins(bext1, n_mask, bext2, mask, off);
+  out2 = bitins(bext3, n_mask, bext4, mask, off);
+  MemoryFence();
+
+  bext1 = (int8_t) bitext((int) Src, 2, 8);
+  bext2 = (int8_t) bitext((int) Src, 2, 10);
+  bext3 = (int8_t) bitext((int) Src, 2, 12);
+  bext4 = (int8_t) bitext((int) Src, 2, 14);
+
+  out3 = bitins(bext1, n_mask, bext2, mask, off);
+  out4 = bitins(bext3, n_mask, bext4, mask, off);
+
+  *((v4s*)pDst) = pack((int8_t) out1, (int8_t) out2, (int8_t) out3, (int8_t) out4);
+  MemoryFence();
+
+  pDst+=4;
+  bext1 = (int8_t) bitext((int) Src, 2, 16);
+  bext2 = (int8_t) bitext((int) Src, 2, 18);
+  bext3 = (int8_t) bitext((int) Src, 2, 20);
+  bext4 = (int8_t) bitext((int) Src, 2, 22);
+
+  out1 = bitins(bext1, n_mask, bext2, mask, off);
+  out2 = bitins(bext3, n_mask, bext4, mask, off);
+  MemoryFence();
+
+  bext1 = (int8_t) bitext((int) Src, 2, 24);
+  bext2 = (int8_t) bitext((int) Src, 2, 26);
+  bext3 = (int8_t) bitext((int) Src, 2, 28);
+  bext4 = (int8_t) bitext((int) Src, 2, 30);
+
+  out3 = bitins(bext1, n_mask, bext2, mask, off);
+  out4 = bitins(bext3, n_mask, bext4, mask, off);
+
+  *((v4s*)pDst) = pack((int8_t) out1, (int8_t) out2, (int8_t) out3, (int8_t) out4);
+
+  return pSrc;
+}
+
+static uint8_t *__attribute__((always_inline)) pulp_nn_u2_to_u4( uint8_t * pSrc, uint8_t * pDst)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  v4u Src = *((v4u*) pSrc);
+  uint8_t bext1, bext2, bext3, bext4;
+  uint8_t out1, out2, out3, out4;
+
+  pSrc+=4;
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 0);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 2);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 4);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 6);
+
+  out1 = bitins(bext1, n_mask, bext2, mask, off);
+  out2 = bitins(bext3, n_mask, bext4, mask, off);
+  MemoryFence();
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 8);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 10);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 12);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 14);
+
+  out3 = bitins(bext1, n_mask, bext2, mask, off);
+  out4 = bitins(bext3, n_mask, bext4, mask, off);
+
+  *((v4u*)pDst) = pack((uint8_t) out1, (uint8_t) out2, (uint8_t) out3, (uint8_t) out4);
+  MemoryFence();
+
+  pDst+=4;
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 16);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 18);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 20);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 22);
+
+  out1 = bitins(bext1, n_mask, bext2, mask, off);
+  out2 = bitins(bext3, n_mask, bext4, mask, off);
+  MemoryFence();
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 24);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 26);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 28);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 30);
+
+  out3 = bitins(bext1, n_mask, bext2, mask, off);
+  out4 = bitins(bext3, n_mask, bext4, mask, off);
+  
+  *((v4u*)pDst) = pack((uint8_t) out1, (uint8_t) out2, (uint8_t) out3, (uint8_t) out4);
+
+  return pSrc;
+}
+
+/*
+ * XpulpV2
+ */
+
+static void __attribute__((noinline)) pulp_zero_mem(uint8_t * pBuffer, unsigned int size)
+{
+  int lfover = size &0x3;
+  for (int i=0; i<(size>>2); i++)
+  {
+    *((v4u *)pBuffer) = (v4u){0,0,0,0};
+    MemoryFence();
+    pBuffer+=4;
+  }
+  while(lfover)
+  {
+    *pBuffer++=0;
+    lfover--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_im2col_u2_to_u8(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_u2_to_u8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 16;
+  }
+  while (lfover)
+  {
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 0);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 2);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 4);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 6);
+    pOutput++;
+    pInput++;
+    lfover -= 4;
+  }
+}
+static void __attribute__((noinline)) pulp_nn_im2col_i2_to_i8(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_i2_to_i8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 16;
+  }
+  while (lfover)
+  {
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 0);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 2);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 4);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 6);
+    pOutput++;
+    pInput++;
+    lfover -= 4;
+  }
+}
+static void __attribute__((noinline)) pulp_nn_im2col_u4_to_u8(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 3u;
+  int lfover = blockSize & 0x07;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_u4_to_u8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 8;
+  }
+  while (lfover)
+  {
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 4, 0);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 4, 4);
+    pOutput++;
+    pInput++;
+    lfover -= 2;
+  }
+}
+static void __attribute__((noinline)) pulp_nn_im2col_i4_to_i8(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 3u;
+  int lfover = blockSize & 0x07;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_i4_to_i8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 8;
+  }
+  while (lfover)
+  {
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 4, 0);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 4, 4);
+    pOutput++;
+    pInput++;
+    lfover -= 2;
+  }
+}
+static void __attribute__((noinline)) pulp_nn_im2col_u8_to_u8(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 2u;
+  int lfover = blockSize & 0x03;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4u*)pOutput) = *((v4u*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    *((uint8_t*)pOutput) = *((uint8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover--;
+  }
+}
+static void __attribute__((noinline)) pulp_nn_im2col_i8_to_i8(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 2u;
+  int lfover = blockSize & 0x03;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4s*)pOutput) = *((v4s*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    *((int8_t*)pOutput) = *((int8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover--;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_im2col_u2_to_u2(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4u*)pOutput) = *((v4u*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    uint8_t extr;
+    *((uint8_t*)pOutput) = *((uint8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover -= 4;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_i2_to_i2(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4s*)pOutput) = *((v4s*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    int8_t extr;
+    *((int8_t*)pOutput) = *((int8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover -= 4;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_u2_to_u4(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_u2_to_u4(pInput, pOutput);
+    MemoryFence();
+    pOutput += 8;
+  }
+  while (lfover)
+  {
+    uint8_t extr;
+    // TODO: this is potentially dangerous/wrong if the number of channels
+    // is not a multiple of 8!
+    *((v4u *)pOutput) = pulp_nn_u2_to_u4_r(pInput);
+    pInput += 2;
+    pOutput += 4;
+    lfover -= 8;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_i2_to_i4(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_i2_to_i4(pInput, pOutput);
+    MemoryFence();
+    pOutput += 8;
+  }
+  while (lfover)
+  {
+    int8_t extr;
+    // TODO: this is potentially dangerous/wrong if the number of channels
+    // is not a multiple of 8!
+    *((v4s *)pOutput) = pulp_nn_i2_to_i4_r(pInput);
+    pInput += 2;
+    pOutput += 4;
+    lfover -= 8;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_u2_to_u8(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_u2_to_u8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 16;
+  }
+  while (lfover)
+  {
+    uint8_t extr;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 0);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 2);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 4);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 6);
+    pOutput++;
+    pInput++;
+    lfover -= 4;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_i2_to_i8(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_i2_to_i8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 16;
+  }
+  while (lfover)
+  {
+    int8_t extr;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 0);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 2);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 4);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 6);
+    pOutput++;
+    pInput++;
+    lfover -= 4;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_u4_to_u4(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 3u;
+  int lfover = blockSize & 0x07;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4u*)pOutput) = *((v4u*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    uint8_t extr;
+    *((uint8_t*)pOutput) = *((uint8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover -= 2;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_i4_to_i4(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 3u;
+  int lfover = blockSize & 0x07;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4s*)pOutput) = *((v4s*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    int8_t extr;
+    *((int8_t*)pOutput) = *((int8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover -= 2;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_u4_to_u8(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 3u;
+  int lfover = blockSize & 0x07;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_u4_to_u8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 8;
+  }
+  while (lfover)
+  {
+    uint8_t extr;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 4, 0);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 4, 4);
+    pOutput++;
+    pInput++;
+    lfover -= 2;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_i4_to_i8(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 3u;
+  int lfover = blockSize & 0x07;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_i4_to_i8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 8;
+  }
+  while (lfover)
+  {
+    int8_t extr;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 4, 0);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 4, 4);
+    pOutput++;
+    pInput++;
+    lfover -= 2;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_u8_to_u8(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 2u;
+  int lfover = blockSize & 0x03;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4u*)pOutput) = *((v4u*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    uint8_t extr;
+    *((uint8_t*)pOutput) = *((uint8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover -= 1;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_i8_to_i8(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 2u;
+  int lfover = blockSize & 0x03;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4s*)pOutput) = *((v4s*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    int8_t extr;
+    *((int8_t*)pOutput) = *((int8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover -= 1;
+  }
+}
+
+
+
+static void __attribute__((noinline)) pulp_nn_compare_and_replace_if_larger_u8(uint8_t * base,
+                                                uint8_t * target,
+                                                uint16_t length)
+{
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+  v4u inp;
+  v4u com;
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    inp = *((v4u*)pIn);
+    com = *((v4u*)pCom);
+
+    *((v4u*)pIn) = max4(inp, com);
+
+    pCom+=4;
+    pIn+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    if(*pIn<*pCom)
+      *pIn=*pCom;
+    
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_compare_and_replace_if_larger_i8(
+    int8_t * base, int8_t * target, uint16_t length) {
+  int8_t *pIn = base;
+  int8_t *pCom = target;
+  v4s inp;
+  v4s com;
+  int cnt = length >> 2;
+
+  while (cnt > 0u) {
+    inp = *((v4s *)pIn);
+    com = *((v4s *)pCom);
+
+    *((v4s *)pIn) = maxs4(inp, com);
+
+    pCom += 4;
+    pIn += 4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left > 0u) {
+    if (*pIn < *pCom)
+      *pIn = *pCom;
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_avg_and_replace_u8(uint8_t * base,
+                                  uint8_t * target,
+                                  uint16_t length)
+{
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+
+  while (length>0u)
+  {
+    *pIn = ((*pIn + *pCom) >> 1);
+
+    pIn++;
+    pCom++;
+    length--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_compare_and_replace_if_larger_u4(uint8_t * base,
+                                                uint8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+  v4u inp[2];
+  v4u com[2];
+  uint8_t *out;
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    pulp_nn_u4_to_u8(pIn, (uint8_t *)inp);
+    pulp_nn_u4_to_u8(pCom, (uint8_t *)com);
+
+    *((v4u *)out) = max4(inp[0], com[0]);
+
+    *((uint8_t*)pIn) = bitins(*out, n_mask, *(out + 1), mask, off);
+    pIn++;
+    *((uint8_t*)pIn) = bitins(*(out + 2), n_mask, *(out + 3), mask, off);
+    pIn++;
+
+    *((v4u *)out) = max4(inp[1], com[1]);
+
+    *((uint8_t*)pIn) = bitins(*out, n_mask, *(out + 1), mask, off);
+    pIn++;
+    *((uint8_t*)pIn) = bitins(*(out + 2), n_mask, *(out + 3), mask, off);
+    pIn++;
+
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 4, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 4, 4);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 4, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 4, 4);
+
+    if(inA0<inB0)
+      inA0=inB0;
+
+    if(inA1<inB1)
+      inA1=inB1;
+
+    *((uint8_t*)pIn) = bitins(inA0, n_mask, inA1, mask, off);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_compare_and_replace_if_larger_i4(int8_t * base,
+                                                int8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  int8_t *pIn = base;
+  int8_t *pCom = target;
+  v4s inp[2];
+  v4s com[2];
+  int8_t *out;
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    pulp_nn_i4_to_i8(pIn, (int8_t *)inp);
+    pulp_nn_i4_to_i8(pCom, (int8_t *)com);
+
+    *((v4s *)out) = maxs4(inp[0], com[0]);
+
+    *((int8_t*)pIn) = bitins(*out, n_mask, *(out + 1), mask, off);
+    pIn++;
+    *((int8_t*)pIn) = bitins(*(out + 2), n_mask, *(out + 3), mask, off);
+    pIn++;
+
+    *((v4s *)out) = maxs4(inp[1], com[1]);
+
+    *((int8_t*)pIn) = bitins(*out, n_mask, *(out + 1), mask, off);
+    pIn++;
+    *((int8_t*)pIn) = bitins(*(out + 2), n_mask, *(out + 3), mask, off);
+    pIn++;
+
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    int8_t inA0 = (int8_t) bitext((int) *pIn, 4, 0);
+    int8_t inA1 = (int8_t) bitext((int) *pIn, 4, 4);
+    int8_t inB0 = (int8_t) bitext((int) *pCom, 4, 0);
+    int8_t inB1 = (int8_t) bitext((int) *pCom, 4, 4);
+
+    if(inA0<inB0)
+      inA0=inB0;
+
+    if(inA1<inB1)
+      inA1=inB1;
+
+    *((int8_t*)pIn) = bitins(inA0, n_mask, inA1, mask, off);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_avg_and_replace_u4(uint8_t * base,
+                                  uint8_t * target,
+                                  uint16_t length)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+
+  while (length>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 4, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 4, 4);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 4, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 4, 4);
+
+    inA0 = ((inA0 + inB0) >> 1);
+    inA1 = ((inA1 + inB1) >> 1);
+
+    *((uint8_t*)pIn) = bitins(inA0, n_mask, inA1, mask, off);
+
+    pIn++;
+    pCom++;
+    length--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_compare_and_replace_if_larger_u2(uint8_t * base,
+                                                uint8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask2 = 0x0c;
+  int8_t n_mask2 = ~ mask2;
+  int8_t mask4 = 0x30;
+  int8_t n_mask4 = ~ mask4;
+  int8_t mask6 = 0xc0;
+  int8_t n_mask6 = ~ mask6;
+  int8_t off2 = 2;
+  int8_t off4 = 4;
+  int8_t off6 = 6;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+  v4u inp[4];
+  v4u com[4];
+  uint8_t *out;
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    pulp_nn_u2_to_u8(pIn, inp);
+    pulp_nn_u2_to_u8(pCom, com);
+
+    *((v4u*)out) = max4(inp[0], com[0]);
+
+    uint8_t inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    *((v4u*)out) = max4(inp[1], com[1]);
+
+    inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    *((v4u*)out) = max4(inp[2], com[2]);
+
+    inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    *((v4u*)out) = max4(inp[3], com[3]);
+
+    inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 2, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 2, 2);
+    uint8_t inA2 = (uint8_t) bitextu((unsigned int) *pIn, 2, 4);
+    uint8_t inA3 = (uint8_t) bitextu((unsigned int) *pIn, 2, 6);
+    v4u inA4 = pack((uint8_t) inA0, (uint8_t) inA1, (uint8_t) inA2, (uint8_t) inA3);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 2, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 2, 2);
+    uint8_t inB2 = (uint8_t) bitextu((unsigned int) *pCom, 2, 4);
+    uint8_t inB3 = (uint8_t) bitextu((unsigned int) *pCom, 2, 6);
+    v4u inB4 = pack((uint8_t) inB0, (uint8_t) inB1, (uint8_t) inB2, (uint8_t) inB3);
+
+    *((v4u*)out) = max4(inA4, inB4);
+
+    uint8_t inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+
+static void __attribute__((noinline)) pulp_nn_compare_and_replace_if_larger_i2(int8_t * base,
+                                                int8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask2 = 0x0c;
+  int8_t n_mask2 = ~ mask2;
+  int8_t mask4 = 0x30;
+  int8_t n_mask4 = ~ mask4;
+  int8_t mask6 = 0xc0;
+  int8_t n_mask6 = ~ mask6;
+  int8_t off2 = 2;
+  int8_t off4 = 4;
+  int8_t off6 = 6;
+
+  int8_t *pIn = base;
+  int8_t *pCom = target;
+  v4s inp[4];
+  v4s com[4];
+  int8_t *out;
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    pulp_nn_i2_to_i8(pIn, inp);
+    pulp_nn_i2_to_i8(pCom, com);
+
+    *((v4s*)out) = maxs4(inp[0], com[0]);
+
+    int8_t inA = (int8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((int8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    *((v4s*)out) = maxs4(inp[1], com[1]);
+
+    inA = (int8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((int8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    *((v4s*)out) = maxs4(inp[2], com[2]);
+
+    inA = (int8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((int8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    *((v4s*)out) = maxs4(inp[3], com[3]);
+
+    inA = (int8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((int8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    int8_t inA0 = (int8_t) bitext((unsigned int) *pIn, 2, 0);
+    int8_t inA1 = (int8_t) bitext((unsigned int) *pIn, 2, 2);
+    int8_t inA2 = (int8_t) bitext((unsigned int) *pIn, 2, 4);
+    int8_t inA3 = (int8_t) bitext((unsigned int) *pIn, 2, 6);
+    v4s inA4 = pack((int8_t) inA0, (int8_t) inA1, (int8_t) inA2, (int8_t) inA3);
+    int8_t inB0 = (int8_t) bitext((unsigned int) *pCom, 2, 0);
+    int8_t inB1 = (int8_t) bitext((unsigned int) *pCom, 2, 2);
+    int8_t inB2 = (int8_t) bitext((unsigned int) *pCom, 2, 4);
+    int8_t inB3 = (int8_t) bitext((unsigned int) *pCom, 2, 6);
+    v4s inB4 = pack((int8_t) inB0, (int8_t) inB1, (int8_t) inB2, (int8_t) inB3);
+
+    *((v4s*)out) = maxs4(inA4, inB4);
+
+    int8_t inA = (int8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((int8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_avg_and_replace_u2(uint8_t * base,
+                                  uint8_t * target,
+                                  uint16_t length)
+{
+  int8_t mask2 = 0x0c;
+  int8_t n_mask2 = ~ mask2;
+  int8_t mask4 = 0x30;
+  int8_t n_mask4 = ~ mask4;
+  int8_t mask6 = 0xc0;
+  int8_t n_mask6 = ~ mask6;
+  int8_t off2 = 2;
+  int8_t off4 = 4;
+  int8_t off6 = 6;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+
+  while (length>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 2, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 2, 2);
+    uint8_t inA2 = (uint8_t) bitextu((unsigned int) *pIn, 2, 4);
+    uint8_t inA3 = (uint8_t) bitextu((unsigned int) *pIn, 2, 6);
+
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 2, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 2, 2);
+    uint8_t inB2 = (uint8_t) bitextu((unsigned int) *pCom, 2, 4);
+    uint8_t inB3 = (uint8_t) bitextu((unsigned int) *pCom, 2, 6);
+
+    inA0 = ((inA0 + inB0) >> 1);
+    inA1 = ((inA1 + inB1) >> 1);
+    inA2 = ((inA2 + inB2) >> 1);
+    inA3 = ((inA3 + inB3) >> 1);
+
+    uint8_t inA = (uint8_t) bitins(inA0, n_mask2, inA1, mask2, off2);
+    inA = bitins(inA, n_mask4, inA2, mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, inA3, mask6, off6);
+
+    pIn++;
+    pCom++;
+    length--;
+  }
+}
+
+/*
+ * XpulpNN
+ */
+
+static void __attribute__((noinline)) xpulp_nn_zero_mem_u8(uint8_t * pBuffer, unsigned int size)
+{
+  int lfover = size &0x3;
+  for (int i=0; i<(size>>2); i++)
+  {
+    *((v4u *)pBuffer) = (v4u){0,0,0,0};
+    MemoryFence();
+    pBuffer+=4;
+  }
+  while(lfover)
+  {
+    *pBuffer++=0;
+    lfover--;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_zero_mem_u4(uint8_t * pBuffer, unsigned int size)
+{
+  int lfover = size &0x7;
+  for (int i=0; i<(size>>3); i++)
+  {
+    *((v4u *)pBuffer) = (v4u){0,0,0,0};
+    MemoryFence();
+    pBuffer+=4;
+  }
+  while(lfover)
+  {
+    *pBuffer++=0;
+    lfover-=2;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_zero_mem_u2(uint8_t * pBuffer, unsigned int size)
+{
+  int lfover = size &0xf;
+  for (int i=0; i<(size>>4); i++)
+  {
+    *((v4u *)pBuffer) = (v4u){0,0,0,0};
+    MemoryFence();
+    pBuffer+=4;
+  }
+  while(lfover)
+  {
+    *pBuffer++=0;
+    lfover-=4;
+  }
+}
+
+
+static void __attribute__((noinline)) xpulp_tnn_zero_mem_ternary(uint8_t * pBuffer, unsigned int size, unsigned int uns)
+{
+  uint8_t pad_val = 0xd9;
+  uint32_t pad_vec = 0xd9d9d9d9;
+  if (uns) {
+    // if we are using an unsigned kernel, we need to pad with -1 because the hardware will add a +1 to ALL values!
+    pad_val = 0xff;
+    pad_vec = 0xffffffff;
+  }
+  int lfover = size &0xf;
+  for (int i=0; i<(size>>4); i++)
+    {
+      *((v4u *)pBuffer) = (v4u)pad_vec;
+      MemoryFence();
+      pBuffer+=4;
+    }
+  while(lfover)
+    {
+      *pBuffer++=pad_val;
+      lfover-=4;
+    }
+}
+
+static void __attribute__((noinline)) xpulp_nn_compare_and_replace_if_larger_u4(uint8_t * base,
+                                                uint8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    *((uint32_t *)pIn) = max8(*((uint32_t *)pIn), *((uint32_t *)pCom));
+
+    pIn+=4;
+    pCom+=4;
+
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 4, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 4, 4);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 4, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 4, 4);
+
+    if(inA0<inB0)
+      inA0=inB0;
+
+    if(inA1<inB1)
+      inA1=inB1;
+
+    *((uint8_t*)pIn) = bitins(inA0, n_mask, inA1, mask, off);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_compare_and_replace_if_larger_i4(
+    int8_t * base, int8_t * target, uint16_t length) {
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~mask;
+  int8_t off = 0x04;
+
+  int8_t *pIn = base;
+  int8_t *pCom = target;
+
+  int cnt = length >> 2;
+
+  while (cnt > 0u) {
+    *((int32_t *)pIn) = maxs8(*((int32_t *)pIn), *((int32_t *)pCom));
+
+    pIn += 4;
+    pCom += 4;
+
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left > 0u) {
+    int8_t inA0 = (int8_t)bitext((int)*pIn, 4, 0);
+    int8_t inA1 = (int8_t)bitext((int)*pIn, 4, 4);
+    int8_t inB0 = (int8_t)bitext((int)*pCom, 4, 0);
+    int8_t inB1 = (int8_t)bitext((int)*pCom, 4, 4);
+
+    if (inA0 < inB0)
+      inA0 = inB0;
+
+    if (inA1 < inB1)
+      inA1 = inB1;
+
+    *((int8_t *)pIn) = bitins(inA0, n_mask, inA1, mask, off);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_avg_and_replace_u4(uint8_t * base,
+                                  uint8_t * target,
+                                  uint16_t length)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+
+  int cnt = length >> 2;
+
+  while (cnt > 0u)
+  {
+    *((uint32_t *)pIn) = avg8(*((uint32_t *)pIn), *((uint32_t *)pCom));
+
+    pIn+=4;
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+
+  while (left>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 4, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 4, 4);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 4, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 4, 4);
+
+    inA0 = ((inA0 + inB0) >> 1);
+    inA1 = ((inA1 + inB1) >> 1);
+
+    *((uint8_t*)pIn) = bitins(inA0, n_mask, inA1, mask, off);
+
+    pIn++;
+    pCom++;
+    length--;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_compare_and_replace_if_larger_u2(uint8_t * base,
+                                                uint8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask2 = 0x0c;
+  int8_t n_mask2 = ~ mask2;
+  int8_t mask4 = 0x30;
+  int8_t n_mask4 = ~ mask4;
+  int8_t mask6 = 0xc0;
+  int8_t n_mask6 = ~ mask6;
+  int8_t off2 = 2;
+  int8_t off4 = 4;
+  int8_t off6 = 6;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+  uint8_t *out;
+
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    *((uint32_t *)pIn) = max16(*((uint32_t *)pIn), *((uint32_t *)pCom));
+
+    pIn+=4;
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 2, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 2, 2);
+    uint8_t inA2 = (uint8_t) bitextu((unsigned int) *pIn, 2, 4);
+    uint8_t inA3 = (uint8_t) bitextu((unsigned int) *pIn, 2, 6);
+    v4u inA4 = pack((uint8_t) inA0, (uint8_t) inA1, (uint8_t) inA2, (uint8_t) inA3);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 2, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 2, 2);
+    uint8_t inB2 = (uint8_t) bitextu((unsigned int) *pCom, 2, 4);
+    uint8_t inB3 = (uint8_t) bitextu((unsigned int) *pCom, 2, 6);
+    v4u inB4 = pack((uint8_t) inB0, (uint8_t) inB1, (uint8_t) inB2, (uint8_t) inB3);
+
+    *((v4u*)out) = max4(inA4, inB4);
+
+    uint8_t inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_compare_and_replace_if_larger_i2(int8_t * base,
+                                                int8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask2 = 0x0c;
+  int8_t n_mask2 = ~ mask2;
+  int8_t mask4 = 0x30;
+  int8_t n_mask4 = ~ mask4;
+  int8_t mask6 = 0xc0;
+  int8_t n_mask6 = ~ mask6;
+  int8_t off2 = 2;
+  int8_t off4 = 4;
+  int8_t off6 = 6;
+
+  int8_t *pIn = base;
+  int8_t *pCom = target;
+  int8_t *out;
+
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    *((int32_t *)pIn) = maxs16(*((int32_t *)pIn), *((int32_t *)pCom));
+
+    pIn+=4;
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    int8_t inA0 = (int8_t) bitext((int) *pIn, 2, 0);
+    int8_t inA1 = (int8_t) bitext((int) *pIn, 2, 2);
+    int8_t inA2 = (int8_t) bitext((int) *pIn, 2, 4);
+    int8_t inA3 = (int8_t) bitext((int) *pIn, 2, 6);
+    v4s inA4 = pack((int8_t) inA0, (int8_t) inA1, (int8_t) inA2, (int8_t) inA3);
+    int8_t inB0 = (int8_t) bitext((int) *pCom, 2, 0);
+    int8_t inB1 = (int8_t) bitext((int) *pCom, 2, 2);
+    int8_t inB2 = (int8_t) bitext((int) *pCom, 2, 4);
+    int8_t inB3 = (int8_t) bitext((int) *pCom, 2, 6);
+    v4s inB4 = pack((int8_t) inB0, (int8_t) inB1, (int8_t) inB2, (int8_t) inB3);
+
+    *((v4s*)out) = maxs4(inA4, inB4);
+
+    int8_t inA = (int8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((int8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_tnn_compare_and_replace_if_larger_ternary(int8_t * base,
+                                                int8_t * target,
+                                                uint16_t length)
+{
+  uint8_t mask2 = 0x0c;
+  uint8_t n_mask2 = ~ mask2;
+  uint8_t mask4 = 0x30;
+  uint8_t n_mask4 = ~ mask4;
+  uint8_t mask6 = 0xc0;
+  uint8_t n_mask6 = ~ mask6;
+  uint8_t off2 = 2;
+  uint8_t off4 = 4;
+  uint8_t off6 = 6;
+
+  uint8_t *pIn = (uint8_t *) base;
+  uint8_t *pCom = (uint8_t *) target;
+  uint8_t *out;
+
+  int cnt = length >> 2;
+  uint32_t result;
+
+  while(cnt > 0u)
+  {
+    uint32_t in1 = *((uint32_t *)pIn);
+    uint32_t in2 = *((int32_t *)pCom);
+    result = maxs20(in1, in2);
+    *((uint32_t *)pIn) = result;
+
+    pIn+=4;
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  if (left>0u)
+  {
+    // do the vector max on the whole word - we won't use the leftover bytes
+    uint32_t in1 = *((uint32_t *)pIn);
+    uint32_t in2 = *((int32_t *)pCom);
+    result = maxs20(in1, in2);
+
+    // ...and copy back the relevant bytes of the result to pIn
+    for (int i=0; i<left; i++)
+      *((uint8_t*)(pIn + i)) = (uint8_t) (result >> (8*i));
+
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_avg_and_replace_u2(uint8_t * base,
+                                  uint8_t * target,
+                                  uint16_t length)
+{
+  int8_t mask2 = 0x0c;
+  int8_t n_mask2 = ~ mask2;
+  int8_t mask4 = 0x30;
+  int8_t n_mask4 = ~ mask4;
+  int8_t mask6 = 0xc0;
+  int8_t n_mask6 = ~ mask6;
+  int8_t off2 = 2;
+  int8_t off4 = 4;
+  int8_t off6 = 6;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+  uint8_t *out;
+
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    *((uint32_t *)pIn) = avg16(*((uint32_t *)pIn), *((uint32_t *)pCom));
+
+    pIn+=4;
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 2, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 2, 2);
+    uint8_t inA2 = (uint8_t) bitextu((unsigned int) *pIn, 2, 4);
+    uint8_t inA3 = (uint8_t) bitextu((unsigned int) *pIn, 2, 6);
+    v4u inA4 = pack((uint8_t) inA0, (uint8_t) inA1, (uint8_t) inA2, (uint8_t) inA3);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 2, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 2, 2);
+    uint8_t inB2 = (uint8_t) bitextu((unsigned int) *pCom, 2, 4);
+    uint8_t inB3 = (uint8_t) bitextu((unsigned int) *pCom, 2, 6);
+    v4u inB4 = pack((uint8_t) inB0, (uint8_t) inB1, (uint8_t) inB2, (uint8_t) inB3);
+
+    *((v4u*)out) = avg4(inA4, inB4);
+
+    uint8_t inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+#endif
diff --git a/rt_nn_tests/xptnn_linear/pulp_nn_utils_xpnn.h b/rt_nn_tests/xptnn_linear/pulp_nn_utils_xpnn.h
new file mode 100644
index 0000000..0c783ae
--- /dev/null
+++ b/rt_nn_tests/xptnn_linear/pulp_nn_utils_xpnn.h
@@ -0,0 +1,1937 @@
+/*
+ * pulp_nn_utils.h
+ * Nazareno   Bruschi  <nazareno.bruschi@unibo.it>
+ * Alessandro Nadalini <alessandro.nadalini3@unibo.it>
+ * Georg Rutishauser   <georgr@iis.ee.ethz.ch>
+ *
+ * Copyright (C) 2019-2020 ETH Zurich & University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef __PULPNN_UTILS__
+#define __PULPNN_UTILS__
+
+#include "pmsis.h"
+#ifdef GAP_SDK
+#include "pulp.h"
+#endif
+
+#define bitext(x,size,off)                                      __builtin_pulp_bextract(x,size,off)
+#define bitextu(x,size,off)                                     __builtin_pulp_bextractu(x,size,off)
+#define bitins(dst,not_mask_imm,src,mask_imm,off)               __builtin_pulp_binsert(dst,not_mask_imm,src,mask_imm,off)
+#define pack(x,y,z,t)                                           __builtin_pulp_pack4(x,y,z,t)
+#define max4(a,b)                                               __builtin_pulp_maxu4(a,b)
+#define maxs4(a, b)                                             __builtin_pulp_max4(a, b)
+#define max8(a, b)                                              __builtin_pulp_maxu8(a, b)
+#define maxs8(a, b)                                             __builtin_pulp_max8(a, b)
+#define max16(a, b)                                             __builtin_pulp_maxu16(a, b)
+#define maxs16(a, b)                                            __builtin_pulp_max16(a, b)
+#define max32(a,b)                                              __builtin_pulp_maxusi(a,b)
+#define maxs32(a,b)                                             __builtin_pulp_maxsi(a,b)
+#define min32(a,b)                                              __builtin_pulp_minusi(a,b)
+#define mins32(a,b)                                             __builtin_pulp_minsi(a,b)
+#define min4(a, b)                                              __builtin_pulp_minu4(a, b)
+#define mins4(a, b)                                             __builtin_pulp_min4(a, b)
+#define min8(a, b)                                              __builtin_pulp_minu8(a, b)
+#define mins8(a, b)                                             __builtin_pulp_min8(a, b)
+#define min16(a, b)                                             __builtin_pulp_minu16(a, b)
+#define mins16(a, b)                                            __builtin_pulp_min16(a, b)
+#define avg4(a,b)                                               __builtin_pulp_avgu4(a,b)
+#define avg8(a,b)                                               __builtin_pulp_avgu8(a,b)
+#define avg16(a,b)                                              __builtin_pulp_avgu16(a,b)
+#define log2(x)                                                 __builtin_pulp_fl1(x)
+#define min(a,b)                                                ((a)<(b)?(a):(b))
+#define SumDotp4(a, b, c)                                       __builtin_pulp_sdotusp4(a, b, c)
+#define SumDotp8(a, b, c)                                       __builtin_pulp_sdotusp8(a, b, c)
+#define SumDotp16(a, b, c)                                      __builtin_pulp_sdotusp16(a, b, c)
+#define SumDotps4(a, b, c)                                      __builtin_pulp_sdotsp4(a, b, c)
+#define SumDotps8(a, b, c)                                      __builtin_pulp_sdotsp8(a, b, c)
+#define SumDotps16(a, b, c)                                     __builtin_pulp_sdotsp16(a, b, c)
+#define clip4(x)                                                __builtin_pulp_clipu_r(x, 15)
+#define clip2(x)                                                __builtin_pulp_clipu_r(x, 3)
+#define clip8(x)                                                __builtin_pulp_clipu_r(x, 255)
+
+#define clips4(x)                                               __builtin_pulp_clip_r(x, 7)
+#define clips2(x)                                               __builtin_pulp_clip_r(x, 1)
+#define clips8(x)                                               __builtin_pulp_clip_r(x, 127)
+#define MacLoadInit(a_update, b_update, a_reg, b_reg, ptr)      __builtin_pulp_mlinitspr_v3(a_update, b_update, a_reg, b_reg, ptr)
+#define MacLoadUpdate(ptr)                                      __builtin_pulp_mlupdatespr_v3(ptr)
+#define MacLoadAssign(ptr)                                      __builtin_pulp_mlassignspr_v3(ptr)
+#define MacLoad4(a_update, b_update, a_reg, b_reg, ptr, sum)    __builtin_pulp_mlsdotsup4_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+#define MacLoad8(a_update, b_update, a_reg, b_reg, ptr, sum)    __builtin_pulp_mlsdotsup8_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+#define MacLoad16(a_update, b_update, a_reg, b_reg, ptr, sum)   __builtin_pulp_mlsdotsup16_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+#define MacLoads4(a_update, b_update, a_reg, b_reg, ptr, sum)   __builtin_pulp_mlsdotsp4_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+#define MacLoads8(a_update, b_update, a_reg, b_reg, ptr, sum)   __builtin_pulp_mlsdotsp8_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+#define MacLoads16(a_update, b_update, a_reg, b_reg, ptr, sum)  __builtin_pulp_mlsdotsp16_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+
+#define PACK_INT8_SIZE(x)                                    (x)
+#define PACK_INT4_SIZE(x)                                    ((x) >> 1)
+#define PACK_INT2_SIZE(x)                                    ((x) >> 2)
+
+#define MemoryFence()                                        asm volatile("":::"memory")
+
+#define LEGACY_MODE(x)                                       asm volatile ("csrwi 0x010," x)
+#define IVEC_FMT(x)                                          asm volatile ("csrwi 0x00D," x)
+#define MIXED_SKIP(x)                                        asm volatile ("csrwi 0x00F," x)
+#define A_ADDRESS(x)                                         asm volatile ("csrw 0x100, %0" :: "r" (x))
+#define W_ADDRESS(x)                                         asm volatile ("csrw 0x101, %0" :: "r" (x))
+#define A_STRIDE(x)                                          asm volatile ("csrw 0x102, %0":: "r" (x))
+#define W_STRIDE(x)                                          asm volatile ("csrw 0x103, %0":: "r" (x))
+#define A_ROLLBACK(x)                                        asm volatile ("csrw 0x104, %0":: "r" (x))
+#define W_ROLLBACK(x)                                        asm volatile ("csrw 0x105, %0":: "r" (x))
+#define A_SKIP(x)                                            asm volatile ("csrwi 0x106," x)
+#define W_SKIP(x)                                            asm volatile ("csrwi 0x107," x)
+
+static uint8_t __attribute__((noinline)) pulp_nn_quant_u2 (
+  int32_t phi,
+  int16_t m,
+  int8_t d
+  ) {
+  int32_t x = (m * phi) >> d;
+  uint8_t res = clip2(x);
+  return res;
+}
+static uint8_t __attribute__((noinline)) pulp_nn_bn_quant_u2 (
+  int32_t phi,
+  int32_t k,
+  int32_t lambda,
+  int8_t d
+  ) {
+  int32_t integer_image_phi = (k * phi) + lambda;
+  int32_t x = (integer_image_phi) >> d;
+  uint8_t res = clip2(x);
+  return res;
+}
+static uint8_t __attribute__((noinline)) pulp_nn_add_quant_u2 (
+  uint8_t pix1,
+  uint8_t pix2,
+  int16_t m1,
+  int16_t m2,
+  int8_t d
+  ) {
+  uint32_t integer_image = pix1*m1 + pix2*m2;
+  uint32_t x = (integer_image) >> d;
+  uint8_t res = clip2(x);
+  return res;
+}
+
+static int8_t __attribute__((noinline)) pulp_nn_quant_i2 (
+  int32_t phi,
+  int16_t m,
+  int8_t d
+  ) {
+  int32_t x = (m * phi) >> d;
+  int8_t res = clips2(x);
+  return res;
+}
+static int8_t __attribute__((noinline)) pulp_nn_bn_quant_i2 (
+  int32_t phi,
+  int32_t k,
+  int32_t lambda,
+  int8_t d
+  ) {
+  int32_t integer_image_phi = (k * phi) + lambda;
+  int32_t x = (integer_image_phi) >> d;
+  int8_t res = clips2(x);
+  return res;
+}
+static int8_t __attribute__((noinline)) pulp_nn_add_quant_i2 (
+  int8_t pix1,
+  int8_t pix2,
+  int16_t m1,
+  int16_t m2,
+  int8_t d
+  ) {
+  int32_t integer_image = pix1*m1 + pix2*m2;
+  int32_t x = (integer_image) >> d;
+  int8_t res = clips2(x);
+  return res;
+}
+
+static uint8_t __attribute__((noinline)) pulp_nn_quant_u4 (
+  int32_t phi,
+  int16_t m,
+  int8_t d
+  ) {
+  int32_t x = (m * phi) >> d;
+  uint8_t res = clip4(x);
+  return res;
+}
+static uint8_t __attribute__((noinline)) pulp_nn_bn_quant_u4 (
+  int32_t phi,
+  int32_t k,
+  int32_t lambda,
+  int8_t d
+  ) {
+  int32_t integer_image_phi = (k * phi) + lambda;
+  int32_t x = (integer_image_phi) >> d;
+  uint8_t res = clip4(x);
+  return res;
+}
+static uint8_t __attribute__((noinline)) pulp_nn_add_quant_u4 (
+  uint8_t pix1,
+  uint8_t pix2,
+  int16_t m1,
+  int16_t m2,
+  int8_t d
+  ) {
+  uint32_t integer_image = pix1*m1 + pix2*m2;
+  uint32_t x = (integer_image) >> d;
+  uint8_t res = clip4(x);
+  return res;
+}
+
+static int8_t __attribute__((noinline)) pulp_nn_quant_i4 (
+  int32_t phi,
+  int16_t m,
+  int8_t d
+  ) {
+  int32_t x = (m * phi) >> d;
+  int8_t res = clips4(x);
+  return res;
+}
+static int8_t __attribute__((noinline)) pulp_nn_bn_quant_i4 (
+  int32_t phi,
+  int32_t k,
+  int32_t lambda,
+  int8_t d
+  ) {
+  int32_t integer_image_phi = (k * phi) + lambda;
+  int32_t x = (integer_image_phi) >> d;
+  int8_t res = clips4(x);
+  return res;
+}
+static int8_t __attribute__((noinline)) pulp_nn_add_quant_i4 (
+  int8_t pix1,
+  int8_t pix2,
+  int16_t m1,
+  int16_t m2,
+  int8_t d
+  ) {
+  int32_t integer_image = pix1*m1 + pix2*m2;
+  int32_t x = (integer_image) >> d;
+  int8_t res = clips4(x);
+  return res;
+}
+
+static uint8_t __attribute__((noinline)) pulp_nn_quant_u8 (
+  int32_t phi,
+  int16_t m,
+  int8_t d
+  ) {
+  int32_t x = (m * phi) >> d;
+  uint8_t res = clip8(x);
+  return res;
+}
+static uint8_t __attribute__((noinline)) pulp_nn_bn_quant_u8 (
+  int32_t phi,
+  int32_t k,
+  int32_t lambda,
+  int8_t d
+  ) {
+  int32_t integer_image_phi = (k * phi) + lambda;
+  int32_t x = (integer_image_phi) >> d;
+  uint8_t res = clip8(x);
+  return res;
+}
+static uint8_t __attribute__((noinline)) pulp_nn_add_quant_u8 (
+  uint8_t pix1,
+  uint8_t pix2,
+  int16_t m1,
+  int16_t m2,
+  int8_t d
+  ) {
+  uint32_t integer_image = pix1*m1 + pix2*m2;
+  uint32_t x = (integer_image) >> d;
+  uint8_t res = clip8(x);
+  return res;
+}
+
+static int8_t __attribute__((noinline)) pulp_nn_quant_i8 (
+  int32_t phi,
+  int16_t m,
+  int8_t d
+  ) {
+  int32_t x = (m * phi) >> d;
+  int8_t res = clips8(x);
+  return res;
+}
+static int8_t __attribute__((noinline)) pulp_nn_bn_quant_i8 (
+  int32_t phi,
+  int32_t k,
+  int32_t lambda,
+  int8_t d
+  ) {
+  int32_t integer_image_phi = (k * phi) + lambda;
+  int32_t x = (integer_image_phi) >> d;
+  int8_t res = clips8(x);
+  return res;
+}
+static int8_t __attribute__((noinline)) pulp_nn_add_quant_i8 (
+  int8_t pix1,
+  int8_t pix2,
+  int16_t m1,
+  int16_t m2,
+  int8_t d
+  ) {
+  int32_t integer_image = pix1*m1 + pix2*m2;
+  int32_t x = (integer_image) >> d;
+  int8_t res = clips8(x);
+  return res;
+}
+
+
+static uint8_t __attribute__((noinline)) pulp_nn_u4_quant(int input, int16_t * pThr)
+{
+  if(input <= pThr[7] )
+  {
+    if( input <= pThr[3])
+    {
+      if( input <= pThr[1])
+      {
+        if( input <= pThr[0])
+          return 0;
+        else
+          return 1;
+      }
+      else
+      {
+        if( input <= pThr[2])
+          return 2;
+        else
+          return 3;
+      }
+    }
+    else
+    {
+      if( input <= pThr[5])
+      {
+        if( input <= pThr[4])
+          return 4;
+        else
+          return 5;
+      }
+      else
+      {
+        if( input <= pThr[6])
+          return 6;
+        else
+          return 7;
+      }
+    }
+  }
+  else
+  {
+    if( input <= pThr[11])
+    {
+      if( input <= pThr[9])
+      {
+        if( input <= pThr[8])
+          return 8;
+        else
+          return 9;
+      }
+      else
+      {
+        if( input <= pThr[10])
+          return 10;
+        else
+          return 11;
+      }
+    }
+    else
+    {
+      if( input <= pThr[13])
+      {
+        if( input <= pThr[12])
+          return 12;
+        else
+          return 13;
+      }
+      else
+      {
+        if( input <= pThr[14])
+          return 14;
+        else
+          return 15;
+      }
+    }
+  }
+}
+
+static uint8_t __attribute__((noinline)) pulp_nn_u2_quant(int input, int16_t * pThr)
+{
+  if( input <= pThr[1])
+  {
+    if( input <= pThr[0])
+        {
+      return 0;
+    }
+        else
+        {
+      return 1;
+    }
+  }
+    else
+    {
+    if( input <= pThr[2])
+        {
+      return 2;
+    }
+        else
+        {
+      return 3;
+    }
+  }
+}
+
+/*
+ * Common
+ */
+
+
+static v4s __attribute__((noinline)) pulp_nn_i4_to_i8_r( int8_t *pSrc)
+{
+  v4s Src = *((v4s*) pSrc);
+  int8_t bext1, bext2, bext3, bext4;
+
+  bext1 = (int8_t) bitext((int) Src, 4, 0);
+  bext2 = (int8_t) bitext((int) Src, 4, 4);
+  bext3 = (int8_t) bitext((int) Src, 4, 8);
+  bext4 = (int8_t) bitext((int) Src, 4, 12);
+  v4s res = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+
+  return res;
+}
+
+static v4u __attribute__((noinline)) pulp_nn_u4_to_u8_r(uint8_t *pSrc)
+{
+  v4u Src = *((v4u*) pSrc);
+  uint8_t bext1, bext2, bext3, bext4;
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 4, 0);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 4, 4);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 4, 8);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 4, 12);
+  v4u res = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+
+  return res;
+}
+
+static v4s __attribute__((noinline)) pulp_nn_i2_to_i8_r( int8_t *pSrc)
+{
+  v4s Src = *((v4s*) pSrc);
+  int8_t bext1, bext2, bext3, bext4;
+
+  bext1 = (int8_t) bitext((int) Src, 2, 0);
+  bext2 = (int8_t) bitext((int) Src, 2, 2);
+  bext3 = (int8_t) bitext((int) Src, 2, 4);
+  bext4 = (int8_t) bitext((int) Src, 2, 6);
+  v4s res = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+
+  return res;
+}
+
+static v4u __attribute__((noinline)) pulp_nn_u2_to_u8_r(uint8_t *pSrc)
+{
+  v4u Src = *((v4u*) pSrc);
+  uint8_t bext1, bext2, bext3, bext4;
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 0);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 2);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 4);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 6);
+  v4u res = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+
+  return res;
+}
+
+static v4s __attribute__((noinline)) pulp_nn_i2_to_i4_r(int8_t *pSrc)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  v4s Src = *((v4s*) pSrc);
+  int8_t bext1, bext2, bext3, bext4;
+  int8_t out1, out2, out3, out4;
+
+  bext1 = (int8_t) bitextu((int) Src, 2, 0);
+  bext2 = (int8_t) bitextu((int) Src, 2, 2);
+  bext3 = (int8_t) bitextu((int) Src, 2, 4);
+  bext4 = (int8_t) bitextu((int) Src, 2, 6);
+
+  out1 = bitins(bext1, n_mask, bext2, mask, off);
+  out2 = bitins(bext3, n_mask, bext4, mask, off);
+
+  bext1 = (int8_t) bitextu((int) Src, 2, 8);
+  bext2 = (int8_t) bitextu((int) Src, 2, 10);
+  bext3 = (int8_t) bitextu((int) Src, 2, 12);
+  bext4 = (int8_t) bitextu((int) Src, 2, 14);
+
+  out3 = bitins(bext1, n_mask, bext2, mask, off);
+  out4 = bitins(bext3, n_mask, bext4, mask, off);
+
+  v4s res = pack((int8_t) out1, (int8_t) out2, (int8_t) out3, (int8_t) out4);
+
+  return res;
+}
+
+static v4u __attribute__((noinline)) pulp_nn_u2_to_u4_r(uint8_t *pSrc)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  v4u Src = *((v4u*) pSrc);
+  uint8_t bext1, bext2, bext3, bext4;
+  uint8_t out1, out2, out3, out4;
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 0);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 2);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 4);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 6);
+
+  out1 = bitins(bext1, n_mask, bext2, mask, off);
+  out2 = bitins(bext3, n_mask, bext4, mask, off);
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 8);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 10);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 12);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 14);
+
+  out3 = bitins(bext1, n_mask, bext2, mask, off);
+  out4 = bitins(bext3, n_mask, bext4, mask, off);
+
+  v4u res = pack((uint8_t) out1, (uint8_t) out2, (uint8_t) out3, (uint8_t) out4);
+
+  return res;
+}
+
+static int8_t *__attribute__((always_inline)) pulp_nn_i4_to_i8( int8_t *pSrc, int8_t *pDst)
+{
+  v4s Src = *((v4s*) pSrc);
+  int8_t bext1, bext2, bext3, bext4;
+
+  pSrc+=4;
+
+  bext1 = (int8_t) bitext((int) Src, 4, 0);
+  bext2 = (int8_t) bitext((int) Src, 4, 4);
+  bext3 = (int8_t) bitext((int) Src, 4, 8);
+  bext4 = (int8_t) bitext((int) Src, 4, 12);
+  *((v4s*)pDst) = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+   MemoryFence();
+  pDst+=4;
+  bext1 = (int8_t) bitext((int) Src, 4, 16);
+  bext2 = (int8_t) bitext((int) Src, 4, 20);
+  bext3 = (int8_t) bitext((int) Src, 4, 24);
+  bext4 = (int8_t) bitext((int) Src, 4, 28);
+  *((v4s*)pDst) = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+
+  return pSrc;
+}
+
+static uint8_t *__attribute__((always_inline)) pulp_nn_u4_to_u8(uint8_t *pSrc, uint8_t *pDst)
+{
+  v4u Src = *((v4u*) pSrc);
+  uint8_t bext1, bext2, bext3, bext4;
+
+  pSrc+=4;
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 4, 0);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 4, 4);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 4, 8);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 4, 12);
+  *((v4u*)pDst) = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 4, 16);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 4, 20);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 4, 24);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 4, 28);
+  *((v4u*)pDst) = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+
+  return pSrc;
+}
+
+static int8_t *__attribute__((always_inline)) pulp_nn_i2_to_i8( int8_t * pSrc, int8_t * pDst)
+{
+  v4s Src = *((v4s*) pSrc);
+  int8_t bext1, bext2, bext3, bext4;
+
+  pSrc+=4;
+
+  bext1 = (int8_t) bitext((int) Src, 2, 0);
+  bext2 = (int8_t) bitext((int) Src, 2, 2);
+  bext3 = (int8_t) bitext((int) Src, 2, 4);
+  bext4 = (int8_t) bitext((int) Src, 2, 6);
+  *((v4s*)pDst) = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (int8_t) bitext((int) Src, 2, 8);
+  bext2 = (int8_t) bitext((int) Src, 2, 10);
+  bext3 = (int8_t) bitext((int) Src, 2, 12);
+  bext4 = (int8_t) bitext((int) Src, 2, 14);
+  *((v4s*)pDst) = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (int8_t) bitext((int) Src, 2, 16);
+  bext2 = (int8_t) bitext((int) Src, 2, 18);
+  bext3 = (int8_t) bitext((int) Src, 2, 20);
+  bext4 = (int8_t) bitext((int) Src, 2, 22);
+  *((v4s*)pDst) = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (int8_t) bitext((int) Src, 2, 24);
+  bext2 = (int8_t) bitext((int) Src, 2, 26);
+  bext3 = (int8_t) bitext((int) Src, 2, 28);
+  bext4 = (int8_t) bitext((int) Src, 2, 30);
+  *((v4s*)pDst) = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+
+  return pSrc;
+}
+
+static uint8_t *__attribute__((always_inline)) pulp_nn_u2_to_u8(uint8_t * pSrc, uint8_t * pDst)
+{
+  v4u Src = *((v4u*) pSrc);
+  uint8_t bext1, bext2, bext3, bext4;
+
+  pSrc+=4;
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 0);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 2);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 4);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 6);
+  *((v4u*)pDst) = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 8);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 10);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 12);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 14);
+  *((v4u*)pDst) = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 16);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 18);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 20);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 22);
+  *((v4u*)pDst) = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 24);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 26);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 28);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 30);
+  *((v4u*)pDst) = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+
+  return pSrc;
+}
+
+static int8_t *__attribute__((always_inline)) pulp_nn_i2_to_i4( int8_t * pSrc, int8_t * pDst)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  v4s Src = *((v4s*) pSrc);
+  int8_t bext1, bext2, bext3, bext4;
+  int8_t out1, out2, out3, out4;
+
+  pSrc+=4;
+
+  bext1 = (int8_t) bitext((int) Src, 2, 0);
+  bext2 = (int8_t) bitext((int) Src, 2, 2);
+  bext3 = (int8_t) bitext((int) Src, 2, 4);
+  bext4 = (int8_t) bitext((int) Src, 2, 6);
+
+  out1 = bitins(bext1, n_mask, bext2, mask, off);
+  out2 = bitins(bext3, n_mask, bext4, mask, off);
+  MemoryFence();
+
+  bext1 = (int8_t) bitext((int) Src, 2, 8);
+  bext2 = (int8_t) bitext((int) Src, 2, 10);
+  bext3 = (int8_t) bitext((int) Src, 2, 12);
+  bext4 = (int8_t) bitext((int) Src, 2, 14);
+
+  out3 = bitins(bext1, n_mask, bext2, mask, off);
+  out4 = bitins(bext3, n_mask, bext4, mask, off);
+
+  *((v4s*)pDst) = pack((int8_t) out1, (int8_t) out2, (int8_t) out3, (int8_t) out4);
+  MemoryFence();
+
+  pDst+=4;
+  bext1 = (int8_t) bitext((int) Src, 2, 16);
+  bext2 = (int8_t) bitext((int) Src, 2, 18);
+  bext3 = (int8_t) bitext((int) Src, 2, 20);
+  bext4 = (int8_t) bitext((int) Src, 2, 22);
+
+  out1 = bitins(bext1, n_mask, bext2, mask, off);
+  out2 = bitins(bext3, n_mask, bext4, mask, off);
+  MemoryFence();
+
+  bext1 = (int8_t) bitext((int) Src, 2, 24);
+  bext2 = (int8_t) bitext((int) Src, 2, 26);
+  bext3 = (int8_t) bitext((int) Src, 2, 28);
+  bext4 = (int8_t) bitext((int) Src, 2, 30);
+
+  out3 = bitins(bext1, n_mask, bext2, mask, off);
+  out4 = bitins(bext3, n_mask, bext4, mask, off);
+
+  *((v4s*)pDst) = pack((int8_t) out1, (int8_t) out2, (int8_t) out3, (int8_t) out4);
+
+  return pSrc;
+}
+
+static uint8_t *__attribute__((always_inline)) pulp_nn_u2_to_u4( uint8_t * pSrc, uint8_t * pDst)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  v4u Src = *((v4u*) pSrc);
+  uint8_t bext1, bext2, bext3, bext4;
+  uint8_t out1, out2, out3, out4;
+
+  pSrc+=4;
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 0);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 2);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 4);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 6);
+
+  out1 = bitins(bext1, n_mask, bext2, mask, off);
+  out2 = bitins(bext3, n_mask, bext4, mask, off);
+  MemoryFence();
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 8);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 10);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 12);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 14);
+
+  out3 = bitins(bext1, n_mask, bext2, mask, off);
+  out4 = bitins(bext3, n_mask, bext4, mask, off);
+
+  *((v4u*)pDst) = pack((uint8_t) out1, (uint8_t) out2, (uint8_t) out3, (uint8_t) out4);
+  MemoryFence();
+
+  pDst+=4;
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 16);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 18);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 20);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 22);
+
+  out1 = bitins(bext1, n_mask, bext2, mask, off);
+  out2 = bitins(bext3, n_mask, bext4, mask, off);
+  MemoryFence();
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 24);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 26);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 28);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 30);
+
+  out3 = bitins(bext1, n_mask, bext2, mask, off);
+  out4 = bitins(bext3, n_mask, bext4, mask, off);
+  
+  *((v4u*)pDst) = pack((uint8_t) out1, (uint8_t) out2, (uint8_t) out3, (uint8_t) out4);
+
+  return pSrc;
+}
+
+/*
+ * XpulpV2
+ */
+
+static void __attribute__((noinline)) pulp_zero_mem(uint8_t * pBuffer, unsigned int size)
+{
+  int lfover = size &0x3;
+  for (int i=0; i<(size>>2); i++)
+  {
+    *((v4u *)pBuffer) = (v4u){0,0,0,0};
+    MemoryFence();
+    pBuffer+=4;
+  }
+  while(lfover)
+  {
+    *pBuffer++=0;
+    lfover--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_im2col_u2_to_u8(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_u2_to_u8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 16;
+  }
+  while (lfover)
+  {
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 0);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 2);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 4);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 6);
+    pOutput++;
+    pInput++;
+    lfover -= 4;
+  }
+}
+static void __attribute__((noinline)) pulp_nn_im2col_i2_to_i8(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_i2_to_i8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 16;
+  }
+  while (lfover)
+  {
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 0);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 2);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 4);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 6);
+    pOutput++;
+    pInput++;
+    lfover -= 4;
+  }
+}
+static void __attribute__((noinline)) pulp_nn_im2col_u4_to_u8(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 3u;
+  int lfover = blockSize & 0x07;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_u4_to_u8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 8;
+  }
+  while (lfover)
+  {
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 4, 0);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 4, 4);
+    pOutput++;
+    pInput++;
+    lfover -= 2;
+  }
+}
+static void __attribute__((noinline)) pulp_nn_im2col_i4_to_i8(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 3u;
+  int lfover = blockSize & 0x07;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_i4_to_i8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 8;
+  }
+  while (lfover)
+  {
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 4, 0);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 4, 4);
+    pOutput++;
+    pInput++;
+    lfover -= 2;
+  }
+}
+static void __attribute__((noinline)) pulp_nn_im2col_u8_to_u8(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 2u;
+  int lfover = blockSize & 0x03;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4u*)pOutput) = *((v4u*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    *((uint8_t*)pOutput) = *((uint8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover--;
+  }
+}
+static void __attribute__((noinline)) pulp_nn_im2col_i8_to_i8(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 2u;
+  int lfover = blockSize & 0x03;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4s*)pOutput) = *((v4s*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    *((int8_t*)pOutput) = *((int8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover--;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_im2col_u2_to_u2(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4u*)pOutput) = *((v4u*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    uint8_t extr;
+    *((uint8_t*)pOutput) = *((uint8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover -= 4;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_i2_to_i2(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4s*)pOutput) = *((v4s*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    int8_t extr;
+    *((int8_t*)pOutput) = *((int8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover -= 4;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_u2_to_u4(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_u2_to_u4(pInput, pOutput);
+    MemoryFence();
+    pOutput += 8;
+  }
+  while (lfover)
+  {
+    uint8_t extr;
+    // TODO: this is potentially dangerous/wrong if the number of channels
+    // is not a multiple of 8!
+    *((v4u *)pOutput) = pulp_nn_u2_to_u4_r(pInput);
+    pInput += 2;
+    pOutput += 4;
+    lfover -= 8;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_i2_to_i4(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_i2_to_i4(pInput, pOutput);
+    MemoryFence();
+    pOutput += 8;
+  }
+  while (lfover)
+  {
+    int8_t extr;
+    // TODO: this is potentially dangerous/wrong if the number of channels
+    // is not a multiple of 8!
+    *((v4s *)pOutput) = pulp_nn_i2_to_i4_r(pInput);
+    pInput += 2;
+    pOutput += 4;
+    lfover -= 8;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_u2_to_u8(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_u2_to_u8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 16;
+  }
+  while (lfover)
+  {
+    uint8_t extr;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 0);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 2);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 4);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 6);
+    pOutput++;
+    pInput++;
+    lfover -= 4;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_i2_to_i8(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_i2_to_i8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 16;
+  }
+  while (lfover)
+  {
+    int8_t extr;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 0);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 2);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 4);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 6);
+    pOutput++;
+    pInput++;
+    lfover -= 4;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_u4_to_u4(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 3u;
+  int lfover = blockSize & 0x07;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4u*)pOutput) = *((v4u*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    uint8_t extr;
+    *((uint8_t*)pOutput) = *((uint8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover -= 2;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_i4_to_i4(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 3u;
+  int lfover = blockSize & 0x07;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4s*)pOutput) = *((v4s*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    int8_t extr;
+    *((int8_t*)pOutput) = *((int8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover -= 2;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_u4_to_u8(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 3u;
+  int lfover = blockSize & 0x07;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_u4_to_u8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 8;
+  }
+  while (lfover)
+  {
+    uint8_t extr;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 4, 0);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 4, 4);
+    pOutput++;
+    pInput++;
+    lfover -= 2;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_i4_to_i8(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 3u;
+  int lfover = blockSize & 0x07;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_i4_to_i8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 8;
+  }
+  while (lfover)
+  {
+    int8_t extr;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 4, 0);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 4, 4);
+    pOutput++;
+    pInput++;
+    lfover -= 2;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_u8_to_u8(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 2u;
+  int lfover = blockSize & 0x03;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4u*)pOutput) = *((v4u*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    uint8_t extr;
+    *((uint8_t*)pOutput) = *((uint8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover -= 1;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_i8_to_i8(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 2u;
+  int lfover = blockSize & 0x03;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4s*)pOutput) = *((v4s*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    int8_t extr;
+    *((int8_t*)pOutput) = *((int8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover -= 1;
+  }
+}
+
+
+
+static void __attribute__((noinline)) pulp_nn_compare_and_replace_if_larger_u8(uint8_t * base,
+                                                uint8_t * target,
+                                                uint16_t length)
+{
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+  v4u inp;
+  v4u com;
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    inp = *((v4u*)pIn);
+    com = *((v4u*)pCom);
+
+    *((v4u*)pIn) = max4(inp, com);
+
+    pCom+=4;
+    pIn+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    if(*pIn<*pCom)
+      *pIn=*pCom;
+    
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_compare_and_replace_if_larger_i8(
+    int8_t * base, int8_t * target, uint16_t length) {
+  int8_t *pIn = base;
+  int8_t *pCom = target;
+  v4s inp;
+  v4s com;
+  int cnt = length >> 2;
+
+  while (cnt > 0u) {
+    inp = *((v4s *)pIn);
+    com = *((v4s *)pCom);
+
+    *((v4s *)pIn) = maxs4(inp, com);
+
+    pCom += 4;
+    pIn += 4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left > 0u) {
+    if (*pIn < *pCom)
+      *pIn = *pCom;
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_avg_and_replace_u8(uint8_t * base,
+                                  uint8_t * target,
+                                  uint16_t length)
+{
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+
+  while (length>0u)
+  {
+    *pIn = ((*pIn + *pCom) >> 1);
+
+    pIn++;
+    pCom++;
+    length--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_compare_and_replace_if_larger_u4(uint8_t * base,
+                                                uint8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+  v4u inp[2];
+  v4u com[2];
+  uint8_t *out;
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    pulp_nn_u4_to_u8(pIn, (uint8_t *)inp);
+    pulp_nn_u4_to_u8(pCom, (uint8_t *)com);
+
+    *((v4u *)out) = max4(inp[0], com[0]);
+
+    *((uint8_t*)pIn) = bitins(*out, n_mask, *(out + 1), mask, off);
+    pIn++;
+    *((uint8_t*)pIn) = bitins(*(out + 2), n_mask, *(out + 3), mask, off);
+    pIn++;
+
+    *((v4u *)out) = max4(inp[1], com[1]);
+
+    *((uint8_t*)pIn) = bitins(*out, n_mask, *(out + 1), mask, off);
+    pIn++;
+    *((uint8_t*)pIn) = bitins(*(out + 2), n_mask, *(out + 3), mask, off);
+    pIn++;
+
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 4, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 4, 4);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 4, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 4, 4);
+
+    if(inA0<inB0)
+      inA0=inB0;
+
+    if(inA1<inB1)
+      inA1=inB1;
+
+    *((uint8_t*)pIn) = bitins(inA0, n_mask, inA1, mask, off);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_compare_and_replace_if_larger_i4(int8_t * base,
+                                                int8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  int8_t *pIn = base;
+  int8_t *pCom = target;
+  v4s inp[2];
+  v4s com[2];
+  int8_t *out;
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    pulp_nn_i4_to_i8(pIn, (int8_t *)inp);
+    pulp_nn_i4_to_i8(pCom, (int8_t *)com);
+
+    *((v4s *)out) = maxs4(inp[0], com[0]);
+
+    *((int8_t*)pIn) = bitins(*out, n_mask, *(out + 1), mask, off);
+    pIn++;
+    *((int8_t*)pIn) = bitins(*(out + 2), n_mask, *(out + 3), mask, off);
+    pIn++;
+
+    *((v4s *)out) = maxs4(inp[1], com[1]);
+
+    *((int8_t*)pIn) = bitins(*out, n_mask, *(out + 1), mask, off);
+    pIn++;
+    *((int8_t*)pIn) = bitins(*(out + 2), n_mask, *(out + 3), mask, off);
+    pIn++;
+
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    int8_t inA0 = (int8_t) bitext((int) *pIn, 4, 0);
+    int8_t inA1 = (int8_t) bitext((int) *pIn, 4, 4);
+    int8_t inB0 = (int8_t) bitext((int) *pCom, 4, 0);
+    int8_t inB1 = (int8_t) bitext((int) *pCom, 4, 4);
+
+    if(inA0<inB0)
+      inA0=inB0;
+
+    if(inA1<inB1)
+      inA1=inB1;
+
+    *((int8_t*)pIn) = bitins(inA0, n_mask, inA1, mask, off);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_avg_and_replace_u4(uint8_t * base,
+                                  uint8_t * target,
+                                  uint16_t length)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+
+  while (length>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 4, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 4, 4);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 4, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 4, 4);
+
+    inA0 = ((inA0 + inB0) >> 1);
+    inA1 = ((inA1 + inB1) >> 1);
+
+    *((uint8_t*)pIn) = bitins(inA0, n_mask, inA1, mask, off);
+
+    pIn++;
+    pCom++;
+    length--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_compare_and_replace_if_larger_u2(uint8_t * base,
+                                                uint8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask2 = 0x0c;
+  int8_t n_mask2 = ~ mask2;
+  int8_t mask4 = 0x30;
+  int8_t n_mask4 = ~ mask4;
+  int8_t mask6 = 0xc0;
+  int8_t n_mask6 = ~ mask6;
+  int8_t off2 = 2;
+  int8_t off4 = 4;
+  int8_t off6 = 6;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+  v4u inp[4];
+  v4u com[4];
+  uint8_t *out;
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    pulp_nn_u2_to_u8(pIn, inp);
+    pulp_nn_u2_to_u8(pCom, com);
+
+    *((v4u*)out) = max4(inp[0], com[0]);
+
+    uint8_t inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    *((v4u*)out) = max4(inp[1], com[1]);
+
+    inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    *((v4u*)out) = max4(inp[2], com[2]);
+
+    inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    *((v4u*)out) = max4(inp[3], com[3]);
+
+    inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 2, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 2, 2);
+    uint8_t inA2 = (uint8_t) bitextu((unsigned int) *pIn, 2, 4);
+    uint8_t inA3 = (uint8_t) bitextu((unsigned int) *pIn, 2, 6);
+    v4u inA4 = pack((uint8_t) inA0, (uint8_t) inA1, (uint8_t) inA2, (uint8_t) inA3);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 2, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 2, 2);
+    uint8_t inB2 = (uint8_t) bitextu((unsigned int) *pCom, 2, 4);
+    uint8_t inB3 = (uint8_t) bitextu((unsigned int) *pCom, 2, 6);
+    v4u inB4 = pack((uint8_t) inB0, (uint8_t) inB1, (uint8_t) inB2, (uint8_t) inB3);
+
+    *((v4u*)out) = max4(inA4, inB4);
+
+    uint8_t inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+
+static void __attribute__((noinline)) pulp_nn_compare_and_replace_if_larger_i2(int8_t * base,
+                                                int8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask2 = 0x0c;
+  int8_t n_mask2 = ~ mask2;
+  int8_t mask4 = 0x30;
+  int8_t n_mask4 = ~ mask4;
+  int8_t mask6 = 0xc0;
+  int8_t n_mask6 = ~ mask6;
+  int8_t off2 = 2;
+  int8_t off4 = 4;
+  int8_t off6 = 6;
+
+  int8_t *pIn = base;
+  int8_t *pCom = target;
+  v4s inp[4];
+  v4s com[4];
+  int8_t *out;
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    pulp_nn_i2_to_i8(pIn, inp);
+    pulp_nn_i2_to_i8(pCom, com);
+
+    *((v4s*)out) = maxs4(inp[0], com[0]);
+
+    int8_t inA = (int8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((int8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    *((v4s*)out) = maxs4(inp[1], com[1]);
+
+    inA = (int8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((int8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    *((v4s*)out) = maxs4(inp[2], com[2]);
+
+    inA = (int8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((int8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    *((v4s*)out) = maxs4(inp[3], com[3]);
+
+    inA = (int8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((int8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    int8_t inA0 = (int8_t) bitext((unsigned int) *pIn, 2, 0);
+    int8_t inA1 = (int8_t) bitext((unsigned int) *pIn, 2, 2);
+    int8_t inA2 = (int8_t) bitext((unsigned int) *pIn, 2, 4);
+    int8_t inA3 = (int8_t) bitext((unsigned int) *pIn, 2, 6);
+    v4s inA4 = pack((int8_t) inA0, (int8_t) inA1, (int8_t) inA2, (int8_t) inA3);
+    int8_t inB0 = (int8_t) bitext((unsigned int) *pCom, 2, 0);
+    int8_t inB1 = (int8_t) bitext((unsigned int) *pCom, 2, 2);
+    int8_t inB2 = (int8_t) bitext((unsigned int) *pCom, 2, 4);
+    int8_t inB3 = (int8_t) bitext((unsigned int) *pCom, 2, 6);
+    v4s inB4 = pack((int8_t) inB0, (int8_t) inB1, (int8_t) inB2, (int8_t) inB3);
+
+    *((v4s*)out) = maxs4(inA4, inB4);
+
+    int8_t inA = (int8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((int8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_avg_and_replace_u2(uint8_t * base,
+                                  uint8_t * target,
+                                  uint16_t length)
+{
+  int8_t mask2 = 0x0c;
+  int8_t n_mask2 = ~ mask2;
+  int8_t mask4 = 0x30;
+  int8_t n_mask4 = ~ mask4;
+  int8_t mask6 = 0xc0;
+  int8_t n_mask6 = ~ mask6;
+  int8_t off2 = 2;
+  int8_t off4 = 4;
+  int8_t off6 = 6;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+
+  while (length>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 2, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 2, 2);
+    uint8_t inA2 = (uint8_t) bitextu((unsigned int) *pIn, 2, 4);
+    uint8_t inA3 = (uint8_t) bitextu((unsigned int) *pIn, 2, 6);
+
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 2, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 2, 2);
+    uint8_t inB2 = (uint8_t) bitextu((unsigned int) *pCom, 2, 4);
+    uint8_t inB3 = (uint8_t) bitextu((unsigned int) *pCom, 2, 6);
+
+    inA0 = ((inA0 + inB0) >> 1);
+    inA1 = ((inA1 + inB1) >> 1);
+    inA2 = ((inA2 + inB2) >> 1);
+    inA3 = ((inA3 + inB3) >> 1);
+
+    uint8_t inA = (uint8_t) bitins(inA0, n_mask2, inA1, mask2, off2);
+    inA = bitins(inA, n_mask4, inA2, mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, inA3, mask6, off6);
+
+    pIn++;
+    pCom++;
+    length--;
+  }
+}
+
+/*
+ * XpulpNN
+ */
+
+static void __attribute__((noinline)) xpulp_nn_zero_mem_u8(uint8_t * pBuffer, unsigned int size)
+{
+  int lfover = size &0x3;
+  for (int i=0; i<(size>>2); i++)
+  {
+    *((v4u *)pBuffer) = (v4u){0,0,0,0};
+    MemoryFence();
+    pBuffer+=4;
+  }
+  while(lfover)
+  {
+    *pBuffer++=0;
+    lfover--;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_zero_mem_u4(uint8_t * pBuffer, unsigned int size)
+{
+  int lfover = size &0x7;
+  for (int i=0; i<(size>>3); i++)
+  {
+    *((v4u *)pBuffer) = (v4u){0,0,0,0};
+    MemoryFence();
+    pBuffer+=4;
+  }
+  while(lfover)
+  {
+    *pBuffer++=0;
+    lfover-=2;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_zero_mem_u2(uint8_t * pBuffer, unsigned int size)
+{
+  int lfover = size &0xf;
+  for (int i=0; i<(size>>4); i++)
+  {
+    *((v4u *)pBuffer) = (v4u){0,0,0,0};
+    MemoryFence();
+    pBuffer+=4;
+  }
+  while(lfover)
+  {
+    *pBuffer++=0;
+    lfover-=4;
+  }
+}
+
+
+static void __attribute__((noinline)) xpulp_nn_compare_and_replace_if_larger_u4(uint8_t * base,
+                                                uint8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    *((uint32_t *)pIn) = max8(*((uint32_t *)pIn), *((uint32_t *)pCom));
+
+    pIn+=4;
+    pCom+=4;
+
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 4, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 4, 4);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 4, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 4, 4);
+
+    if(inA0<inB0)
+      inA0=inB0;
+
+    if(inA1<inB1)
+      inA1=inB1;
+
+    *((uint8_t*)pIn) = bitins(inA0, n_mask, inA1, mask, off);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_compare_and_replace_if_larger_i4(
+    int8_t * base, int8_t * target, uint16_t length) {
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~mask;
+  int8_t off = 0x04;
+
+  int8_t *pIn = base;
+  int8_t *pCom = target;
+
+  int cnt = length >> 2;
+
+  while (cnt > 0u) {
+    *((int32_t *)pIn) = maxs8(*((int32_t *)pIn), *((int32_t *)pCom));
+
+    pIn += 4;
+    pCom += 4;
+
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left > 0u) {
+    int8_t inA0 = (int8_t)bitext((int)*pIn, 4, 0);
+    int8_t inA1 = (int8_t)bitext((int)*pIn, 4, 4);
+    int8_t inB0 = (int8_t)bitext((int)*pCom, 4, 0);
+    int8_t inB1 = (int8_t)bitext((int)*pCom, 4, 4);
+
+    if (inA0 < inB0)
+      inA0 = inB0;
+
+    if (inA1 < inB1)
+      inA1 = inB1;
+
+    *((int8_t *)pIn) = bitins(inA0, n_mask, inA1, mask, off);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_avg_and_replace_u4(uint8_t * base,
+                                  uint8_t * target,
+                                  uint16_t length)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+
+  int cnt = length >> 2;
+
+  while (cnt > 0u)
+  {
+    *((uint32_t *)pIn) = avg8(*((uint32_t *)pIn), *((uint32_t *)pCom));
+
+    pIn+=4;
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+
+  while (left>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 4, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 4, 4);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 4, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 4, 4);
+
+    inA0 = ((inA0 + inB0) >> 1);
+    inA1 = ((inA1 + inB1) >> 1);
+
+    *((uint8_t*)pIn) = bitins(inA0, n_mask, inA1, mask, off);
+
+    pIn++;
+    pCom++;
+    length--;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_compare_and_replace_if_larger_u2(uint8_t * base,
+                                                uint8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask2 = 0x0c;
+  int8_t n_mask2 = ~ mask2;
+  int8_t mask4 = 0x30;
+  int8_t n_mask4 = ~ mask4;
+  int8_t mask6 = 0xc0;
+  int8_t n_mask6 = ~ mask6;
+  int8_t off2 = 2;
+  int8_t off4 = 4;
+  int8_t off6 = 6;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+  uint8_t *out;
+
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    *((uint32_t *)pIn) = max16(*((uint32_t *)pIn), *((uint32_t *)pCom));
+
+    pIn+=4;
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 2, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 2, 2);
+    uint8_t inA2 = (uint8_t) bitextu((unsigned int) *pIn, 2, 4);
+    uint8_t inA3 = (uint8_t) bitextu((unsigned int) *pIn, 2, 6);
+    v4u inA4 = pack((uint8_t) inA0, (uint8_t) inA1, (uint8_t) inA2, (uint8_t) inA3);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 2, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 2, 2);
+    uint8_t inB2 = (uint8_t) bitextu((unsigned int) *pCom, 2, 4);
+    uint8_t inB3 = (uint8_t) bitextu((unsigned int) *pCom, 2, 6);
+    v4u inB4 = pack((uint8_t) inB0, (uint8_t) inB1, (uint8_t) inB2, (uint8_t) inB3);
+
+    *((v4u*)out) = max4(inA4, inB4);
+
+    uint8_t inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_compare_and_replace_if_larger_i2(int8_t * base,
+                                                int8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask2 = 0x0c;
+  int8_t n_mask2 = ~ mask2;
+  int8_t mask4 = 0x30;
+  int8_t n_mask4 = ~ mask4;
+  int8_t mask6 = 0xc0;
+  int8_t n_mask6 = ~ mask6;
+  int8_t off2 = 2;
+  int8_t off4 = 4;
+  int8_t off6 = 6;
+
+  int8_t *pIn = base;
+  int8_t *pCom = target;
+  int8_t *out;
+
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    *((int32_t *)pIn) = maxs16(*((int32_t *)pIn), *((int32_t *)pCom));
+
+    pIn+=4;
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    int8_t inA0 = (int8_t) bitext((int) *pIn, 2, 0);
+    int8_t inA1 = (int8_t) bitext((int) *pIn, 2, 2);
+    int8_t inA2 = (int8_t) bitext((int) *pIn, 2, 4);
+    int8_t inA3 = (int8_t) bitext((int) *pIn, 2, 6);
+    v4s inA4 = pack((int8_t) inA0, (int8_t) inA1, (int8_t) inA2, (int8_t) inA3);
+    int8_t inB0 = (int8_t) bitext((int) *pCom, 2, 0);
+    int8_t inB1 = (int8_t) bitext((int) *pCom, 2, 2);
+    int8_t inB2 = (int8_t) bitext((int) *pCom, 2, 4);
+    int8_t inB3 = (int8_t) bitext((int) *pCom, 2, 6);
+    v4s inB4 = pack((int8_t) inB0, (int8_t) inB1, (int8_t) inB2, (int8_t) inB3);
+
+    *((v4s*)out) = maxs4(inA4, inB4);
+
+    int8_t inA = (int8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((int8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_avg_and_replace_u2(uint8_t * base,
+                                  uint8_t * target,
+                                  uint16_t length)
+{
+  int8_t mask2 = 0x0c;
+  int8_t n_mask2 = ~ mask2;
+  int8_t mask4 = 0x30;
+  int8_t n_mask4 = ~ mask4;
+  int8_t mask6 = 0xc0;
+  int8_t n_mask6 = ~ mask6;
+  int8_t off2 = 2;
+  int8_t off4 = 4;
+  int8_t off6 = 6;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+  uint8_t *out;
+
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    *((uint32_t *)pIn) = avg16(*((uint32_t *)pIn), *((uint32_t *)pCom));
+
+    pIn+=4;
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 2, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 2, 2);
+    uint8_t inA2 = (uint8_t) bitextu((unsigned int) *pIn, 2, 4);
+    uint8_t inA3 = (uint8_t) bitextu((unsigned int) *pIn, 2, 6);
+    v4u inA4 = pack((uint8_t) inA0, (uint8_t) inA1, (uint8_t) inA2, (uint8_t) inA3);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 2, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 2, 2);
+    uint8_t inB2 = (uint8_t) bitextu((unsigned int) *pCom, 2, 4);
+    uint8_t inB3 = (uint8_t) bitextu((unsigned int) *pCom, 2, 6);
+    v4u inB4 = pack((uint8_t) inB0, (uint8_t) inB1, (uint8_t) inB2, (uint8_t) inB3);
+
+    *((v4u*)out) = avg4(inA4, inB4);
+
+    uint8_t inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+
+#endif
diff --git a/rt_nn_tests/xptnn_linear/test.c b/rt_nn_tests/xptnn_linear/test.c
new file mode 100644
index 0000000..51cc048
--- /dev/null
+++ b/rt_nn_tests/xptnn_linear/test.c
@@ -0,0 +1,171 @@
+#include <stdio.h>
+#include <stdint.h>
+#include <stdlib.h>
+
+
+#include "xpulp_tnn_matmul_ternary.h"
+#include "xpulp_tnn_matmul_ternary_4x1.h"
+#include "xpulp_tnn_conv_ternary.h"
+#include "xpulp_tnn_conv_ternary_signed.h"
+#ifndef PROFILE // HACKYTIME
+#include "xpulp_tnn_conv1d_ternary.h"
+#include "xpulp_tnn_conv1d_ternary_signed.h"
+#endif
+#include "xpulp_tnn_maxpool_ternary.h"
+#include "xpulp_tnn_linear_ternary_i32_signed.h"
+#include "xpulp_tnn_linear_ternary_i32.h"
+#include "pmsis.h"
+
+#include "data_statstest.h"
+//#include "pulp_nn_kernels.h"
+#include "pulp_nn_mix_kernels.h"
+
+
+#define start_cycle_counter()                                   asm volatile("csrw 0xCC0, 0x01;")
+#define stop_cycle_counter()                                    asm volatile("csrw 0xCC0, 0x00;")
+#define read_cycle_counter(x)                                   asm volatile("csrr %0, 0x780;" : "=r" (x))
+#define reset_cycle_counter()                                   asm volatile("csrw 0x780, 0x0;")
+
+uint8_t im2col[IM2COL_DIM] = {0};
+uint8_t outputs[OUTPUT_DIM] = {0};
+
+int32_t outputs_fp[OUTPUT_DIM_FP] = {0};
+
+#ifndef PROFILE
+int num_cycles;
+#else
+int im2col_cycles;
+int hotloop_prep_cycles;
+int hotloop_cycles;
+int threshold_cycles;
+int requant_cycles;
+int hotloop_leftover_cycles;
+int matmul4x2_leftover_cycles;
+#endif
+
+void call_krnl_0();
+void test_0();
+
+int main(int argc, char *argv[])
+{
+#if KRAKEN_PTEST == 1
+  kraken_padframe_aon_pad_gpioa_cfg_rxe_set(24, 0);
+  kraken_padframe_aon_pad_gpioa_cfg_trie_set(24, 0);
+  kraken_padframe_aon_pad_gpioa_cfg_chip2pad_set(24, 0);
+#endif
+  int n_mismatches = 0;
+  int mismatches_tot = 0;
+
+    
+
+  #ifndef PROFILE
+  num_cycles = 0;
+  #endif
+
+  if (get_core_id() == 0) {
+  printf("===> TEST 0: Running xpulp_tnn_linear_ternary_i32_signed...\n");
+  printf("  ch_in/out   = [100, 100]\n");
+  //printf("  padding_y_top    = [%d]\n", padding_y_top);
+  //printf("  padding_y_bottom = [%d]\n", padding_y_bottom);
+  //printf("  padding_x_left   = [%d]\n", padding_x_left);
+  //printf("  padding_x_right  = [%d]\n", padding_x_right);
+  //printf("  stride_x         = [%d]\n", stride_x);
+  //printf("  stride_y         = [%d]\n", stride_y);
+  }
+  test_0();
+  #ifndef PROFILE
+  stop_cycle_counter();
+  read_cycle_counter(num_cycles);
+  if (get_core_id() == 0) {
+  printf("===> TEST 0: Finished running xpulp_tnn_linear_ternary_i32_signed\n");
+  printf("num_cycles = %d\n", num_cycles);
+  printf("MACs = 10000\n");
+  printf("MACs/cycle = %.4f\n", 10000/num_cycles);
+  }
+  #endif
+  if (get_core_id() == 0) {
+  printf("Checking for mismatches..\n");
+  n_mismatches = 0;
+
+  for(int i=0; i < 100; i++) {
+    if (outputs_fp[i] != exp_outp_0[i]){
+        printf("***Mismatch in test 0 at iteration %d: Expected: %x, got: %x\n", i, exp_outp_0[i], outputs_fp[i]);
+        n_mismatches++;
+    }
+  }
+  }
+  mismatches_tot += n_mismatches;
+
+  if (get_core_id() == 0) {
+  printf("Got %d mismatches in %d tests\n", mismatches_tot, 1);
+  }
+  return mismatches_tot;
+}
+
+
+void call_krnl_0(void) {
+  uint8_t * pInp;
+  uint8_t * pIm2ColBuffer;
+  int8_t * pBias = NULL;
+  uint8_t * pOut;
+  int8_t * pWeight;
+  uint32_t * pThr;
+  int32_t * pKappa, pLambda;
+  pInp = inp_l1;
+  
+  pOut = outp_l1;
+  pIm2ColBuffer = im2col_l1;
+  pWeight = wt_l1;
+  pThr = threshs_l1;
+  pKappa = kappa_l1;
+  pLambda = lambda_l1;
+#if KRAKEN_PTEST == 1
+  if (pi_core_id() == 0) {
+    kraken_padframe_aon_pad_gpioa_cfg_chip2pad_set(24, 1);
+  } 
+#endif
+  xpulp_tnn_linear_ternary_i32_signed(
+    pInp,
+    pBias,
+    (int32_t *) pOut,
+    pWeight,
+    100,
+    100);
+
+#if KRAKEN_PTEST == 1
+  if (pi_core_id() == 0) {
+    kraken_padframe_aon_pad_gpioa_cfg_chip2pad_set(24, 0);
+  }
+#endif
+}
+
+void test_0(void) {
+  // DMA transfer inputs from L2 to L1
+  if (pi_core_id() == 0) {
+     plp_dma_memcpy(pIn_0, inp_l1, 20, 1);
+     plp_dma_barrier();
+  }
+  pi_cl_team_barrier(0);
+  // transfer weights
+  if (pi_core_id() == 0) {
+     plp_dma_memcpy(pWeight_0, wt_l1, 2000, 1);
+     plp_dma_barrier();
+  }
+  pi_cl_team_barrier(0);
+  call_krnl_0();
+  // get outputs back with DMA
+  if (pi_core_id() == 0) {
+     plp_dma_memcpy(outputs_fp, outp_l1, 400, 0);
+     plp_dma_barrier();
+  }
+}
+
+
+
+
+
+
+
+
+
+
diff --git a/rt_nn_tests/xptnn_linear/xpulp_tnn_conv1d_ternary.h b/rt_nn_tests/xptnn_linear/xpulp_tnn_conv1d_ternary.h
new file mode 100644
index 0000000..f1a08a6
--- /dev/null
+++ b/rt_nn_tests/xptnn_linear/xpulp_tnn_conv1d_ternary.h
@@ -0,0 +1,35 @@
+#ifndef __XPULP_NN_CONV1D_TERNARY_H
+#define __XPULP_NN_CONV1D_TERNARY_H
+#include <stdio.h>
+#include "pulp_nn_utils.h"
+
+
+
+// TODO: review argument order
+void __attribute__((noinline)) xpulp_tnn_conv1d_ternary(
+                                                        uint8_t  *pIn,
+                                                        uint8_t  *pIm2ColBuffer,
+                                                        int8_t   *pBias,
+                                                        uint8_t  *pOut,
+                                                        uint8_t  *pWeight,
+                                                        uint32_t *pThr,
+                                                        uint16_t  dim_in_x,
+                                                        uint16_t  ch_in,
+                                                        uint16_t  dim_out_x,
+                                                        uint16_t  ch_out,
+                                                        uint16_t  dim_kernel_x,
+                                                        uint16_t  padding_x_left,
+                                                        uint16_t  padding_x_right,
+                                                        uint16_t  stride_x,
+#ifndef PROFILE                                         
+                                                        uint16_t  dilation_x);
+#else
+                                                        uint16_t  dilation_x,
+                                                        uint32_t *im2col_cycles,
+                                                        uint32_t *hotloop_prep_cycles,
+                                                        uint32_t *hotloop_cycles,
+                                                        uint32_t *threshold_cycles,
+                                                        uint32_t *hotloop_leftover_cycles,
+                                                        uint32_t *matmul4x2_leftover_cycles);
+#endif
+#endif
diff --git a/rt_nn_tests/xptnn_linear/xpulp_tnn_conv1d_ternary_signed.h b/rt_nn_tests/xptnn_linear/xpulp_tnn_conv1d_ternary_signed.h
new file mode 100644
index 0000000..0f8ba08
--- /dev/null
+++ b/rt_nn_tests/xptnn_linear/xpulp_tnn_conv1d_ternary_signed.h
@@ -0,0 +1,35 @@
+#ifndef __XPULP_NN_CONV1D_TERNARY_SIGNED_H
+#define __XPULP_NN_CONV1D_TERNARY_SIGNED_H
+#include <stdio.h>
+#include "pulp_nn_utils.h"
+
+
+
+// TODO: review argument order
+void __attribute__((noinline)) xpulp_tnn_conv1d_ternary_signed(
+                                                        uint8_t  *pIn,
+                                                        uint8_t  *pIm2ColBuffer,
+                                                        int8_t   *pBias,
+                                                        uint8_t  *pOut,
+                                                        uint8_t  *pWeight,
+                                                        uint32_t *pThr,
+                                                        uint16_t  dim_in_x,
+                                                        uint16_t  ch_in,
+                                                        uint16_t  dim_out_x,
+                                                        uint16_t  ch_out,
+                                                        uint16_t  dim_kernel_x,
+                                                        uint16_t  padding_x_left,
+                                                        uint16_t  padding_x_right,
+                                                        uint16_t  stride_x,
+#ifndef PROFILE                                         
+                                                        uint16_t  dilation_x);
+#else
+                                                        uint16_t  dilation_x,
+                                                        uint32_t *im2col_cycles,
+                                                        uint32_t *hotloop_prep_cycles,
+                                                        uint32_t *hotloop_cycles,
+                                                        uint32_t *threshold_cycles,
+                                                        uint32_t *hotloop_leftover_cycles,
+                                                        uint32_t *matmul4x2_leftover_cycles);
+#endif
+#endif
diff --git a/rt_nn_tests/xptnn_linear/xpulp_tnn_conv_ternary.h b/rt_nn_tests/xptnn_linear/xpulp_tnn_conv_ternary.h
new file mode 100644
index 0000000..f4a8759
--- /dev/null
+++ b/rt_nn_tests/xptnn_linear/xpulp_tnn_conv_ternary.h
@@ -0,0 +1,40 @@
+#ifndef __XPULP_NN_CONV_TERNARY_H
+#define __XPULP_NN_CONV_TERNARY_H
+#include <stdio.h>
+#include "pulp_nn_utils.h"
+
+
+
+// TODO: review argument order
+void __attribute__((noinline)) xpulp_tnn_conv_ternary(
+                                                     uint8_t  *pIn,
+                                                     uint8_t  *pIm2ColBuffer,
+                                                     int8_t   *pBias,
+                                                     uint8_t  *pOut,
+                                                     uint8_t  *pWeight,
+                                                     uint32_t *pThr,
+                                                     uint16_t  dim_in_x,
+                                                     uint16_t  dim_in_y,
+                                                     uint16_t  ch_in,
+                                                     uint16_t  dim_out_x,
+                                                     uint16_t  dim_out_y,
+                                                     uint16_t  ch_out,
+                                                     uint16_t  dim_kernel_x,
+                                                     uint16_t  dim_kernel_y,
+                                                     uint16_t  padding_y_top,
+                                                     uint16_t  padding_y_bottom,
+                                                     uint16_t  padding_x_left,
+                                                     uint16_t  padding_x_right,
+                                                     uint16_t  stride_x,
+#ifndef PROFILE
+                                                     uint16_t  stride_y);
+#else
+                                                     uint16_t  stride_y,
+                                                     uint32_t *im2col_cycles,
+                                                     uint32_t *hotloop_prep_cycles,
+                                                     uint32_t *hotloop_cycles,
+                                                     uint32_t *threshold_cycles,
+                                                     uint32_t *hotloop_leftover_cycles,
+                                                     uint32_t *matmul4x2_leftover_cycles);
+#endif
+#endif
diff --git a/rt_nn_tests/xptnn_linear/xpulp_tnn_conv_ternary_signed.h b/rt_nn_tests/xptnn_linear/xpulp_tnn_conv_ternary_signed.h
new file mode 100644
index 0000000..d05c2ab
--- /dev/null
+++ b/rt_nn_tests/xptnn_linear/xpulp_tnn_conv_ternary_signed.h
@@ -0,0 +1,40 @@
+#ifndef __XPULP_NN_CONV_TERNARY_SIGNED_H
+#define __XPULP_NN_CONV_TERNARY_SIGNED_H
+#include <stdio.h>
+#include "pulp_nn_utils.h"
+
+
+
+// TODO: review argument order
+void __attribute__((noinline)) xpulp_tnn_conv_ternary_signed(
+                                                     uint8_t  *pIn,
+                                                     uint8_t  *pIm2ColBuffer,
+                                                     int8_t   *pBias,
+                                                     uint8_t  *pOut,
+                                                     uint8_t  *pWeight,
+                                                     uint32_t *pThr,
+                                                     uint16_t  dim_in_x,
+                                                     uint16_t  dim_in_y,
+                                                     uint16_t  ch_in,
+                                                     uint16_t  dim_out_x,
+                                                     uint16_t  dim_out_y,
+                                                     uint16_t  ch_out,
+                                                     uint16_t  dim_kernel_x,
+                                                     uint16_t  dim_kernel_y,
+                                                     uint16_t  padding_y_top,
+                                                     uint16_t  padding_y_bottom,
+                                                     uint16_t  padding_x_left,
+                                                     uint16_t  padding_x_right,
+                                                     uint16_t  stride_x,
+#ifndef PROFILE
+                                                     uint16_t  stride_y);
+#else
+                                                     uint16_t  stride_y,
+                                                     uint32_t *im2col_cycles,
+                                                     uint32_t *hotloop_prep_cycles,
+                                                     uint32_t *hotloop_cycles,
+                                                     uint32_t *threshold_cycles,
+                                                     uint32_t *hotloop_leftover_cycles,
+                                                     uint32_t *matmul4x2_leftover_cycles);
+#endif
+#endif
diff --git a/rt_nn_tests/xptnn_linear/xpulp_tnn_linear_ternary_i32.h b/rt_nn_tests/xptnn_linear/xpulp_tnn_linear_ternary_i32.h
new file mode 100644
index 0000000..5c353d3
--- /dev/null
+++ b/rt_nn_tests/xptnn_linear/xpulp_tnn_linear_ternary_i32.h
@@ -0,0 +1,12 @@
+#ifndef __XPULP_TNN_LINEAR_TERNARY_I32_H
+#define __XPULP_TNN_LINEAR_TERNARY_I32_H
+
+void __attribute__((noinline)) xpulp_tnn_linear_ternary_i32(
+                   uint8_t *pIn,
+                   int8_t *pBias,
+                   int8_t *pOut,
+                   int8_t *pWeight,
+                   uint16_t dim_vec,
+                   uint16_t num_o_neurons);
+
+#endif
\ No newline at end of file
diff --git a/rt_nn_tests/xptnn_linear/xpulp_tnn_linear_ternary_i32_signed.c b/rt_nn_tests/xptnn_linear/xpulp_tnn_linear_ternary_i32_signed.c
new file mode 100644
index 0000000..36aeb29
--- /dev/null
+++ b/rt_nn_tests/xptnn_linear/xpulp_tnn_linear_ternary_i32_signed.c
@@ -0,0 +1,95 @@
+#include "pmsis.h"
+#include "pulp_nn_utils.h"
+#include "xpulp_tnn_linear_ternary_i32_signed.h"
+
+void __attribute__((noinline)) xpulp_tnn_linear_ternary_i32_signed(
+                   int8_t *pIn,
+                   int8_t *pBias,
+                   int8_t *pOut,
+                   int8_t *pWeight,
+                   uint16_t dim_vec,
+                   uint16_t num_o_neurons)
+
+{
+  uint16_t dim_vec_in = dim_vec/5; // number of bytes
+  uint16_t dim_vec_wt = dim_vec_in;
+  uint16_t dim_vec_words = dim_vec_in >> 2; // number of words
+
+#ifdef FC_TEST
+  int core_id = 0;
+#else
+  int core_id = pi_core_id();
+#endif
+
+  int Log2Core = log2(NUM_CORES);
+  int chunk = (num_o_neurons >> Log2Core) + ((num_o_neurons & (NUM_CORES-1))!=0);
+  int start = min(chunk * core_id, num_o_neurons);
+  int stop = min(start + chunk, num_o_neurons);
+  int32_t * pOutBuffer = ((int32_t *) pOut) + start;
+
+  for(int i=start; i<stop; i++)
+    {
+      int sum = 0;
+
+      if (pBias != NULL)
+        {
+          sum = *(int32_t *)(pBias + 4*i);
+        }
+
+      uint8_t *pA = pWeight + (i * dim_vec_wt);
+
+      uint8_t *pB = pIn;
+
+      int32_t *ptrA  = (int32_t *) pA;
+
+      int32_t *ptrB  = (int32_t *) pB;
+
+      ptrA = MacLoadInit(1, 0, 0, 0, ptrA);
+      ptrB = MacLoadInit(0, 1, 0, 0, ptrB);
+
+      for (int j=0; j<dim_vec_words; j++)
+      {
+        sum = MacLoads20(1, 0, 0, 0, ptrA, sum);
+        ptrA = MacLoadUpdate(ptrA);
+        ptrB = MacLoadInit(0, 1, 0, 0, ptrB);
+        //ensure enough instructions in the HW loop - otherwise it will work on GVSOC but not in real hardware!
+        asm volatile("nop;");
+      }
+
+      uint16_t col_cnt = dim_vec_in & 0x3;
+
+      if (col_cnt) {
+        int32_t valB;
+        pA += (dim_vec_words << 2);
+        pB += (dim_vec_words << 2);
+        if (col_cnt == 1)
+          {
+            valB = 0xd9 << 24 | 0xd9 << 16 | 0xd9 << 8 | *pB;
+          }
+        else if (col_cnt == 2)
+          {
+            valB = 0xd9 << 24 | 0xd9 << 16 | *(pB+1) << 8 | *pB;
+          }
+        else if (col_cnt == 3)
+          {
+            valB = 0xd9 << 24 | *(pB+2) << 16 | *(pB+1) << 8 | *pB;
+          }
+        // we can leave ptrA and the associated nn-rf as is because we
+        // zero-padded the B input so the "leftover" data in the A word doesn't
+        // matter!
+        uint32_t * pB_p = &valB;
+        pB_p = MacLoadInit(0, 1, 0, 0, pB_p);
+        // try to work around HW bug with misaligned NNRF addresses by initializing NNRF twice...
+        pB_p = &valB;
+        pB_p = MacLoadInit(0, 1, 0, 0, pB_p);
+        sum = MacLoads20(0, 0, 0, 0, pB_p, sum);
+        // work around broken/non-existent forwarding
+        asm volatile("nop;");
+      }
+      *(pOutBuffer++) = sum;
+
+    }
+  #ifndef FC_TEST
+  pi_cl_team_barrier(0);
+  #endif
+}
diff --git a/rt_nn_tests/xptnn_linear/xpulp_tnn_linear_ternary_i32_signed.h b/rt_nn_tests/xptnn_linear/xpulp_tnn_linear_ternary_i32_signed.h
new file mode 100644
index 0000000..218c5c4
--- /dev/null
+++ b/rt_nn_tests/xptnn_linear/xpulp_tnn_linear_ternary_i32_signed.h
@@ -0,0 +1,12 @@
+#ifndef __XPULP_TNN_LINEAR_TERNARY_I32_SIGNED_H
+#define __XPULP_TNN_LINEAR_TERNARY_I32_SIGNED_H
+
+void __attribute__((noinline)) xpulp_tnn_linear_ternary_i32_signed(
+                   int8_t *pIn,
+                   int8_t *pBias,
+                   int8_t *pOut,
+                   int8_t *pWeight,
+                   uint16_t dim_vec,
+                   uint16_t num_o_neurons);
+
+#endif
\ No newline at end of file
diff --git a/rt_nn_tests/xptnn_linear/xpulp_tnn_matmul_ternary.h b/rt_nn_tests/xptnn_linear/xpulp_tnn_matmul_ternary.h
new file mode 100644
index 0000000..914daee
--- /dev/null
+++ b/rt_nn_tests/xptnn_linear/xpulp_tnn_matmul_ternary.h
@@ -0,0 +1,113 @@
+#ifndef __XPULP_TNN_MATMUL_TERNARY_H
+#define __XPULP_TNN_MATMUL_TERNARY_H
+
+#include "pulp_nn_utils.h"
+
+uint8_t * __attribute__((noinline)) xpulp_tnn_matmul_ternary(
+                        uint8_t  *pIn,
+                        int8_t   *pBias,
+                        uint32_t *pThr,
+                        uint8_t  *pOut,
+                        uint8_t  *pOut2,
+                        uint8_t  *pWeight,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+#ifndef PROFILE
+                        uint32_t *thrc_res1,
+                        uint32_t *thrc_res2);
+#else
+                        uint32_t *thrc_res1,
+                        uint32_t *thrc_res2,
+                        uint32_t *hotloop_prep_cycles,
+                        uint32_t *hotloop_cycles,
+                        uint32_t *threshold_cycles,
+                        uint32_t *hotloop_leftover_cycles);
+#endif
+
+inline void __attribute__((aligned(4))) hotloop_4x2(
+  int *sum, int *sum2, int *sum3, int *sum4, int *sum5, int *sum6, int *sum7, int *sum8,
+  uint32_t *ptrA, uint32_t *ptrA2, uint32_t *ptrA3, uint32_t *ptrA4,
+  uint32_t *ptrB, uint32_t *ptrB2, int num_col_im2col_words
+){
+  for (int j=0; j<num_col_im2col_words; j++)
+  {
+    *sum = MacLoad20(0, 0, 0, 0, ptrA, *sum);
+    *sum2 = MacLoad20(0, 0, 1, 0, ptrA2, *sum2);
+    *sum3 = MacLoad20(0, 0, 2, 0, ptrA3, *sum3);
+    *sum4 = MacLoad20(0, 1, 3, 0, ptrB, *sum4);
+    ptrB = MacLoadUpdate(ptrB);
+
+    *sum5 = MacLoad20(1, 0, 0, 1, ptrA, *sum5);
+    ptrA = MacLoadUpdate(ptrA);
+    *sum6 = MacLoad20(1, 0, 1, 1, ptrA2, *sum6);
+    ptrA2 = MacLoadUpdate(ptrA2);
+    *sum7 = MacLoad20(1, 0, 2, 1, ptrA3, *sum7);
+    ptrA3 = MacLoadUpdate(ptrA3);
+    *sum8 = MacLoad20(1, 0, 3, 1, ptrA4, *sum8);
+    ptrA4 = MacLoadUpdate(ptrA4);
+
+    ptrB2 = MacLoadInit(0, 1, 0, 1, ptrB2);
+  }
+}
+
+
+inline void __attribute__((aligned(4))) hotloop_3x2(
+  int *sum, int *sum2, int *sum3, int *sum4, int *sum5, int *sum6,
+  uint32_t *ptrA, uint32_t *ptrA2, uint32_t *ptrA3, uint32_t *ptrB, uint32_t *ptrB2, int num_col_im2col_words
+){
+  for (int j=0; j<num_col_im2col_words; j++)
+  {
+    *sum = MacLoad20(0, 0, 0, 0, ptrA, *sum);
+    *sum2 = MacLoad20(0, 0, 1, 0, ptrA2, *sum2);
+    *sum3 = MacLoad20(0, 1, 2, 0, ptrB, *sum3);
+    ptrB = MacLoadUpdate(ptrB);
+
+    *sum4 = MacLoad20(1, 0, 0, 1, ptrA, *sum4);
+    ptrA = MacLoadUpdate(ptrA);
+    *sum5 = MacLoad20(1, 0, 1, 1, ptrA2, *sum5);
+    ptrA2 = MacLoadUpdate(ptrA2);
+    *sum6 = MacLoad20(1, 0, 2, 1, ptrA3, *sum6);
+    ptrA3 = MacLoadUpdate(ptrA3);
+
+    ptrB2 = MacLoadInit(0, 1, 0, 1, ptrB2);
+  }
+}
+
+
+inline void __attribute__((aligned(4))) hotloop_2x2(
+  int *sum, int *sum2, int *sum3, int *sum4,
+  uint32_t *ptrA, uint32_t *ptrA2, uint32_t *ptrB, uint32_t *ptrB2, int num_col_im2col_words
+){
+  for (int j=0; j<num_col_im2col_words; j++)
+  {
+    *sum = MacLoad20(0, 0, 0, 0, ptrA, *sum);
+    *sum2 = MacLoad20(0, 1, 1, 0, ptrB, *sum2);
+    ptrB = MacLoadUpdate(ptrB);
+
+    *sum3 = MacLoad20(1, 0, 0, 1, ptrA, *sum3);
+    ptrA = MacLoadUpdate(ptrA);
+    *sum4 = MacLoad20(1, 0, 1, 1, ptrA2, *sum4);
+    ptrA2 = MacLoadUpdate(ptrA2);
+
+    ptrB2 = MacLoadInit(0, 1, 0, 1, ptrB2);
+  }
+}
+
+
+inline void __attribute__((aligned(4))) hotloop_1x2(
+  int *sum, int *sum2, uint32_t *ptrA, uint32_t *ptrB, uint32_t *ptrB2, int num_col_im2col_words
+){
+  for (int j=0; j<num_col_im2col_words; j++)
+  {
+    *sum = MacLoad20(0, 1, 0, 0, ptrB, *sum);
+    ptrB = MacLoadUpdate(ptrB);
+    *sum2 = MacLoad20(1, 0, 0, 1, ptrA, *sum2);
+    ptrA = MacLoadUpdate(ptrA);
+
+    ptrB2 = MacLoadInit(0, 1, 0, 1, ptrB2);
+  }
+}
+
+
+
+#endif
\ No newline at end of file
diff --git a/rt_nn_tests/xptnn_linear/xpulp_tnn_matmul_ternary_4x1.h b/rt_nn_tests/xptnn_linear/xpulp_tnn_matmul_ternary_4x1.h
new file mode 100644
index 0000000..fd06e93
--- /dev/null
+++ b/rt_nn_tests/xptnn_linear/xpulp_tnn_matmul_ternary_4x1.h
@@ -0,0 +1,96 @@
+#ifndef __XPULP_TNN_MATMUL_TERNARY_4X1_H
+#define __XPULP_TNN_MATMUL_TERNARY_4X1_H
+
+#include "pulp_nn_utils.h"
+
+uint8_t * __attribute__((noinline)) xpulp_tnn_matmul_ternary_4x1(
+                        uint8_t  *pIn,
+                        int8_t   *pBias,
+                        uint32_t *pThr,
+                        uint8_t  *pOut,
+                        uint8_t  *pWeight,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+#ifndef PROFILE
+                        uint32_t *thrc_res);
+#else
+                        uint32_t *thrc_res,
+                        uint32_t *hotloop_prep_cycles,
+                        uint32_t *hotloop_cycles,
+                        uint32_t *threshold_cycles,
+                        uint32_t *hotloop_leftover_cycles);
+#endif
+
+inline void __attribute__((aligned(4))) hotloop_4x1(
+  int *sum, int *sum2, int *sum3, int *sum4,
+  uint32_t *ptrA, uint32_t *ptrA2, uint32_t *ptrA3, uint32_t *ptrA4,
+  uint32_t *ptrB, int num_col_im2col_words
+){
+  for (int j=0; j<num_col_im2col_words; j++)
+  {
+    *sum = MacLoad20(1, 0, 0, 0, ptrA, *sum);
+    ptrA = MacLoadUpdate(ptrA);
+    *sum2 = MacLoad20(1, 0, 1, 0, ptrA2, *sum2);
+    ptrA2 = MacLoadUpdate(ptrA2);
+    *sum3 = MacLoad20(1, 0, 2, 0, ptrA3, *sum3);
+    ptrA3 = MacLoadUpdate(ptrA3);
+    *sum4 = MacLoad20(1, 0, 3, 0, ptrA4, *sum4);
+    ptrA4 = MacLoadUpdate(ptrA4);
+
+    ptrB = MacLoadInit(0, 1, 0, 0, ptrB);
+    asm volatile ("nop;");
+  }
+}
+
+
+inline void __attribute__((aligned(4))) hotloop_3x1(
+  int *sum, int *sum2, int *sum3,
+  uint32_t *ptrA, uint32_t *ptrA2, uint32_t *ptrA3, uint32_t *ptrB, int num_col_im2col_words
+){
+  for (int j=0; j<num_col_im2col_words; j++)
+  {
+    *sum = MacLoad20(1, 0, 0, 0, ptrA, *sum);
+    ptrA = MacLoadUpdate(ptrA);
+    *sum2 = MacLoad20(1, 0, 1, 0, ptrA2, *sum2);
+    ptrA2 = MacLoadUpdate(ptrA2);
+    *sum3 = MacLoad20(1, 0, 2, 0, ptrA3, *sum3);
+    ptrA3 = MacLoadUpdate(ptrA3);
+
+    ptrB = MacLoadInit(0, 1, 0, 0, ptrB);
+    asm volatile ("nop;");
+  }
+}
+
+
+inline void __attribute__((aligned(4))) hotloop_2x1(
+  int *sum, int *sum2, uint32_t *ptrA, uint32_t *ptrA2, uint32_t *ptrB, int num_col_im2col_words
+){
+  for (int j=0; j<num_col_im2col_words; j++)
+  {
+    *sum = MacLoad20(1, 0, 0, 0, ptrA, *sum);
+    ptrA = MacLoadUpdate(ptrA);
+    *sum2 = MacLoad20(1, 0, 1, 0, ptrA2, *sum2);
+    ptrA2 = MacLoadUpdate(ptrA2);
+
+    ptrB = MacLoadInit(0, 1, 0, 0, ptrB);
+    asm volatile ("nop;");
+  }
+}
+
+
+inline void __attribute__((aligned(4))) hotloop_1x1(
+  int *sum, uint32_t *ptrA, uint32_t *ptrB, int num_col_im2col_words
+){
+  for (int j=0; j<num_col_im2col_words; j++)
+  {
+    *sum = MacLoad20(1, 0, 0, 0, ptrA, *sum);
+    ptrA = MacLoadUpdate(ptrA);
+
+    ptrB = MacLoadInit(0, 1, 0, 0, ptrB);
+    asm volatile ("nop;");
+  }
+}
+
+
+
+#endif
\ No newline at end of file
diff --git a/rt_nn_tests/xptnn_linear/xpulp_tnn_matmul_ternary_signed.h b/rt_nn_tests/xptnn_linear/xpulp_tnn_matmul_ternary_signed.h
new file mode 100644
index 0000000..1493b5d
--- /dev/null
+++ b/rt_nn_tests/xptnn_linear/xpulp_tnn_matmul_ternary_signed.h
@@ -0,0 +1,113 @@
+#ifndef __XPULP_TNN_MATMUL_TERNARY_SIGNED_H
+#define __XPULP_TNN_MATMUL_TERNARY_SIGNED_H
+
+#include "pulp_nn_utils.h"
+
+uint8_t * __attribute__((noinline)) xpulp_tnn_matmul_ternary_signed(
+                        uint8_t  *pIn,
+                        int8_t   *pBias,
+                        uint32_t *pThr,
+                        uint8_t  *pOut,
+                        uint8_t  *pOut2,
+                        uint8_t  *pWeight,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+#ifndef PROFILE
+                        uint32_t *thrc_res1,
+                        uint32_t *thrc_res2);
+#else
+                        uint32_t *thrc_res1,
+                        uint32_t *thrc_res2,
+                        uint32_t *hotloop_prep_cycles,
+                        uint32_t *hotloop_cycles,
+                        uint32_t *threshold_cycles,
+                        uint32_t *hotloop_leftover_cycles);
+#endif
+
+inline void __attribute__((aligned(4))) hotloop_4x2(
+  int *sum, int *sum2, int *sum3, int *sum4, int *sum5, int *sum6, int *sum7, int *sum8,
+  uint32_t *ptrA, uint32_t *ptrA2, uint32_t *ptrA3, uint32_t *ptrA4,
+  uint32_t *ptrB, uint32_t *ptrB2, int num_col_im2col_words
+){
+  for (int j=0; j<num_col_im2col_words; j++)
+  {
+    *sum = MacLoads20(0, 0, 0, 0, ptrA, *sum);
+    *sum2 = MacLoads20(0, 0, 1, 0, ptrA2, *sum2);
+    *sum3 = MacLoads20(0, 0, 2, 0, ptrA3, *sum3);
+    *sum4 = MacLoads20(0, 1, 3, 0, ptrB, *sum4);
+    ptrB = MacLoadUpdate(ptrB);
+
+    *sum5 = MacLoads20(1, 0, 0, 1, ptrA, *sum5);
+    ptrA = MacLoadUpdate(ptrA);
+    *sum6 = MacLoads20(1, 0, 1, 1, ptrA2, *sum6);
+    ptrA2 = MacLoadUpdate(ptrA2);
+    *sum7 = MacLoads20(1, 0, 2, 1, ptrA3, *sum7);
+    ptrA3 = MacLoadUpdate(ptrA3);
+    *sum8 = MacLoads20(1, 0, 3, 1, ptrA4, *sum8);
+    ptrA4 = MacLoadUpdate(ptrA4);
+
+    ptrB2 = MacLoadInit(0, 1, 0, 1, ptrB2);
+  }
+}
+
+
+inline void __attribute__((aligned(4))) hotloop_3x2(
+  int *sum, int *sum2, int *sum3, int *sum4, int *sum5, int *sum6,
+  uint32_t *ptrA, uint32_t *ptrA2, uint32_t *ptrA3, uint32_t *ptrB, uint32_t *ptrB2, int num_col_im2col_words
+){
+  for (int j=0; j<num_col_im2col_words; j++)
+  {
+    *sum = MacLoads20(0, 0, 0, 0, ptrA, *sum);
+    *sum2 = MacLoads20(0, 0, 1, 0, ptrA2, *sum2);
+    *sum3 = MacLoads20(0, 1, 2, 0, ptrB, *sum3);
+    ptrB = MacLoadUpdate(ptrB);
+
+    *sum4 = MacLoads20(1, 0, 0, 1, ptrA, *sum4);
+    ptrA = MacLoadUpdate(ptrA);
+    *sum5 = MacLoads20(1, 0, 1, 1, ptrA2, *sum5);
+    ptrA2 = MacLoadUpdate(ptrA2);
+    *sum6 = MacLoads20(1, 0, 2, 1, ptrA3, *sum6);
+    ptrA3 = MacLoadUpdate(ptrA3);
+
+    ptrB2 = MacLoadInit(0, 1, 0, 1, ptrB2);
+  }
+}
+
+
+inline void __attribute__((aligned(4))) hotloop_2x2(
+  int *sum, int *sum2, int *sum3, int *sum4,
+  uint32_t *ptrA, uint32_t *ptrA2, uint32_t *ptrB, uint32_t *ptrB2, int num_col_im2col_words
+){
+  for (int j=0; j<num_col_im2col_words; j++)
+  {
+    *sum = MacLoads20(0, 0, 0, 0, ptrA, *sum);
+    *sum2 = MacLoads20(0, 1, 1, 0, ptrB, *sum2);
+    ptrB = MacLoadUpdate(ptrB);
+
+    *sum3 = MacLoads20(1, 0, 0, 1, ptrA, *sum3);
+    ptrA = MacLoadUpdate(ptrA);
+    *sum4 = MacLoads20(1, 0, 1, 1, ptrA2, *sum4);
+    ptrA2 = MacLoadUpdate(ptrA2);
+
+    ptrB2 = MacLoadInit(0, 1, 0, 1, ptrB2);
+  }
+}
+
+
+inline void __attribute__((aligned(4))) hotloop_1x2(
+  int *sum, int *sum2, uint32_t *ptrA, uint32_t *ptrB, uint32_t *ptrB2, int num_col_im2col_words
+){
+  for (int j=0; j<num_col_im2col_words; j++)
+  {
+    *sum = MacLoads20(0, 1, 0, 0, ptrB, *sum);
+    ptrB = MacLoadUpdate(ptrB);
+    *sum2 = MacLoads20(1, 0, 0, 1, ptrA, *sum2);
+    ptrA = MacLoadUpdate(ptrA);
+
+    ptrB2 = MacLoadInit(0, 1, 0, 1, ptrB2);
+  }
+}
+
+
+
+#endif
\ No newline at end of file
diff --git a/rt_nn_tests/xptnn_linear/xpulp_tnn_matmul_ternary_signed_4x1.h b/rt_nn_tests/xptnn_linear/xpulp_tnn_matmul_ternary_signed_4x1.h
new file mode 100644
index 0000000..a847bfb
--- /dev/null
+++ b/rt_nn_tests/xptnn_linear/xpulp_tnn_matmul_ternary_signed_4x1.h
@@ -0,0 +1,96 @@
+#ifndef __XPULP_TNN_MATMUL_TERNARY_SIGNED_4X1_H
+#define __XPULP_TNN_MATMUL_TERNARY_SIGNED_4X1_H
+
+#include "pulp_nn_utils.h"
+
+uint8_t * __attribute__((noinline)) xpulp_tnn_matmul_ternary_signed_4x1(
+                        uint8_t  *pIn,
+                        int8_t   *pBias,
+                        uint32_t *pThr,
+                        uint8_t  *pOut,
+                        uint8_t  *pWeight,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+#ifndef PROFILE
+                        uint32_t *thrc_res);
+#else
+                        uint32_t *thrc_res,
+                        uint32_t *hotloop_prep_cycles,
+                        uint32_t *hotloop_cycles,
+                        uint32_t *threshold_cycles,
+                        uint32_t *hotloop_leftover_cycles);
+#endif
+
+inline void __attribute__((aligned(4))) hotloop_4x1(
+  int *sum, int *sum2, int *sum3, int *sum4,
+  uint32_t *ptrA, uint32_t *ptrA2, uint32_t *ptrA3, uint32_t *ptrA4,
+  uint32_t *ptrB, int num_col_im2col_words
+){
+  for (int j=0; j<num_col_im2col_words; j++)
+  {
+    *sum = MacLoads20(1, 0, 0, 0, ptrA, *sum);
+    ptrA = MacLoadUpdate(ptrA);
+    *sum2 = MacLoads20(1, 0, 1, 0, ptrA2, *sum2);
+    ptrA2 = MacLoadUpdate(ptrA2);
+    *sum3 = MacLoads20(1, 0, 2, 0, ptrA3, *sum3);
+    ptrA3 = MacLoadUpdate(ptrA3);
+    *sum4 = MacLoads20(1, 0, 3, 0, ptrA4, *sum4);
+    ptrA4 = MacLoadUpdate(ptrA4);
+
+    ptrB = MacLoadInit(0, 1, 0, 0, ptrB);
+    asm volatile ("nop;");
+  }
+}
+
+
+inline void __attribute__((aligned(4))) hotloop_3x1(
+  int *sum, int *sum2, int *sum3,
+  uint32_t *ptrA, uint32_t *ptrA2, uint32_t *ptrA3, uint32_t *ptrB, int num_col_im2col_words
+){
+  for (int j=0; j<num_col_im2col_words; j++)
+  {
+    *sum = MacLoads20(1, 0, 0, 0, ptrA, *sum);
+    ptrA = MacLoadUpdate(ptrA);
+    *sum2 = MacLoads20(1, 0, 1, 0, ptrA2, *sum2);
+    ptrA2 = MacLoadUpdate(ptrA2);
+    *sum3 = MacLoads20(1, 0, 2, 0, ptrA3, *sum3);
+    ptrA3 = MacLoadUpdate(ptrA3);
+
+    ptrB = MacLoadInit(0, 1, 0, 0, ptrB);
+    asm volatile ("nop;");
+  }
+}
+
+
+inline void __attribute__((aligned(4))) hotloop_2x1(
+  int *sum, int *sum2, uint32_t *ptrA, uint32_t *ptrA2, uint32_t *ptrB, int num_col_im2col_words
+){
+  for (int j=0; j<num_col_im2col_words; j++)
+  {
+    *sum = MacLoads20(1, 0, 0, 0, ptrA, *sum);
+    ptrA = MacLoadUpdate(ptrA);
+    *sum2 = MacLoads20(1, 0, 1, 0, ptrA2, *sum2);
+    ptrA2 = MacLoadUpdate(ptrA2);
+
+    ptrB = MacLoadInit(0, 1, 0, 0, ptrB);
+    asm volatile ("nop;");
+  }
+}
+
+
+inline void __attribute__((aligned(4))) hotloop_1x1(
+  int *sum, uint32_t *ptrA, uint32_t *ptrB, int num_col_im2col_words
+){
+  for (int j=0; j<num_col_im2col_words; j++)
+  {
+    *sum = MacLoads20(1, 0, 0, 0, ptrA, *sum);
+    ptrA = MacLoadUpdate(ptrA);
+
+    ptrB = MacLoadInit(0, 1, 0, 0, ptrB);
+    asm volatile ("nop;");
+  }
+}
+
+
+
+#endif
\ No newline at end of file
diff --git a/rt_nn_tests/xptnn_linear/xpulp_tnn_maxpool_ternary.h b/rt_nn_tests/xptnn_linear/xpulp_tnn_maxpool_ternary.h
new file mode 100644
index 0000000..2b65fca
--- /dev/null
+++ b/rt_nn_tests/xptnn_linear/xpulp_tnn_maxpool_ternary.h
@@ -0,0 +1,20 @@
+#ifndef __XPULP_TNN_MAXPOOL_TERNARY_H
+#define __XPULP_TNN_MAXPOOL_TERNARY_H
+void __attribute__ ((noinline)) xpulp_tnn_maxpool_ternary(
+                                                          int8_t * pIn,
+                                                          int8_t * pOut,
+                                                          uint16_t  dim_im_in_x,
+                                                          uint16_t  dim_im_in_y,
+                                                          // ch_im_in: UNCOMPRESSED number of input/output channels
+                                                          uint16_t  ch_im_in,
+                                                          uint16_t  dim_im_out_x,
+                                                          uint16_t  dim_im_out_y,
+                                                          uint16_t  dim_kernel_x,
+                                                          uint16_t  dim_kernel_y,
+                                                          uint16_t  padding_t,
+                                                          uint16_t  padding_b,
+                                                          uint16_t  padding_l,
+                                                          uint16_t  padding_r,
+                                                          uint16_t  stride_x,
+                                                          uint16_t  stride_y);
+#endif
diff --git a/rt_nn_tests/xptnn_maxpool/Makefile b/rt_nn_tests/xptnn_maxpool/Makefile
new file mode 100644
index 0000000..3925fa5
--- /dev/null
+++ b/rt_nn_tests/xptnn_maxpool/Makefile
@@ -0,0 +1,18 @@
+APP = test
+PULP_APP = test
+PULP_APP_SRCS = test.c
+PULP_APP_SRCS += xpulp_tnn_maxpool_ternary.c
+
+CORE=8
+
+
+PULP_CFLAGS += -DNUM_CORES=$(CORE) -I. -O3
+PULP_LDFLAGS += -lc -lm -lgcc -Wl,-print-memory-usage 
+
+PULP_APP_CFLAGS += -DNUM_CORES=$(CORE) -I. -O3
+PULP_APP_LDFLAGS += -lc -lm -lgcc -Wl,-print-memory-usage 
+
+PULP_CFLAGS += -DPULP_RUNTIME -DSINGLE_CORE_DMA -DALWAYS_BLOCK_DMA_TRANSFERS
+
+
+include $(PULP_SDK_HOME)/install/rules/pulp.mk
diff --git a/rt_nn_tests/xptnn_maxpool/data_statstest.h b/rt_nn_tests/xptnn_maxpool/data_statstest.h
new file mode 100644
index 0000000..c7077af
--- /dev/null
+++ b/rt_nn_tests/xptnn_maxpool/data_statstest.h
@@ -0,0 +1,10270 @@
+// this file is generated automatically by the python script `generate_statstest.py`
+
+#ifndef _DATA_H
+#define _DATA_H
+
+
+
+#define IM2COL_DIM (32 * NUM_CORES * 2) // for NUM_CORES cores
+#define OUTPUT_DIM 2560
+#define OUTPUT_DIM_FP (1)
+
+uint8_t pIn_0 [] = {
+0x77,
+0x6c,
+0xdb,
+0x1b,
+0x1d,
+0x1e,
+0xd5,
+0x36,
+0xf0,
+0xf3,
+0x0b,
+0x7a,
+0xa8,
+0x41,
+0xfe,
+0x0a,
+0x6f,
+0x7c,
+0x8b,
+0x0e,
+0xe1,
+0x67,
+0xc1,
+0xf3,
+0xee,
+0x3a,
+0x51,
+0xb6,
+0x49,
+0x27,
+0x2c,
+0x0a,
+0xaf,
+0xd6,
+0xca,
+0xd0,
+0x71,
+0xc7,
+0x8b,
+0x33,
+0x88,
+0x9f,
+0x34,
+0x70,
+0x25,
+0x73,
+0x9b,
+0xe9,
+0xc4,
+0xc3,
+0x46,
+0x35,
+0xa5,
+0xfa,
+0x3d,
+0x03,
+0xc2,
+0x71,
+0xc0,
+0xee,
+0x00,
+0x60,
+0x0a,
+0xe4,
+0x76,
+0x7c,
+0x5f,
+0xfd,
+0xaa,
+0xa5,
+0xa7,
+0x0f,
+0x7d,
+0x7c,
+0xa6,
+0x97,
+0x44,
+0xdc,
+0x33,
+0x05,
+0x8c,
+0x97,
+0xea,
+0xf1,
+0x60,
+0x8c,
+0x6a,
+0x37,
+0x28,
+0x80,
+0xed,
+0x06,
+0x69,
+0xb1,
+0x42,
+0xc2,
+0x3b,
+0x84,
+0x0d,
+0x3d,
+0x51,
+0xc9,
+0xdc,
+0xe0,
+0x37,
+0x09,
+0x75,
+0xba,
+0x00,
+0xa7,
+0x42,
+0x5b,
+0x01,
+0x49,
+0xe3,
+0x91,
+0x7b,
+0x3d,
+0xec,
+0x58,
+0x2c,
+0x9d,
+0xf5,
+0xed,
+0xa0,
+0xb0,
+0x6f,
+0xbb,
+0x89,
+0x80,
+0x2a,
+0x2d,
+0x16,
+0x7c,
+0x9c,
+0xcd,
+0xad,
+0x34,
+0x0c,
+0x02,
+0x44,
+0x87,
+0x73,
+0x84,
+0x82,
+0xca,
+0xb1,
+0xcc,
+0x38,
+0xa7,
+0xfd,
+0xa3,
+0x4c,
+0xde,
+0xce,
+0xd6,
+0x21,
+0xe5,
+0x91,
+0x7a,
+0x82,
+0x3c,
+0x11,
+0xa9,
+0x0b,
+0xc1,
+0xfc,
+0x33,
+0x3c,
+0xc7,
+0x49,
+0x45,
+0x25,
+0xa3,
+0xab,
+0xeb,
+0xce,
+0x4d,
+0x4e,
+0x9d,
+0x4f,
+0x9d,
+0x9c,
+0xec,
+0x84,
+0x73,
+0x9c,
+0x3a,
+0xfd,
+0x50,
+0x29,
+0x96,
+0x01,
+0xec,
+0xf5,
+0x13,
+0x4f,
+0xc6,
+0x1e,
+0x37,
+0xe7,
+0xbb,
+0x92,
+0xc8,
+0xc3,
+0xed,
+0xc0,
+0xce,
+0x08,
+0xf1,
+0x41,
+0x8f,
+0x06,
+0x61,
+0x44,
+0x0a,
+0xd6,
+0xfb,
+0x03,
+0xc5,
+0x81,
+0xf7,
+0x27,
+0xca,
+0x3a,
+0x7d,
+0x07,
+0x24,
+0x82,
+0xdd,
+0xd8,
+0x30,
+0x5c,
+0x26,
+0xee,
+0xfb,
+0xbf,
+0x98,
+0xf0,
+0xa2,
+0xc7,
+0x51,
+0x95,
+0xb1,
+0x11,
+0xb0,
+0x31,
+0xf8,
+0xa1,
+0x6a,
+0x4b,
+0xf6,
+0xcf,
+0xe5,
+0x02,
+0x13,
+0xc1,
+0x7f,
+0x61,
+0x16,
+0xce,
+0x3a,
+0x24,
+0xa2,
+0xeb,
+0xb0,
+0xb4,
+0x94,
+0x2c,
+0xf1,
+0xa2,
+0xa7,
+0x20,
+0xdf,
+0x74,
+0x8f,
+0xd9,
+0x35,
+0x0e,
+0x11,
+0xff,
+0x2a,
+0xd3,
+0x58,
+0xb1,
+0x78,
+0xcd,
+0xb6,
+0x32,
+0xe8,
+0xc6,
+0x88,
+0xab,
+0x61,
+0x70,
+0x67,
+0x49,
+0xec,
+0x50,
+0xa6,
+0x20,
+0xa2,
+0xbb,
+0x7f,
+0x14,
+0x33,
+0x7c,
+0x2a,
+0xcf,
+0x1e,
+0x84,
+0x5c,
+0x0c,
+0xe8,
+0xf8,
+0x7a,
+0xc7,
+0xe5,
+0x5f,
+0x48,
+0x12,
+0xdc,
+0x00,
+0x11,
+0x44,
+0x09,
+0xca,
+0x34,
+0x2e,
+0x3a,
+0xfe,
+0x94,
+0xf7,
+0xf5,
+0x5a,
+0xff,
+0xce,
+0xfa,
+0xda,
+0x35,
+0x4d,
+0xc4,
+0xf0,
+0x29,
+0xbf,
+0xdb,
+0x4b,
+0xce,
+0xca,
+0xa3,
+0x97,
+0xdf,
+0xac,
+0x90,
+0x21,
+0x8d,
+0xf6,
+0x32,
+0x6b,
+0x20,
+0xd7,
+0x29,
+0xb7,
+0x18,
+0x7d,
+0x04,
+0xf8,
+0x27,
+0xb8,
+0x67,
+0x8d,
+0xdf,
+0x38,
+0xb6,
+0x8c,
+0x40,
+0xf5,
+0x34,
+0x1d,
+0x5a,
+0x0f,
+0x1a,
+0x91,
+0xb7,
+0x40,
+0x18,
+0x74,
+0x33,
+0x87,
+0xcc,
+0xb7,
+0xda,
+0xa6,
+0x86,
+0x15,
+0x49,
+0x70,
+0x84,
+0xb6,
+0x81,
+0x00,
+0xe4,
+0x39,
+0x01,
+0x3e,
+0xcd,
+0x11,
+0xc4,
+0x1b,
+0xbf,
+0x0d,
+0x39,
+0xa3,
+0x25,
+0x22,
+0x82,
+0x4c,
+0x58,
+0xd7,
+0x1b,
+0x53,
+0x5a,
+0xbd,
+0x62,
+0xa4,
+0xe9,
+0xd4,
+0x62,
+0x6a,
+0xed,
+0x54,
+0x0c,
+0x7f,
+0x9f,
+0x36,
+0x5e,
+0x1e,
+0x07,
+0x98,
+0x80,
+0xc2,
+0xa6,
+0x88,
+0xf4,
+0x88,
+0x51,
+0xd7,
+0x80,
+0x2d,
+0x44,
+0x03,
+0x24,
+0xc1,
+0x32,
+0x6f,
+0xea,
+0xaa,
+0x8f,
+0xfe,
+0xbc,
+0x26,
+0x0f,
+0x5e,
+0xfb,
+0x30,
+0x3e,
+0xab,
+0xa8,
+0xe4,
+0x56,
+0xe4,
+0xc6,
+0xf5,
+0x90,
+0x51,
+0x34,
+0x73,
+0xa8,
+0x87,
+0xf7,
+0x45,
+0x4e,
+0xc9,
+0x83,
+0x38,
+0x5b,
+0xd4,
+0x67,
+0xea,
+0xa9,
+0xc7,
+0x26,
+0x9d,
+0x75,
+0xba,
+0x98,
+0xe5,
+0x5f,
+0xdd,
+0x5d,
+0x42,
+0x2a,
+0x47,
+0xe0,
+0x63,
+0x6a,
+0x0e,
+0x7a,
+0xa0,
+0xaa,
+0xe4,
+0x7b,
+0x04,
+0x46,
+0x6e,
+0xaf,
+0x96,
+0x41,
+0x1c,
+0x94,
+0x4d,
+0xab,
+0x09,
+0x6f,
+0xc5,
+0x73,
+0x37,
+0x42,
+0x0b,
+0x9f,
+0x9b,
+0xcc,
+0x07,
+0x1f,
+0x27,
+0x28,
+0xfc,
+0xa5,
+0x13,
+0x07,
+0x5a,
+0x12,
+0xb3,
+0x75,
+0xe2,
+0x5e,
+0x97,
+0xb2,
+0x0e,
+0x65,
+0x29,
+0x21,
+0x5f,
+0x21,
+0xea,
+0x73,
+0xa0,
+0x01,
+0x3b,
+0x34,
+0x04,
+0xdf,
+0xaa,
+0x53,
+0x07,
+0x56,
+0xea,
+0xf6,
+0x48,
+0x3d,
+0xc3,
+0x9b,
+0xfd,
+0x4e,
+0x7e,
+0x3c,
+0x13,
+0x54,
+0xbb,
+0xb4,
+0x44,
+0x7c,
+0x2f,
+0x7d,
+0x43,
+0x5a,
+0x31,
+0x40,
+0x5a,
+0xf4,
+0x40,
+0x5e,
+0x95,
+0x66,
+0x2b,
+0x88,
+0xd6,
+0x71,
+0x46,
+0x33,
+0x36,
+0x45,
+0x29,
+0x62,
+0x31,
+0x57,
+0x74,
+0x1c,
+0x50,
+0xc7,
+0x85,
+0x21,
+0xd6,
+0x34,
+0x4c,
+0x48,
+0xef,
+0xc5,
+0x50,
+0xa2,
+0x44,
+0x67,
+0x15,
+0x7d,
+0x2f,
+0x33,
+0xe6,
+0x64,
+0x3b,
+0x4b,
+0xf1,
+0xfe,
+0x39,
+0xd7,
+0xf0,
+0x22,
+0x15,
+0x3f,
+0x0d,
+0x9a,
+0xe1,
+0xcc,
+0xb1,
+0xd1,
+0xeb,
+0x04,
+0x95,
+0xa0,
+0x51,
+0xd7,
+0x70,
+0x73,
+0xf3,
+0x8f,
+0xa3,
+0x21,
+0xdb,
+0x2e,
+0xee,
+0x20,
+0x60,
+0x1d,
+0x09,
+0x86,
+0xeb,
+0x33,
+0xa9,
+0xba,
+0xcb,
+0x7a,
+0xea,
+0xd5,
+0xfc,
+0xdc,
+0x24,
+0xc9,
+0xe0,
+0xe0,
+0xc5,
+0x69,
+0xc4,
+0x8b,
+0x4a,
+0x11,
+0x6b,
+0x76,
+0xfc,
+0x69,
+0xe6,
+0x77,
+0xe5,
+0x4a,
+0x0f,
+0xbd,
+0x90,
+0xb6,
+0xc6,
+0x8c,
+0x5e,
+0x45,
+0xc1,
+0x66,
+0xd1,
+0xb3,
+0x07,
+0x6c,
+0x6e,
+0x5b,
+0xd9,
+0x0e,
+0x01,
+0xf6,
+0xe8,
+0x96,
+0x7c,
+0xbc,
+0x27,
+0xf4,
+0x77,
+0x49,
+0x30,
+0xdd,
+0x00,
+0xac,
+0xac,
+0x00,
+0x43,
+0x4c,
+0x15,
+0x8b,
+0x24,
+0x3d,
+0xc2,
+0x0d,
+0xb5,
+0xb5,
+0xad,
+0x8a,
+0x7e,
+0x48,
+0xf0,
+0xfa,
+0xe6,
+0xa9,
+0x61,
+0xb3,
+0x73,
+0xf6,
+0xb2,
+0x2a,
+0x61,
+0x5f,
+0x3a,
+0x84,
+0xd8,
+0xa2,
+0x39,
+0x31,
+0xef,
+0x81,
+0x2d,
+0xfa,
+0x27,
+0x01,
+0x10,
+0x7d,
+0x40,
+0x98,
+0xfa,
+0xdb,
+0xfa,
+0x54,
+0xfd,
+0x33,
+0xf0,
+0x2c,
+0xd7,
+0xab,
+0xeb,
+0xa0,
+0x67,
+0x3f,
+0x44,
+0x06,
+0x34,
+0x1b,
+0xac,
+0xa8,
+0xe5,
+0x1c,
+0xf8,
+0x91,
+0x47,
+0x54,
+0xc2,
+0xc8,
+0x6d,
+0x95,
+0xdd,
+0x4e,
+0xe4,
+0x7f,
+0xc1,
+0xec,
+0x84,
+0x9b,
+0xf6,
+0x73,
+0x2d,
+0x92,
+0xbc,
+0x54,
+0x62,
+0x37,
+0x2a,
+0x44,
+0x6f,
+0xd7,
+0x38,
+0x51,
+0x66,
+0x1c,
+0x56,
+0x4b,
+0xfc,
+0x73,
+0x8d,
+0x4e,
+0xca,
+0x5f,
+0xb6,
+0x00,
+0x86,
+0x0a,
+0x49,
+0x33,
+0x3d,
+0x5a,
+0x98,
+0x73,
+0xc6,
+0xcf,
+0x1f,
+0x6d,
+0x2b,
+0x8f,
+0xc9,
+0x89,
+0x9a,
+0xf0,
+0x48,
+0x41,
+0x3c,
+0x73,
+0x49,
+0xa5,
+0x2e,
+0xa4,
+0x12,
+0xc3,
+0x7c,
+0xd9,
+0x76,
+0x21,
+0x85,
+0xfa,
+0x0d,
+0x2c,
+0x9a,
+0x9a,
+0xa3,
+0x29,
+0x8b,
+0x6b,
+0xaa,
+0x4f,
+0x87,
+0x49,
+0x8d,
+0xce,
+0x74,
+0xc1,
+0x85,
+0xdc,
+0x04,
+0xeb,
+0x73,
+0x6f,
+0x69,
+0xe5,
+0x15,
+0x51,
+0x01,
+0x78,
+0x62,
+0xa8,
+0xa9,
+0x94,
+0x5c,
+0x83,
+0xf8,
+0x44,
+0x6f,
+0x27,
+0x64,
+0xe4,
+0xfd,
+0x24,
+0x09,
+0x07,
+0x60,
+0x9d,
+0xfa,
+0x3f,
+0x2d,
+0xe3,
+0x51,
+0xea,
+0x18,
+0xc4,
+0xa6,
+0x3e,
+0x0a,
+0x57,
+0x38,
+0x6e,
+0x6c,
+0xb6,
+0x97,
+0x5b,
+0xe7,
+0xc7,
+0x6b,
+0x7a,
+0xe8,
+0xe5,
+0x69,
+0x85,
+0x62,
+0xdf,
+0x01,
+0x37,
+0xad,
+0x48,
+0xd8,
+0x66,
+0x55,
+0x28,
+0xc6,
+0x63,
+0x26,
+0x88,
+0x22,
+0x8b,
+0x2b,
+0xe2,
+0xd9,
+0x24,
+0x13,
+0x9f,
+0xb3,
+0xfb,
+0x2d,
+0x96,
+0x44,
+0x75,
+0x9c,
+0x1a,
+0x7c,
+0x55,
+0x98,
+0xce,
+0x34,
+0xaa,
+0x63,
+0x77,
+0xc6,
+0x21,
+0xe4,
+0xcf,
+0x9f,
+0x7f,
+0xa3,
+0x13,
+0x2b,
+0xc4,
+0xa9,
+0xc2,
+0xc2,
+0x29,
+0x26,
+0x41,
+0xc4,
+0x40,
+0x74,
+0xb7,
+0x91,
+0x36,
+0x23,
+0x91,
+0xce,
+0xa2,
+0xc3,
+0x2c,
+0x76,
+0x87,
+0x4e,
+0xb5,
+0x3e,
+0xca,
+0xc3,
+0x67,
+0xc0,
+0x60,
+0xb4,
+0xea,
+0x39,
+0x90,
+0x81,
+0x23,
+0xe6,
+0xc2,
+0xfa,
+0x64,
+0x29,
+0x29,
+0x88,
+0xe9,
+0x1a,
+0xba,
+0x68,
+0x15,
+0x3c,
+0xc5,
+0x3c,
+0x50,
+0x64,
+0x83,
+0x84,
+0xe2,
+0xab,
+0x83,
+0xa6,
+0xee,
+0x93,
+0x46,
+0x4a,
+0xe7,
+0x37,
+0x70,
+0xc2,
+0x0b,
+0x53,
+0x5d,
+0x3a,
+0x67,
+0x84,
+0x65,
+0x97,
+0xc4,
+0x70,
+0x9f,
+0x34,
+0x08,
+0x04,
+0x11,
+0x34,
+0xad,
+0xdb,
+0xe2,
+0x31,
+0x1c,
+0x50,
+0x68,
+0xe8,
+0x75,
+0x14,
+0x8d,
+0x0a,
+0x17,
+0x9b,
+0xf3,
+0x4d,
+0x20,
+0xa6,
+0xe1,
+0x4a,
+0x0a,
+0xa5,
+0x66,
+0xe2,
+0xe8,
+0xe5,
+0x5e,
+0x62,
+0x2d,
+0xa1,
+0x1f,
+0x69,
+0xac,
+0xf5,
+0x42,
+0x80,
+0xa0,
+0xde,
+0x54,
+0xd6,
+0xd6,
+0x8d,
+0x80,
+0xfc,
+0x0a,
+0x4a,
+0x33,
+0x20,
+0x81,
+0x2f,
+0xb0,
+0xaa,
+0x3f,
+0x92,
+0x36,
+0xe6,
+0xfe,
+0x30,
+0xcc,
+0x92,
+0x77,
+0x37,
+0x00,
+0x4a,
+0x2c,
+0x40,
+0x45,
+0xa3,
+0x3b,
+0x7f,
+0x88,
+0x65,
+0x73,
+0x82,
+0x12,
+0x16,
+0x8a,
+0xff,
+0x6c,
+0x87,
+0xb2,
+0xfd,
+0x89,
+0xa2,
+0xf6,
+0x0a,
+0x69,
+0x67,
+0xd4,
+0xf8,
+0x20,
+0x93,
+0x5d,
+0x57,
+0x39,
+0x92,
+0xdd,
+0xc0,
+0xc5,
+0xdf,
+0x74,
+0x71,
+0x7f,
+0xfe,
+0x92,
+0x9b,
+0x5c,
+0xc0,
+0x22,
+0x4f,
+0x54,
+0xb2,
+0xe9,
+0x21,
+0x64,
+0x4d,
+0xe6,
+0x0a,
+0x66,
+0x35,
+0x73,
+0x5a,
+0x7f,
+0xf8,
+0x67,
+0x3b,
+0x12,
+0xa0,
+0xa1,
+0x62,
+0x18,
+0xfb,
+0xbd,
+0xbd,
+0xef,
+0xa2,
+0x00,
+0x1d,
+0xbf,
+0x0a,
+0x01,
+0xbf,
+0xe0,
+0xba,
+0xe0,
+0x87,
+0xe2,
+0xb3,
+0x66,
+0xaa,
+0x2c,
+0x35,
+0xc9,
+0x3d,
+0x4d,
+0x9d,
+0xa8,
+0x6e,
+0x13,
+0x3c,
+0xff,
+0xe5,
+0xd6,
+0x3d,
+0xc0,
+0x7b,
+0xda,
+0x80,
+0x73,
+0xe2,
+0x30,
+0x24,
+0x3b,
+0xc7,
+0x13,
+0xe4,
+0x34,
+0xb6,
+0x75,
+0xfc,
+0x0d,
+0x09,
+0x7f,
+0x5b,
+0xce,
+0x26,
+0x61,
+0xa0,
+0xeb,
+0x94,
+0x02,
+0x1c,
+0xfd,
+0x3e,
+0x7f,
+0x00,
+0xf6,
+0xe3,
+0x8a,
+0xc4,
+0x6a,
+0x01,
+0xee,
+0x93,
+0x10,
+0x2d,
+0x73,
+0x34,
+0x1e,
+0x60,
+0x30,
+0xee,
+0xf3,
+0x3b,
+0x98,
+0x0d,
+0x60,
+0xb6,
+0x83,
+0xfe,
+0xe7,
+0x28,
+0x88,
+0x34,
+0x26,
+0x39,
+0x8d,
+0xd6,
+0x8d,
+0x7e,
+0x13,
+0x65,
+0xd4,
+0xd4,
+0x3d,
+0xa6,
+0x00,
+0xb6,
+0x35,
+0x2c,
+0xe9,
+0x90,
+0xad,
+0x50,
+0x69,
+0xe4,
+0xa6,
+0x29,
+0xc0,
+0x5b,
+0xed,
+0x07,
+0xc8,
+0x3a,
+0x9b,
+0xad,
+0xc5,
+0xea,
+0x39,
+0x56,
+0xfd,
+0xb9,
+0x4f,
+0x6f,
+0x35,
+0x90,
+0xdc,
+0x07,
+0x6b,
+0x57,
+0xef,
+0x6d,
+0x3c,
+0xa2,
+0x05,
+0x75,
+0x8f,
+0x7b,
+0x2b,
+0x48,
+0xcd,
+0x3e,
+0x5f,
+0x8f,
+0x96,
+0x14,
+0x17,
+0x6d,
+0xd6,
+0x46,
+0xb7,
+0x83,
+0x41,
+0x54,
+0x85,
+0xec,
+0x70,
+0x81,
+0x64,
+0xd6,
+0x0b,
+0x00,
+0xd1,
+0x0c,
+0x6f,
+0x1b,
+0x2a,
+0xc2,
+0xb6,
+0x11,
+0x5f,
+0x49,
+0xfa,
+0x35,
+0xa3,
+0x4c,
+0x67,
+0x78,
+0x24,
+0x18,
+0x4d,
+0x0b,
+0xe3,
+0x2c,
+0x24,
+0x6a,
+0x0b,
+0x37,
+0x00,
+0x3e,
+0x5c,
+0x14,
+0x5d,
+0x0a,
+0x91,
+0x78,
+0x57,
+0xd9,
+0x53,
+0x3a,
+0x61,
+0x93,
+0xe6,
+0xcc,
+0x27,
+0x09,
+0xee,
+0x27,
+0xd8,
+0x83,
+0xa1,
+0x8c,
+0x5c,
+0xc2,
+0x4a,
+0x63,
+0xcb,
+0xe3,
+0x08,
+0x21,
+0x7c,
+0x9d,
+0x4d,
+0x87,
+0x3a,
+0xe8,
+0x57,
+0xf5,
+0x1d,
+0x8f,
+0x3a,
+0xed,
+0xce,
+0xe9,
+0xcc,
+0xc7,
+0x39,
+0xca,
+0x1b,
+0xc6,
+0xea,
+0xf1,
+0x1b,
+0x65,
+0xa1,
+0x62,
+0x2c,
+0xa2,
+0x49,
+0xcc,
+0x4e,
+0xee,
+0xe1,
+0x6c,
+0x57,
+0x5e,
+0xd3,
+0x63,
+0x57,
+0xdc,
+0x86,
+0x03,
+0x66,
+0xcd,
+0x34,
+0x74,
+0xa9,
+0x6c,
+0xc5,
+0x43,
+0x4a,
+0x75,
+0xe6,
+0xf3,
+0x85,
+0x78,
+0xf1,
+0xc3,
+0xc1,
+0x1e,
+0xc8,
+0xc9,
+0x7e,
+0x57,
+0x56,
+0x3a,
+0xe2,
+0xfa,
+0xb9,
+0xa6,
+0x69,
+0x70,
+0xf0,
+0xe8,
+0x45,
+0xd3,
+0xb3,
+0xc2,
+0x08,
+0x63,
+0x34,
+0x23,
+0xf8,
+0x87,
+0x9d,
+0x49,
+0x39,
+0x58,
+0x53,
+0xa7,
+0xaf,
+0xc6,
+0xec,
+0xba,
+0x21,
+0x03,
+0x86,
+0x0d,
+0xd5,
+0xc2,
+0x80,
+0x56,
+0x83,
+0x58,
+0x8a,
+0x7b,
+0xba,
+0xe4,
+0x49,
+0xa7,
+0x91,
+0x31,
+0x77,
+0x7f,
+0x70,
+0x49,
+0x2f,
+0x65,
+0x25,
+0xd3,
+0xc6,
+0x6f,
+0xf6,
+0x5d,
+0xa9,
+0x6c,
+0x07,
+0xfa,
+0xba,
+0xd0,
+0xb8,
+0x58,
+0x78,
+0xdb,
+0x30,
+0xd8,
+0xf3,
+0xea,
+0xc6,
+0x90,
+0x6d,
+0xda,
+0x25,
+0x97,
+0xcc,
+0x4d,
+0xaa,
+0x83,
+0xd6,
+0x90,
+0xe0,
+0x70,
+0x53,
+0x06,
+0xce,
+0xb9,
+0xce,
+0xfc,
+0x2e,
+0x27,
+0x3e,
+0xc5,
+0xfd,
+0x8a,
+0xba,
+0xfa,
+0x5f,
+0x3f,
+0x23,
+0x85,
+0xdb,
+0x3d,
+0x0d,
+0x71,
+0xb7,
+0x48,
+0x47,
+0xe9,
+0x0f,
+0x51,
+0x0e,
+0x86,
+0xab,
+0xf0,
+0xac,
+0xe1,
+0x03,
+0xdd,
+0x2b,
+0x68,
+0x7a,
+0xee,
+0x4f,
+0x85,
+0xf0,
+0x45,
+0x5e,
+0x13,
+0x98,
+0x0f,
+0x32,
+0xe6,
+0xa7,
+0x8a,
+0x8a,
+0xe3,
+0x31,
+0x3e,
+0x60,
+0x38,
+0x09,
+0xc4,
+0x1e,
+0x91,
+0x54,
+0xd8,
+0x53,
+0x6e,
+0xe2,
+0xad,
+0xfa,
+0x62,
+0x0f,
+0xb8,
+0x6e,
+0x75,
+0x6e,
+0x9f,
+0xd8,
+0x2b,
+0x9d,
+0x05,
+0x5b,
+0x49,
+0xe7,
+0x94,
+0x91,
+0x6b,
+0xaa,
+0x58,
+0x0c,
+0xeb,
+0xe5,
+0x0e,
+0xfe,
+0x00,
+0xe0,
+0x22,
+0x4b,
+0x46,
+0x40,
+0xde,
+0x80,
+0x76,
+0xa5,
+0xa7,
+0x6d,
+0x36,
+0xd9,
+0x3f,
+0x43,
+0x26,
+0x61,
+0x20,
+0xc4,
+0xc6,
+0x62,
+0x96,
+0x03,
+0xc3,
+0x31,
+0x9f,
+0x5d,
+0xd3,
+0xa6,
+0xf8,
+0x7c,
+0xe1,
+0x57,
+0xad,
+0x64,
+0x0b,
+0x7d,
+0x58,
+0xd5,
+0xac,
+0x76,
+0xf5,
+0xc0,
+0x5b,
+0x65,
+0x8f,
+0x21,
+0xca,
+0xe2,
+0x5c,
+0x2d,
+0x3e,
+0xb9,
+0xb3,
+0x3e,
+0x28,
+0x14,
+0x24,
+0x41,
+0xb4,
+0x35,
+0x4e,
+0x29,
+0x9c,
+0x48,
+0x40,
+0xa9,
+0xba,
+0xd6,
+0x2f,
+0x3c,
+0x25,
+0xe8,
+0xdd,
+0xab,
+0x55,
+0xa5,
+0xe5,
+0xe5,
+0xaa,
+0x95,
+0xd8,
+0x92,
+0x12,
+0x34,
+0x2f,
+0x8a,
+0x4e,
+0x3c,
+0xd7,
+0x78,
+0xda,
+0xf7,
+0xc6,
+0x6b,
+0x5f,
+0x88,
+0xe2,
+0xe7,
+0x3e,
+0x7f,
+0x4a,
+0xa1,
+0x9d,
+0xe7,
+0x8a,
+0x73,
+0xa0,
+0xde,
+0xa0,
+0x91,
+0x3f,
+0xe1,
+0x90,
+0xda,
+0x77,
+0xaf,
+0x34,
+0xba,
+0x73,
+0x26,
+0x4f,
+0x9d,
+0xfe,
+0x61,
+0xc4,
+0x2f,
+0xcb,
+0x80,
+0x35,
+0xe0,
+0xa2,
+0x7c,
+0xfd,
+0xf4,
+0x2e,
+0x2c,
+0x9d,
+0x60,
+0x2b,
+0xd6,
+0x15,
+0x8b,
+0x57,
+0xad,
+0x8f,
+0x7e,
+0x83,
+0x3b,
+0xde,
+0xa1,
+0x2f,
+0x15,
+0xe7,
+0xa7,
+0x58,
+0x15,
+0xc7,
+0xc5,
+0x27,
+0x55,
+0x5e,
+0xad,
+0x2d,
+0x4c,
+0xeb,
+0x66,
+0xa9,
+0xec,
+0x89,
+0x13,
+0x71,
+0x24,
+0x12,
+0xa5,
+0x7d,
+0xb6,
+0x67,
+0x80,
+0x76,
+0x94,
+0x22,
+0x17,
+0x37,
+0xd7,
+0x9a,
+0x25,
+0xdb,
+0x25,
+0xb4,
+0x0b,
+0x96,
+0xb9,
+0x22,
+0xa8,
+0x47,
+0x48,
+0x86,
+0x6e,
+0xd5,
+0x42,
+0x0c,
+0x8b,
+0xfa,
+0xb6,
+0x84,
+0xd0,
+0x24,
+0x1c,
+0x64,
+0x38,
+0x1e,
+0xcd,
+0x4a,
+0x3c,
+0xdf,
+0x6c,
+0x34,
+0xce,
+0xfb,
+0x54,
+0x28,
+0xeb,
+0x21,
+0x80,
+0x2c,
+0xaa,
+0xa5,
+0x0c,
+0x3e,
+0x87,
+0xdc,
+0x6b,
+0xa2,
+0x73,
+0xc8,
+0xca,
+0x09,
+0x3c,
+0x51,
+0x89,
+0x31,
+0x8f,
+0x95,
+0x77,
+0xd5,
+0x56,
+0xd7,
+0x6c,
+0x06,
+0x4c,
+0x6c,
+0x84,
+0xcd,
+0x1f,
+0x47,
+0x53,
+0x45,
+0x77,
+0xff,
+0xe5,
+0x03,
+0x6d,
+0xea,
+0xfa,
+0xe5,
+0x8a,
+0xc6,
+0xf3,
+0x4a,
+0xd6,
+0x88,
+0xcd,
+0x80,
+0xee,
+0x28,
+0x1f,
+0x92,
+0x64,
+0x41,
+0xd6,
+0x4f,
+0x9d,
+0xe2,
+0x55,
+0xee,
+0x1b,
+0x10,
+0xa7,
+0x61,
+0x61,
+0x87,
+0xf5,
+0x3b,
+0x5f,
+0x55,
+0xf7,
+0xb1,
+0xe8,
+0xc9,
+0x4a,
+0x9d,
+0x73,
+0x05,
+0x97,
+0x1c,
+0x50,
+0x29,
+0xcc,
+0x97,
+0xa6,
+0xa2,
+0xf3,
+0x1d,
+0xe8,
+0x04,
+0x18,
+0xc3,
+0xdf,
+0x73,
+0x92,
+0xa0,
+0x68,
+0x73,
+0x91,
+0xff,
+0xd9,
+0x0d,
+0x93,
+0x1b,
+0x40,
+0x39,
+0x1e,
+0x85,
+0x1a,
+0xc7,
+0x03,
+0xf3,
+0x70,
+0x44,
+0x66,
+0x0c,
+0xf6,
+0x9c,
+0xb6,
+0x47,
+0x43,
+0x8a,
+0x36,
+0x04,
+0x43,
+0x1b,
+0xe2,
+0x5a,
+0x44,
+0xb6,
+0x54,
+0xcf,
+0x05,
+0x53,
+0x2d,
+0xe1,
+0x8f,
+0x66,
+0xc8,
+0xf4,
+0xa1,
+0xfb,
+0xca,
+0x8a,
+0x0b,
+0xbb,
+0x37,
+0x12,
+0x18,
+0x34,
+0xdb,
+0x45,
+0xd6,
+0x5b,
+0x43,
+0x4b,
+0x2a,
+0xc1,
+0x4a,
+0x58,
+0x0a,
+0x23,
+0x51,
+0xee,
+0x9a,
+0x49,
+0xa4,
+0xc4,
+0x51,
+0x9b,
+0xe2,
+0xba,
+0x95,
+0x6a,
+0x41,
+0x93,
+0xf6,
+0x88,
+0xed,
+0x14,
+0xbd,
+0xaa,
+0xde,
+0x77,
+0x39,
+0x90,
+0x88,
+0xed,
+0x5d,
+0xcc,
+0x23,
+0x23,
+0x27,
+0x5b,
+0x4e,
+0x9a,
+0x6e,
+0xa7,
+0x01,
+0xfe,
+0x5c,
+0x12,
+0xc0,
+0xcc,
+0x0b,
+0x92,
+0x6e,
+0x87,
+0x4f,
+0x6b,
+0x90,
+0xcc,
+0x4e,
+0x1a,
+0xc4,
+0xe8,
+0xac,
+0x2b,
+0x78,
+0xa8,
+0x8a,
+0xd9,
+0x36,
+0xc6,
+0xcc,
+0x82,
+0x31,
+0x51,
+0x62,
+0x1c,
+0x4d,
+0xad,
+0x2d,
+0xd6,
+0x1c,
+0x7e,
+0x80,
+0x8f,
+0x4e,
+0x9d,
+0x02,
+0xbc,
+0x92,
+0x8b,
+0xb2,
+0x3b,
+0x0e,
+0x9a,
+0xc4,
+0xf3,
+0x15,
+0x75,
+0x09,
+0xd3,
+0x1c,
+0xbd,
+0x5a,
+0xf8,
+0xb2,
+0x5e,
+0x9f,
+0xbf,
+0xcd,
+0xf7,
+0x5a,
+0x12,
+0xc9,
+0xe0,
+0x02,
+0x64,
+0x24,
+0x1b,
+0xa2,
+0xdd,
+0x96,
+0x5d,
+0x6c,
+0x66,
+0xb4,
+0x57,
+0x65,
+0x15,
+0x82,
+0x26,
+0xec,
+0x2d,
+0x86,
+0x46,
+0xe2,
+0xcf,
+0x3c,
+0x0a,
+0xcf,
+0xdb,
+0x54,
+0x87,
+0x8a,
+0xb8,
+0x76,
+0x32,
+0xe8,
+0x6c,
+0x45,
+0xbb,
+0x34,
+0x41,
+0x65,
+0x17,
+0xf3,
+0x6c,
+0xb9,
+0x86,
+0xa6,
+0xb2,
+0x05,
+0xd7,
+0x45,
+0x4b,
+0x07,
+0x13,
+0x1c,
+0xb5,
+0x75,
+0xf3,
+0xf5,
+0xbb,
+0x23,
+0xf4,
+0x51,
+0x0e,
+0xf0,
+0x68,
+0x98,
+0x84,
+0x22,
+0x0c,
+0xa2,
+0x32,
+0x94,
+0xdc,
+0x9d,
+0x90,
+0xdb,
+0xe5,
+0xd9,
+0xab,
+0xdf,
+0xe1,
+0xa0,
+0x56,
+0x87,
+0x28,
+0xc2,
+0xe8,
+0x30,
+0x71,
+0x10,
+0xc4,
+0xca,
+0x1c,
+0x11,
+0x08,
+0x76,
+0xa9,
+0xee,
+0xf7,
+0x8a,
+0xa4,
+0x9b,
+0x28,
+0x78,
+0xca,
+0xa1,
+0xc9,
+0xc4,
+0x60,
+0xc7,
+0x21,
+0x9d,
+0x80,
+0xcf,
+0x07,
+0x48,
+0xc9,
+0x8d,
+0xca,
+0x9d,
+0x13,
+0xbf,
+0x0a,
+0x89,
+0xaa,
+0x93,
+0xbb,
+0x78,
+0x5f,
+0x9f,
+0x5a,
+0x49,
+0x60,
+0xfd,
+0xfa,
+0x67,
+0xc5,
+0xa0,
+0x63,
+0x55,
+0x94,
+0x63,
+0xe8,
+0xe3,
+0x81,
+0x29,
+0x23,
+0xd9,
+0x88,
+0xeb,
+0x71,
+0x8c,
+0x9d,
+0xe8,
+0x2b,
+0xd1,
+0xac,
+0x71,
+0xf8,
+0x10,
+0x4f,
+0x4c,
+0x2d,
+0x29,
+0x13,
+0x45,
+0x1b,
+0xfc,
+0x64,
+0x85,
+0x2f,
+0x32,
+0xa1,
+0x53,
+0xd9,
+0x92,
+0x9b,
+0x95,
+0xd9,
+0xa8,
+0x9f,
+0x4e,
+0xdb,
+0x5d,
+0xab,
+0xc3,
+0xa1,
+0x50,
+0x9c,
+0x56,
+0x05,
+0xcf,
+0x41,
+0xf4,
+0x62,
+0xef,
+0x03,
+0x06,
+0x6c,
+0x2d,
+0xe5,
+0x60,
+0x5d,
+0xdf,
+0xfc,
+0xf5,
+0x46,
+0x38,
+0x37,
+0x6f,
+0x77,
+0x4a,
+0x39,
+0xf1,
+0xe1,
+0xcd,
+0x36,
+0x3f,
+0xb1,
+0x09,
+0x1c,
+0xc6,
+0x4c,
+0x51,
+0x80,
+0xbd,
+0x00,
+0x7e,
+0x55,
+0x2c,
+0x1c,
+0xc3,
+0x4e,
+0xd9,
+0x6d,
+0x80,
+0x23,
+0xfa,
+0xf3,
+0x91,
+0xab,
+0xf3,
+0x1c,
+0x58,
+0x14,
+0xd0,
+0x90,
+0xab,
+0x75,
+0xb4,
+0x93,
+0xb0,
+0x2a,
+0x91,
+0x07,
+0xbf,
+0xa1,
+0x26,
+0xbd,
+0x1b,
+0x2f,
+0xee,
+0xb5,
+0xe4,
+0x73,
+0x3e,
+0x65,
+0x70,
+0x8a,
+0x38,
+0xb5,
+0x6e,
+0xdb,
+0x5c,
+0xd6,
+0x35,
+0x7c,
+0x67,
+0x60,
+0xa7,
+0x91,
+0x44,
+0x06,
+0xfc,
+0x83,
+0x2b,
+0xfb,
+0xf0,
+0x47,
+0x37,
+0xbc,
+0xf8,
+0xe2,
+0x5c,
+0x89,
+0x33,
+0xf5,
+0x88,
+0x76,
+0x75,
+0xd4,
+0x12,
+0x7e,
+0xb4,
+0xdd,
+0x6c,
+0x5c,
+0x76,
+0x96,
+0x2e,
+0x76,
+0x87,
+0x7f,
+0x0b,
+0x63,
+0x89,
+0x18,
+0x90,
+0xf1,
+0x86,
+0x80,
+0x8f,
+0xd6,
+0x4e,
+0xad,
+0x0d,
+0x51,
+0x45,
+0x6b,
+0xaa,
+0x34,
+0xfa,
+0xbc,
+0x18,
+0x34,
+0x38,
+0x8c,
+0x34,
+0x3e,
+0xb0,
+0x95,
+0x78,
+0xdf,
+0x8d,
+0x4b,
+0xe3,
+0x61,
+0x7d,
+0xcb,
+0x20,
+0x50,
+0x53,
+0x03,
+0x01,
+0x97,
+0x35,
+0x2c,
+0xcb,
+0xda,
+0xaa,
+0xfe,
+0x00,
+0x93,
+0xa7,
+0x83,
+0x8f,
+0xfb,
+0x3d,
+0x76,
+0x83,
+0x6d,
+0xe2,
+0xe7,
+0x25,
+0x4a,
+0x51,
+0x5c,
+0xc7,
+0x02,
+0xc8,
+0x41,
+0x6c,
+0x44,
+0x03,
+0x8d,
+0x05,
+0x12,
+0x45,
+0x0a,
+0x4d,
+0x7b,
+0x4a,
+0xe1,
+0x21,
+0xa1,
+0xa4,
+0x50,
+0x95,
+0x9f,
+0x3c,
+0x39,
+0x0c,
+0x66,
+0x77,
+0x3c,
+0x73,
+0x45,
+0x0c,
+0xf6,
+0xa6,
+0xf0,
+0x61,
+0x2c,
+0xb5,
+0xf6,
+0xfa,
+0xce,
+0xd8,
+0x2d,
+0xbd,
+0xaf,
+0x44,
+0x7d,
+0x40,
+0xf7,
+0x2e,
+0x3f,
+0xee,
+0x39,
+0x4b,
+0xad,
+0xfe,
+0xda,
+0x63,
+0x31,
+0x30,
+0xfd,
+0x20,
+0xee,
+0x5b,
+0x69,
+0x57,
+0xd0,
+0xfb,
+0x48,
+0xee,
+0x0a,
+0x40,
+0x04,
+0x0e,
+0xa7,
+0x8b,
+0x73,
+0xe7,
+0xff,
+0x3e,
+0x02,
+0x97,
+0x18,
+0x26,
+0xe5,
+0x0b,
+0xc2,
+0x35,
+0xe8,
+0x87,
+0xa9,
+0x51,
+0xf8,
+0xfd,
+0xe9,
+0x0f,
+0x00,
+0x2e,
+0xfd,
+0xb6,
+0x9b,
+0x25,
+0x69,
+0x39,
+0x20,
+0x1d,
+0xc6,
+0x00,
+0xfb,
+0xd6,
+0x00,
+0x30,
+0xc9,
+0x5c,
+0x29,
+0x3b,
+0x87,
+0x8c,
+0x07,
+0x7f,
+0x55,
+0x8c,
+0xaa,
+0xb2,
+0xaa,
+0xc5,
+0x84,
+0x3b,
+0x67,
+0x3b,
+0x4d,
+0xf5,
+0xa9,
+0xe7,
+0xff,
+0xd4,
+0x87,
+0xf1,
+0x2d,
+0xa5,
+0x32,
+0x0a,
+0x63,
+0x0e,
+0x76,
+0x11,
+0xf6,
+0x8d,
+0x50,
+0xee,
+0xe4,
+0x44,
+0x93,
+0xcc,
+0x42,
+0xb3,
+0x0b,
+0xff,
+0x44,
+0x3d,
+0x67,
+0xc2,
+0xcf,
+0xff,
+0xef,
+0x23,
+0x65,
+0xe6,
+0xc3,
+0x6e,
+0x1b,
+0x7d,
+0x76,
+0x0e,
+0xdd,
+0xc1,
+0xa2,
+0x76,
+0xe9,
+0x22,
+0x93,
+0xf5,
+0xcd,
+0xba,
+0x89,
+0x96,
+0x11,
+0x54,
+0x16,
+0xde,
+0xc5,
+0xc9,
+0xa0,
+0xde,
+0x61,
+0x1b,
+0xff,
+0xf7,
+0x6a,
+0xbb,
+0xd0,
+0x10,
+0x53,
+0x6a,
+0xa8,
+0x57,
+0x2c,
+0xfd,
+0x7c,
+0x6f,
+0x31,
+0x78,
+0x9d,
+0xab,
+0x1d,
+0xc3,
+0x10,
+0xef,
+0x63,
+0x3a,
+0x2e,
+0x18,
+0x53,
+0x8d,
+0x45,
+0xf3,
+0x68,
+0x6d,
+0x22,
+0x44,
+0xfa,
+0x9c,
+0x20,
+0x5a,
+0x66,
+0x6a,
+0x68,
+0x2f,
+0xd4,
+0x08,
+0xe1,
+0x29,
+0x4a,
+0x9c,
+0xb5,
+0xb6,
+0x3d,
+0xed,
+0x0b,
+0x58,
+0x21,
+0xdf,
+0xc1,
+0x73,
+0xbf,
+0x8d,
+0x4b,
+0xe2,
+0x2e,
+0xbb,
+0xca,
+0x21,
+0x27,
+0xfe,
+0xa5,
+0x69,
+0xea,
+0x7c,
+0x5c,
+0x9c,
+0x68,
+0x1b,
+0x64,
+0x6f,
+0xd1,
+0xba,
+0x82,
+0xd1,
+0x6a,
+0xb4,
+0x89,
+0x32,
+0xf8,
+0x4f,
+0x9b,
+0x0d,
+0x23,
+0xa6,
+0xd8,
+0x62,
+0x14,
+0xdb,
+0xea,
+0x7c,
+0xd6,
+0x42,
+0x2f,
+0xa9,
+0xa4,
+0x78,
+0x47,
+0xcb,
+0x2e,
+0x63,
+0x1e,
+0xbf,
+0x61,
+0x43,
+0x8a,
+0x95,
+0xd6,
+0x92,
+0x9a,
+0xa4,
+0xd4,
+0x36,
+0xee,
+0xf3,
+0x24,
+0xf4,
+0x4b,
+0x89,
+0x8f,
+0xb8,
+0xc4,
+0x04,
+0xad,
+0x25,
+0xa8,
+0x3a,
+0x7a,
+0x9d,
+0x33,
+0xbc,
+0x1f,
+0x74,
+0x9c,
+0x98,
+0xfe,
+0x16,
+0xa3,
+0x4e,
+0xd9,
+0x7e,
+0x3b,
+0xc8,
+0x7e,
+0x44,
+0x54,
+0x7c,
+0x9a,
+0xb7,
+0x2c,
+0x3a,
+0xf6,
+0x87,
+0x93,
+0x49,
+0xc4,
+0xb7,
+0x6e,
+0x5b,
+0x6e,
+0xfe,
+0x94,
+0x87,
+0x36,
+0x5f,
+0xe7,
+0xd1,
+0x14,
+0xba,
+0xe8,
+0xe1,
+0x85,
+0xfc,
+0xaf,
+0xc2,
+0x14,
+0x3f,
+0xc4,
+0x32,
+0x7b,
+0x22,
+0x01,
+0x98,
+0x26,
+0x70,
+0x0e,
+0x16,
+0x31,
+0x1a,
+0xee,
+0x09,
+0x02,
+0x09,
+0xea,
+0x3f,
+0x74,
+0x0f,
+0x9f,
+0xaa,
+0x53,
+0xd3,
+0xe1,
+0x62,
+0x60,
+0xa8,
+0x5b,
+0x0f,
+0xc7,
+0xbb,
+0x29,
+0xd8,
+0x08,
+0xfa,
+0xa9,
+0x7e,
+0x55,
+0x4c,
+0x2e,
+0x7d,
+0xab,
+0xd1,
+0xdd,
+0xfb,
+0x2b,
+0xe9,
+0x0e,
+0x1e,
+0xe1,
+0x92,
+0x96,
+0x87,
+0x0a,
+0xb1,
+0x98,
+0xe0,
+0x40,
+0x97,
+0xe6,
+0x00,
+0x5b,
+0x73,
+0x51,
+0x0e,
+0x9a,
+0xc1,
+0x75,
+0xda,
+0xa8,
+0xfe,
+0x51,
+0xdc,
+0x56,
+0xc5,
+0x5c,
+0x1d,
+0x00,
+0x73,
+0xea,
+0x93,
+0x2c,
+0x87,
+0x4f,
+0x7d,
+0x85,
+0xdf,
+0xe0,
+0xde,
+0x07,
+0x00,
+0xe2,
+0x22,
+0x0a,
+0xd3,
+0x0c,
+0x46,
+0x26,
+0xbb,
+0x41,
+0x17,
+0xb3,
+0x3b,
+0xc7,
+0xb4,
+0x2b,
+0x40,
+0x4c,
+0x7d,
+0xad,
+0x40,
+0x22,
+0xf1,
+0x0f,
+0x25,
+0xc3,
+0x00,
+0xa3,
+0x5f,
+0xa9,
+0xef,
+0x2f,
+0x56,
+0xfd,
+0xaa,
+0x61,
+0x7d,
+0x40,
+0x30,
+0xca,
+0x53,
+0xd1,
+0x0b,
+0x4f,
+0xc6,
+0xff,
+0x3d,
+0xb3,
+0xa9,
+0x4d,
+0x0b,
+0xd4,
+0x6a,
+0x20,
+0x1c,
+0x6b,
+0x85,
+0x8d,
+0xd4,
+0x89,
+0xcd,
+0x00,
+0x88,
+0x86,
+0xe5,
+0x84,
+0x74,
+0xf1,
+0x68,
+0xaa,
+0x09,
+0xce,
+0xba,
+0xea,
+0x61,
+0x5f,
+0xa4,
+0x4a,
+0xa1,
+0xb9,
+0xbd,
+0xf1,
+0x2f,
+0x11,
+0x34,
+0x10,
+0x2d,
+0x2a,
+0x24,
+0xd7,
+0x9b,
+0xf1,
+0x39,
+0xe5,
+0x76,
+0x04,
+0xb0,
+0xb1,
+0x90,
+0x24,
+0x56,
+0x5c,
+0xb1,
+0x96,
+0x33,
+0x4f,
+0xff,
+0x06,
+0x78,
+0x6d,
+0xc1,
+0x58,
+0x6d,
+0x38,
+0x38,
+0x21,
+0x66,
+0x08,
+0x38,
+0x5e,
+0x24,
+0x02,
+0xd1,
+0x63,
+0x39,
+0xf4,
+0xba,
+0x98,
+0xba,
+0x67,
+0x31,
+0xd1,
+0x0a,
+0x33,
+0x11,
+0x7a,
+0x7c,
+0x2e,
+0xad,
+0xb5,
+0x0f,
+0x90,
+0xbf,
+0x4e,
+0x3f,
+0x09,
+0x01,
+0x28,
+0x64,
+0xa6,
+0xa9,
+0x60,
+0x55,
+0x48,
+0xe1,
+0x75,
+0xc7,
+0x38,
+0x8b,
+0x24,
+0x93,
+0xd7,
+0x90,
+0x7f,
+0x26,
+0xf1,
+0xff,
+0xb7,
+0x78,
+0xff,
+0x67,
+0xb7,
+0x31,
+0x77,
+0x83,
+0xf3,
+0x53,
+0x98,
+0xe8,
+0xe7,
+0xdd,
+0x21,
+0xc9,
+0xe2,
+0xd4,
+0x3c,
+0x33,
+0xf0,
+0x33,
+0x44,
+0xda,
+0x43,
+0x47,
+0xa3,
+0x37,
+0x05,
+0x4f,
+0x4a,
+0x9c,
+0x1c,
+0xe1,
+0xfb,
+0xef,
+0x7d,
+0x14,
+0x6e,
+0x12,
+0xe3,
+0x3d,
+0xb5,
+0xd3,
+0x26,
+0xe3,
+0x96,
+0x93,
+0x47,
+0x49,
+0x0b,
+0x64,
+0xaf,
+0x34,
+0xd9,
+0xf0,
+0x9c,
+0x38,
+0x80,
+0xfa,
+0x6a,
+0x0d,
+0xdb,
+0x53,
+0x21,
+0x63,
+0x40,
+0x41,
+0x35,
+0xee,
+0xd9,
+0x71,
+0xbd,
+0x60,
+0x7f,
+0x00,
+0x56,
+0xea,
+0xc6,
+0x77,
+0xf3,
+0x88,
+0xcc,
+0x07,
+0x66,
+0xe7,
+0x7f,
+0xdd,
+0x17,
+0x54,
+0x5a,
+0xd9,
+0x21,
+0x65,
+0xf6,
+0x69,
+0xbb,
+0x07,
+0x63,
+0x2b,
+0x94,
+0x73,
+0x6c,
+0xfb,
+0x80,
+0x65,
+0x10,
+0xc2,
+0xcf,
+0x8c,
+0xc4,
+0xdb,
+0x6c,
+0x40,
+0xed,
+0x22,
+0x7e,
+0x80,
+0xcd,
+0x2e,
+0xc5,
+0xb1,
+0xe2,
+0x1d,
+0x8c,
+0xa2,
+0xc2,
+0xef,
+0x77,
+0xfc,
+0x5d,
+0x96,
+0xb1,
+0xa5,
+0x4f,
+0x16,
+0x28,
+0x0f,
+0x33,
+0x40,
+0xdf,
+0x21,
+0x61,
+0xc7,
+0x80,
+0x4d,
+0x42,
+0xac,
+0x20,
+0x6e,
+0xe6,
+0xf3,
+0x94,
+0x06,
+0x5e,
+0x21,
+0xe3,
+0x0d,
+0x46,
+0xc3,
+0xcd,
+0x9d,
+0x4b,
+0xa9,
+0x67,
+0xa0,
+0x78,
+0xa1,
+0xcc,
+0xa2,
+0x04,
+0xa2,
+0x5e,
+0xbd,
+0x86,
+0x0d,
+0x1b,
+0xe4,
+0x3d,
+0x27,
+0x30,
+0x06,
+0x21,
+0xcb,
+0xb3,
+0x78,
+0x63,
+0xb2,
+0xcc,
+0x81,
+0x53,
+0x42,
+0x31,
+0x81,
+0x3e,
+0x5c,
+0x14,
+0xd9,
+0xd0,
+0x9d,
+0x14,
+0x98,
+0x2b,
+0xbb,
+0xfe,
+0x4c,
+0x6e,
+0xf5,
+0x39,
+0xce,
+0x68,
+0x6e,
+0xd6,
+0x0a,
+0x97,
+0xad,
+0xd1,
+0x48,
+0xfb,
+0xbf,
+0x01,
+0x66,
+0x37,
+0x70,
+0x80,
+0xcb,
+0x07,
+0xe5,
+0xcb,
+0x5c,
+0x12,
+0x01,
+0xc9,
+0x2f,
+0x45,
+0x5c,
+0xb4,
+0x8d,
+0xe2,
+0x22,
+0x68,
+0xc4,
+0x96,
+0x9f,
+0xcf,
+0x11,
+0x08,
+0xe5,
+0x42,
+0xd6,
+0xde,
+0x2a,
+0x6c,
+0xfe,
+0xaf,
+0x87,
+0x05,
+0x7e,
+0x84,
+0x13,
+0x83,
+0x2d,
+0x2e,
+0xa1,
+0x07,
+0xb2,
+0x12,
+0xa9,
+0xb5,
+0xd5,
+0x64,
+0x75,
+0xcf,
+0xc9,
+0x67,
+0x7f,
+0x98,
+0x9d,
+0x37,
+0xf4,
+0x3a,
+0xde,
+0xf8,
+0x35,
+0x39,
+0x39,
+0xce,
+0xec,
+0x8a,
+0x1e,
+0xde,
+0x91,
+0x75,
+0x7d,
+0x7a,
+0xf0,
+0xd4,
+0x43,
+0x74,
+0x66,
+0xf6,
+0x84,
+0x4e,
+0xf0,
+0x5e,
+0x8c,
+0x0f,
+0x4a,
+0x12,
+0x03,
+0xd8,
+0x73,
+0xac,
+0xb6,
+0x11,
+0x92,
+0x00,
+0x84,
+0x5d,
+0x3d,
+0x48,
+0xab,
+0x7b,
+0x35,
+0x7a,
+0x8d,
+0x88,
+0xa6,
+0xc4,
+0xa5,
+0x16,
+0x01,
+0xd0,
+0xfb,
+0xde,
+0x2b,
+0x5d,
+0xe1,
+0xcc,
+0x3c,
+0xec,
+0x10,
+0x00,
+0x5a,
+0x77,
+0xa9,
+0x70,
+0xc1,
+0x2c,
+0xaf,
+0x78,
+0xfa,
+0x1f,
+0x3f,
+0x08,
+0xa2,
+0x55,
+0xfb,
+0xcc,
+0xeb,
+0x94,
+0x67,
+0x57,
+0x4f,
+0xda,
+0xc2,
+0xc5,
+0x9c,
+0xe0,
+0xda,
+0x12,
+0xe9,
+0x9a,
+0x0f,
+0xcd,
+0xdb,
+0xa0,
+0x45,
+0x6f,
+0x35,
+0x0d,
+0x08,
+0x90,
+0xc6,
+0x7c,
+0x09,
+0xcf,
+0xc9,
+0xf5,
+0xac,
+0x8b,
+0x80,
+0x7e,
+0xa7,
+0x53,
+0x6e,
+0x3a,
+0xa2,
+0x8b,
+0xd6,
+0x09,
+0x8f,
+0xd8,
+0xaa,
+0xc1,
+0x48,
+0x24,
+0x0f,
+0x87,
+0x70,
+0x36,
+0xf3,
+0x01,
+0x44,
+0x71,
+0x6d,
+0x64,
+0x77,
+0x2a,
+0x13,
+0xb6,
+0xfa,
+0x5f,
+0x5f,
+0xa4,
+0x3a,
+0xd9,
+0x14,
+0x4c,
+0x93,
+0x9d,
+0x77,
+0xa3,
+0xf6,
+0xd5,
+0xf0,
+0x94,
+0xca,
+0x95,
+0xe3,
+0x0e,
+0x56,
+0x29,
+0x74,
+0xff,
+0x5b,
+0x01,
+0x90,
+0xb7,
+0xa0,
+0x65,
+0xc5,
+0x09,
+0x30,
+0x7c,
+0x16,
+0x78,
+0x11,
+0x67,
+0x0b,
+0x31,
+0x48,
+0x4b,
+0x69,
+0xf7,
+0x63,
+0xf0,
+0x89,
+0xd3,
+0x82,
+0xdb,
+0x28,
+0x47,
+0x73,
+0x13,
+0xa7,
+0xbd,
+0x6a,
+0x5e,
+0x12,
+0x48,
+0x32,
+0x29,
+0xc7,
+0x50,
+0xa9,
+0xcb,
+0xf3,
+0xe2,
+0x0e,
+0xeb,
+0xd7,
+0x65,
+0x69,
+0x69,
+0xd5,
+0xd3,
+0x27,
+0xe6,
+0x76,
+0x4b,
+0x8f,
+0xe1,
+0xbd,
+0x11,
+0x71,
+0x1d,
+0x09,
+0xda,
+0xb9,
+0xc2,
+0x40,
+0xce,
+0xcc,
+0x2a,
+0x75,
+0x14,
+0xec,
+0x28,
+0xb5,
+0x6b,
+0x27,
+0x85,
+0xad,
+0x40,
+0x01,
+0x6c,
+0x46,
+0xb5,
+0x88,
+0x2c,
+0x32,
+0x10,
+0x50,
+0xa2,
+0x4e,
+0xc7,
+0x84,
+0x05,
+0x3d,
+0x08,
+0x83,
+0xcf,
+0x83,
+0xe1,
+0xf3,
+0xa4,
+0xe7,
+0x23,
+0x23,
+0x66,
+0x04,
+0x26,
+0xe5,
+0x48,
+0x24,
+0xd3,
+0xec,
+0xe6,
+0x69,
+0xfc,
+0x06,
+0x36,
+0x86,
+0x6b,
+0xed,
+0x98,
+0x81,
+0x00,
+0x38,
+0x55,
+0x11,
+0x5a,
+0xdd,
+0x89,
+0x88,
+0xff,
+0xb3,
+0x94,
+0x1b,
+0xbc,
+0xd5,
+0x23,
+0x50,
+0x74,
+0xc1,
+0x63,
+0x44,
+0xc4,
+0x16,
+0xa3,
+0xb3,
+0xe0,
+0x78,
+0x3a,
+0xa2,
+0x1a,
+0x28,
+0x31,
+0xea,
+0xea,
+0xe5,
+0x4c,
+0x4e,
+0x78,
+0x6b,
+0x6a,
+0xfd,
+0xa2,
+0x67,
+0x47,
+0x1d,
+0x27,
+0xfa,
+0xe9,
+0xa4,
+0x2d,
+0x2f,
+0x92,
+0x2a,
+0xb2,
+0xa7,
+0xa0,
+0x66,
+0xc3,
+0xb9,
+0x98,
+0xad,
+0xdb,
+0x8c,
+0x5e,
+0x9f,
+0xfc,
+0xe8,
+0xd6,
+0x81,
+0xe7,
+0xd0,
+0x80,
+0x17,
+0xa9,
+0x7a,
+0x27,
+0x5a,
+0xbf,
+0x74,
+0x8d,
+0x39,
+0x04,
+0xf7,
+0x03,
+0x81,
+0xd6,
+0xba,
+0xa8,
+0x5e,
+0x97,
+0x3b,
+0x97,
+0xdf,
+0x02,
+0x56,
+0xdf,
+0xe1,
+0x35,
+0x2d,
+0xde,
+0x78,
+0x67,
+0xc3,
+0x4b,
+0x2a,
+0xc8,
+0xbc,
+0x55,
+0x6c,
+0x51,
+0xdd,
+0xfc,
+0x77,
+0x02,
+0x89,
+0x0b,
+0x4f,
+0x0b,
+0x8b,
+0xdc,
+0x62,
+0xb3,
+0x74,
+0xf1,
+0xa1,
+0xda,
+0x58,
+0x25,
+0x58,
+0x55,
+0xb0,
+0x6b,
+0xad,
+0xa9,
+0x1e,
+0x0b,
+0xa8,
+0x05,
+0x5f,
+0xc7,
+0x00,
+0xe2,
+0x35,
+0xa9,
+0xc3,
+0x8b,
+0xf0,
+0xab,
+0x88,
+0x34,
+0xe1,
+0x4c,
+0xa9,
+0xda,
+0xb5,
+0x16,
+0x98,
+0x24,
+0xa6,
+0x4c,
+0xb6,
+0x43,
+0xf3,
+0x4e,
+0xec,
+0x69,
+0x8b,
+0x77,
+0xdd,
+0x0e,
+0xf5,
+0x9d,
+0x65,
+0x09,
+0x0d,
+0x40,
+0x4f,
+0xba,
+0xce,
+0xa5,
+0xbd,
+0x76,
+0xb3,
+0x48,
+0x0d,
+0x82,
+0xed,
+0x81,
+0x01,
+0x62,
+0x5e,
+0x1f,
+0x87,
+0xe9,
+0x55,
+0x62,
+0x29,
+0xcc,
+0xeb,
+0xca,
+0xef,
+0x8c,
+0xeb,
+0x13,
+0xc6,
+0xff,
+0xf3,
+0x26,
+0x71,
+0xca,
+0x81,
+0x3e,
+0x91,
+0x28,
+0x0b,
+0x34,
+0xbd,
+0x50,
+0x37,
+0x83,
+0x29,
+0x7e,
+0x35,
+0xfe,
+0x0a,
+0xb8,
+0x02,
+0x76,
+0x20,
+0xd6,
+0x29,
+0x77,
+0x65,
+0xc6,
+0xb4,
+0xe2,
+0x42,
+0xa8,
+0x12,
+0xe0,
+0xa8,
+0xe1,
+0xf3,
+0x29,
+0xe0,
+0x71,
+0xf4,
+0x86,
+0xa1,
+0x6a,
+0x65,
+0x76,
+0x97,
+0x06,
+0x1f,
+0xdb,
+0xa9,
+0x2d,
+0x2e,
+0xac,
+0xb1,
+0xb3,
+0x67,
+0x04,
+0x15,
+0x38,
+0x7e,
+0xf5,
+0x11,
+0x0a,
+0x74,
+0x1b,
+0x70,
+0x6c,
+0x49,
+0x74,
+0x2d,
+0xd4,
+0xb2,
+0x86,
+0xa3,
+0xdd,
+0x44,
+0xe2,
+0x5a,
+0x01,
+0x7f,
+0xa5,
+0x6c,
+0x3c,
+0x71,
+0x02,
+0xd7,
+0x29,
+0xf3,
+0x4c,
+0x28,
+0xc6,
+0xec,
+0x65,
+0xfe,
+0xd3,
+0x42,
+0x80,
+0x04,
+0x2a,
+0xbd,
+0x6e,
+0xa2,
+0xd8,
+0xe6,
+0x1e,
+0xa8,
+0x08,
+0x4e,
+0x22,
+0x7f,
+0xe8,
+0xcc,
+0x33,
+0xc1,
+0x58,
+0x15,
+0xa3,
+0x20,
+0x88,
+0xc2,
+0x3b,
+0xf1,
+0x17,
+0xe6,
+0x18,
+0x5f,
+0x70,
+0x91,
+0xc8,
+0x39,
+0x65,
+0x1b,
+0xc8,
+0x3e,
+0x00,
+0x4d,
+0x4f,
+0x6a,
+0x8b,
+0x4d,
+0x63,
+0xcb,
+0x24,
+0x91,
+0x44,
+0xdc,
+0x13,
+0x9d,
+0xd3,
+0xf7,
+0x8b,
+0x06,
+0xe8,
+0xc3,
+0x6e,
+0x47,
+0xe1,
+0x8a,
+0xec,
+0x98,
+0x22,
+0xee,
+0x50,
+0x9a,
+0xd5,
+0x74,
+0xf8,
+0xac,
+0x45,
+0xcb,
+0xeb,
+0xd8,
+0xf4,
+0x3e,
+0xbb,
+0x5d,
+0x6c,
+0xed,
+0xc4,
+0x3b,
+0x3a,
+0x69,
+0x16,
+0xc5,
+0x16,
+0x43,
+0xb6,
+0x21,
+0xd4,
+0xce,
+0x90,
+0x2d,
+0xfc,
+0x88,
+0x4c,
+0xd6,
+0x46,
+0xc8,
+0x55,
+0xc2,
+0x7f,
+0x81,
+0xb7,
+0xc0,
+0x10,
+0x9d,
+0xdc,
+0x1c,
+0xed,
+0x45,
+0xb7,
+0xe3,
+0x6d,
+0x36,
+0x7c,
+0x0e,
+0x68,
+0x16,
+0xe3,
+0x00,
+0x9c,
+0x69,
+0x40,
+0xcf,
+0x11,
+0x30,
+0x5c,
+0x9f,
+0xfd,
+0xed,
+0x5c,
+0xd0,
+0x26,
+0x2b,
+0x66,
+0x54,
+0x53,
+0x9c,
+0xa7,
+0xfb,
+0x08,
+0x92,
+0xaf,
+0xcf,
+0x18,
+0xe9,
+0x6e,
+0x11,
+0x8b,
+0xc8,
+0xe4,
+0x2c,
+0xe7,
+0xc1,
+0x3c,
+0x2a,
+0x75,
+0x54,
+0x14,
+0x97,
+0xdd,
+0x49,
+0x8c,
+0x1a,
+0x74,
+0x89,
+0x83,
+0x88,
+0x74,
+0xde,
+0xf1,
+0x90,
+0x13,
+0x1b,
+0x15,
+0xda,
+0x93,
+0xb3,
+0x81,
+0x50,
+0x46,
+0x5d,
+0x60,
+0x64,
+0xcd,
+0x4e,
+0xa2,
+0x3d,
+0x7f,
+0x43,
+0xb0,
+0x7f,
+0x09,
+0xd4,
+0xa0,
+0x32,
+0x63,
+0x70,
+0x03,
+0xc6,
+0xcd,
+0xbc,
+0x43,
+0xf6,
+0xf0,
+0xa5,
+0x4b,
+0x92,
+0x25,
+0x70,
+0x8d,
+0x76,
+0xa6,
+0xe4,
+0x11,
+0xdc,
+0xc4,
+0xbd,
+0x4e,
+0xe7,
+0x31,
+0x7a,
+0xca,
+0xc6,
+0xf3,
+0xc6,
+0x8c,
+0x40,
+0xd8,
+0x21,
+0xdd,
+0x4c,
+0x5a,
+0x50,
+0x3a,
+0x12,
+0x6c,
+0x12,
+0x85,
+0x9b,
+0xb1,
+0x78,
+0x27,
+0x61,
+0xee,
+0x68,
+0xb5,
+0xed,
+0x73,
+0x71,
+0x9a,
+0xa4,
+0x9b,
+0x62,
+0x16,
+0x1e,
+0xd6,
+0xe5,
+0x4c,
+0x8a,
+0x6d,
+0x31,
+0x0b,
+0x03,
+0x64,
+0x8c,
+0x40,
+0xca,
+0x85,
+0xd3,
+0x15,
+0x7a,
+0xd5,
+0xf0,
+0xad,
+0xc6,
+0x0d,
+0x6f,
+0x0a,
+0xf6,
+0xaf,
+0x3f,
+0x5c,
+0x54,
+0xa3,
+0x2b,
+0x78,
+0x03,
+0x97,
+0xe3,
+0x91,
+0x17,
+0xaf,
+0x95,
+0x82,
+0xa6,
+0xab,
+0x50,
+0xd6,
+0x16,
+0xed,
+0x7f,
+0x70,
+0xd1,
+0xef,
+0x74,
+0x6f,
+0x5a,
+0xc7,
+0xe4,
+0x7b,
+0xf5,
+0x05,
+0xa3,
+0x02,
+0xe3,
+0x43,
+0xfa,
+0x7a,
+0x83,
+0x9a,
+0x7e,
+0x70,
+0xed,
+0x8c,
+0x17,
+0xa6,
+0x50,
+0xd8,
+0x6e,
+0x36,
+0x86,
+0x39,
+0xff,
+0x5f,
+0x4b,
+0x6e,
+0xc0,
+0x6a,
+0x28,
+0x5a,
+0x81,
+0xde,
+0x5c,
+0x96,
+0xb0,
+0xe9,
+0x39,
+0x01,
+0xd3,
+0x84,
+0xc8,
+0xf0,
+0xe5,
+0xc8,
+0x43,
+0x01,
+0x38,
+0x2b,
+0x63,
+0xdd,
+0xd4,
+0xfb,
+0x69,
+0x9b,
+0xc2,
+0x0d,
+0x11,
+0x88,
+0xd4,
+0xf7,
+0x28,
+0x40,
+0x7f,
+0x34,
+0x09,
+0xe8,
+0xb5,
+0xcf,
+0xb9,
+0x7d,
+0x98,
+0x0f,
+0x77,
+0xaf,
+0xf0,
+0x64,
+0x09,
+0x23,
+0x88,
+0xc3,
+0xd1,
+0xe9,
+0x4c,
+0x0c,
+0xc6,
+0x34,
+0xa7,
+0x00,
+0x26,
+0xd7,
+0xd0,
+0x5b,
+0x62,
+0xaa,
+0xfb,
+0x18,
+0xaf,
+0xa2,
+0x4d,
+0xb9,
+0x32,
+0x44,
+0xbd,
+0x87,
+0xa7,
+0x2d,
+0x15,
+0x5f,
+0x63,
+0xea,
+0x18,
+0x20,
+0x66,
+0x37,
+0x78,
+0xdf,
+0xe4,
+0x49,
+0x34,
+0x07,
+0xe2,
+0x2f,
+0x57,
+0xd9,
+0x6d,
+0xcb,
+0xc4,
+0x1e,
+0x9b,
+0xb4,
+0x24,
+0xe0,
+0x10,
+0x47,
+0x7a,
+0xb8,
+0x2c,
+0xd5,
+0x67,
+0xe9,
+0x60,
+0x95,
+0xf1,
+0x49,
+0xeb,
+0x85,
+0x96,
+0xee,
+0xad,
+0x77,
+0x11,
+0x4f,
+0x8a,
+0xf8,
+0x61,
+0x49,
+0x47,
+0x35,
+0x54,
+0xca,
+0xa2,
+0xf4,
+0x30,
+0x2e,
+0x1e,
+0xd5,
+0x6b,
+0x32,
+0xda,
+0x1e,
+0xa4,
+0x27,
+0x41,
+0x0c,
+0x92,
+0x94,
+0xe8,
+0x37,
+0x9a,
+0x21,
+0xcc,
+0x4d,
+0x20,
+0x2b,
+0xcf,
+0xe0,
+0x57,
+0x69,
+0x92,
+0x61,
+0x4e,
+0x75,
+0xe7,
+0xc1,
+0xc0,
+0xd5,
+0x33,
+0x13,
+0xd8,
+0x69,
+0xb5,
+0xb7,
+0x37,
+0xf1,
+0xd4,
+0xff,
+0x46,
+0x20,
+0xc3,
+0x6f,
+0x82,
+0x7f,
+0x5d,
+0x13,
+0xcc,
+0x94,
+0x2d,
+0xf3,
+0x8b,
+0x68,
+0x58,
+0x3b,
+0x81,
+0x7f,
+0x85,
+0xe5,
+0xd5,
+0xbf,
+0xca,
+0xb2,
+0xd8,
+0x7a,
+0xfb,
+0x7c,
+0x90,
+0xfd,
+0x33,
+0x9a,
+0x54,
+0xf3,
+0x20,
+0xbc,
+0x05,
+0x60,
+0x4c,
+0xed,
+0x3c,
+0x81,
+0x6e,
+0x37,
+0xe3,
+0x03,
+0x3a,
+0xe9,
+0x55,
+0x56,
+0x8c,
+0x5e,
+0xbd,
+0x07,
+0xfd,
+0x4e,
+0x1c,
+0xc6,
+0x60,
+0x5c,
+0x7d,
+0x0b,
+0xfe,
+0xad,
+0x8d,
+0x0c,
+0xad,
+0xb7,
+0x20,
+0x18,
+0x9a,
+0x0d,
+0x29,
+0xed,
+0xfe,
+0x29,
+0xc4,
+0x70,
+0x6a,
+0xbd,
+0xb2,
+0x48,
+0x14,
+0x92,
+0x7e,
+0x73,
+0xe3,
+0xf7,
+0xdd,
+0x10,
+0xa7,
+0x02,
+0x07,
+0x35,
+0x85,
+0xd6,
+0xd1,
+0x06,
+0x58,
+0x24,
+0xda,
+0x4d,
+0xe1,
+0x04,
+0xde,
+0x53,
+0x7b,
+0xf6,
+0x89,
+0xca,
+0xce,
+0x6d,
+0x24,
+0x3a,
+0xe8,
+0x78,
+0x48,
+0x7f,
+0x28,
+0x4d,
+0x82,
+0x0b,
+0x06,
+0xea,
+0xd8,
+0x9c,
+0x50,
+0x7b,
+0x5f,
+0x15,
+0xba,
+0x4f,
+0x97,
+0x8c,
+0x94,
+0x3a,
+0x0d,
+0x92,
+0x10,
+0xa3,
+0x35,
+0x92,
+0xb3,
+0x84,
+0x40,
+0xf1,
+0x18,
+0xdc,
+0x24,
+0x1d,
+0xb0,
+0x4b,
+0x44,
+0x30,
+0xb7,
+0x56,
+0x65,
+0xb6,
+0x3b,
+0x02,
+0xe9,
+0xff,
+0x7a,
+0x7c,
+0xd7,
+0x8b,
+0xb7,
+0x6d,
+0xb9,
+0x7a,
+0xa2,
+0x0e,
+0x5a,
+0x0d,
+0xc3,
+0xb2,
+0xc6,
+0xe8,
+0x27,
+0x3a,
+0xc2,
+0x48,
+0x90,
+0x8f,
+0xe7,
+0xb1,
+0x50,
+0xb7,
+0xf6,
+0xb3,
+0x9f,
+0xa2,
+0xf4,
+0x7e,
+0xfd,
+0x0f,
+0x32,
+0x5e,
+0x01,
+0x97,
+0xc0,
+0x13,
+0x6b,
+0xfa,
+0x12,
+0x3d,
+0x0f,
+0xc6,
+0xa8,
+0xb4,
+0xb1,
+0x0e,
+0xcc,
+0x25,
+0x5f,
+0x39,
+0xe1,
+0x62,
+0xa7,
+0x2e,
+0xf7,
+0x61,
+0xc0,
+0xfc,
+0x65,
+0xd3,
+0x80,
+0x6b,
+0x5e,
+0xfc,
+0xa8,
+0xdf,
+0xe7,
+0x3f,
+0x9f,
+0x02,
+0x6e,
+0xa1,
+0x3f,
+0xd7,
+0x85,
+0x6f,
+0x08,
+0x95,
+0xf4,
+0x71,
+0x58,
+0x40,
+0xe7,
+0x20,
+0x00,
+0x4a,
+0x65,
+0x87,
+0x31,
+0xe2,
+0x73,
+0xa1,
+0x27,
+0x4d,
+0xc0,
+0x82,
+0x7a,
+0x24,
+0xc8,
+0x6c,
+0x4b,
+0xe6,
+0x7d,
+0x67,
+0x54,
+0x2d,
+0xcc,
+0xa4,
+0xca,
+0x09,
+0xb2,
+0xb4,
+0xec,
+0x6f,
+0x49,
+0xb4,
+0x5d,
+0x34,
+0x9c,
+0x0a,
+0x8f,
+0x3b,
+0x57,
+0x98,
+0x49,
+0xdc,
+0x51,
+0xa7,
+0x35,
+0x56,
+0x44,
+0xa5,
+0x7a,
+0xd3,
+0x0d,
+0x09,
+0x2a,
+0x22,
+0x95,
+0xe7,
+0xee,
+0xc2,
+0x89,
+0xfa,
+0x09,
+0xa4,
+0xdd,
+0x54,
+0x32,
+0x0d,
+0x3c,
+0xf4,
+0xc8,
+0xff,
+0x64,
+0x6a,
+0xf8,
+0xf6,
+0x61,
+0x64,
+0x4f,
+0x9b,
+0x96,
+0xcc,
+0x24,
+0xcc,
+0xcd,
+0x55,
+0xe2,
+0x03,
+0x26,
+0x73,
+0xf5,
+0x28,
+0x7d,
+0x0a,
+0xe6,
+0xaf,
+0x23,
+0x02,
+0xf7,
+0x6e,
+0x4c,
+0x6f,
+0x4a,
+0x49,
+0x12,
+0x5a,
+0x63,
+0x18,
+0xe2,
+0x4c,
+0x95,
+0x37,
+0x26,
+0xdb,
+0x3a,
+0x95,
+0x11,
+0x51,
+0x9c,
+0xd9,
+0xe9,
+0x45,
+0x5f,
+0x55,
+0x8c,
+0x97,
+0x2a,
+0x36,
+0x16,
+0xb1,
+0x0b,
+0x2f,
+0x54,
+0x68,
+0x5b,
+0x15,
+0x1b,
+0xaf,
+0x7a,
+0xa3,
+0xd0,
+0x24,
+0x0a,
+0x3e,
+0x03,
+0x31,
+0x26,
+0x08,
+0xdc,
+0xdb,
+0x34,
+0x98,
+0xe6,
+0xe5,
+0x15,
+0xf8,
+0x14,
+0x85,
+0xea,
+0x5c,
+0xa5,
+0xab,
+0xf0,
+0x3a,
+0x3a,
+0x14,
+0x4a,
+0x5f,
+0x4b,
+0x6d,
+0xa6,
+0xdf,
+0x9f,
+0x5f,
+0x13,
+0xa5,
+0x81,
+0x1e,
+0xd3,
+0x16,
+0x96,
+0x2b,
+0x92,
+0xb1,
+0xed,
+0x39,
+0x08,
+0xf5,
+0x7c,
+0xb3,
+0x57,
+0xb4,
+0x62,
+0xd1,
+0x60,
+0x0a,
+0x90,
+0xaf,
+0xd0,
+0x89,
+0xdf,
+0x7b,
+0xde,
+0x82,
+0xa2,
+0xa8,
+0xb8,
+0x66,
+0x27,
+0x0a,
+0x78,
+0x71,
+0x26,
+0x67,
+0x08,
+0x54,
+0x44,
+0x2b,
+0x27,
+0xfd,
+0x08,
+0xe1,
+0xd4,
+0xb9,
+0x75,
+0x6c,
+0x02,
+0xbf,
+0x89,
+0xdc,
+0x7c,
+0x51,
+0x29,
+0xa9,
+0xa8,
+0x63,
+0x3f,
+0xb3,
+0x3f,
+0x11,
+0x4c,
+0x20,
+0x0f,
+0x9b,
+0x7c,
+0x85,
+0x68,
+0x5f,
+0x00,
+0x53,
+0x1f,
+0xf1,
+0xeb,
+0x3d,
+0x58,
+0x5d,
+0x85,
+0xe5,
+0xc9,
+0xd7,
+0xb5,
+0x23,
+0x2d,
+0x94,
+0xb4,
+0x6b,
+0x2f,
+0x41,
+0xe8,
+0xa8,
+0xbf,
+0xfa,
+0x57,
+0xe4,
+0xbb,
+0xc0,
+0x58,
+0x4a,
+0x7a,
+0xb0,
+0xcc,
+0x87,
+0x36,
+0x61,
+0x7d,
+0x24,
+0x6a,
+0xe2,
+0x40,
+0xd7,
+0xfe,
+0x78,
+0xea,
+0xbd,
+0x47,
+0xe6,
+0x63,
+0x8c,
+0x91,
+0xb1,
+0xec,
+0xdc,
+0x65,
+0x45,
+0x31,
+0xa8,
+0x26,
+0xec,
+0x6c,
+0xa6,
+0x65,
+0x13,
+0x90,
+0xb0,
+0x9c,
+0xb5,
+0x4f,
+0x55,
+0x2c,
+0xef,
+0xa5,
+0xff,
+0x9b,
+0xc1,
+0xbb,
+0xee,
+0xb9,
+0x81,
+0x2b,
+0xe4,
+0x81,
+0x8c,
+0xdd,
+0x1a,
+0x08,
+0xbf,
+0xf4,
+0x71,
+0x1d,
+0xd3,
+0xcf,
+0x80,
+0xd4,
+0xba,
+0x6c,
+0xe5,
+0x1a,
+0x57,
+0x4b,
+0xf5,
+0x34,
+0x30,
+0x14,
+0xdc,
+0x05,
+0x3d,
+0x0c,
+0xd8,
+0xa0,
+0xbf,
+0x07,
+0xd1,
+0x91,
+0xe5,
+0x6c,
+0x2f,
+0x85,
+0x58,
+0x16,
+0x6c,
+0x9c,
+0x15,
+0x5b,
+0x24,
+0x54,
+0xd8,
+0x67,
+0x4c,
+0x61,
+0x37,
+0x9b,
+0xeb,
+0x54,
+0x95,
+0x9d,
+0xcc,
+0x2b,
+0xd5,
+0xfb,
+0x76,
+0x9a,
+0xa8,
+0x86,
+0x12,
+0xea,
+0xa2,
+0xd5,
+0x94,
+0xb2,
+0x06,
+0xcd,
+0x68,
+0x2c,
+0xb2,
+0x91,
+0xd7,
+0xbb,
+0x3a,
+0xea,
+0x97,
+0x5e,
+0xee,
+0xe1,
+0x44,
+0xa4,
+0x20,
+0x73,
+0xde,
+0x17,
+0xeb,
+0x29,
+0x83,
+0x05,
+0x16,
+0xee,
+0x7c,
+0x38,
+0x60,
+0x69,
+0x1d,
+0xad,
+0x26,
+0x51,
+0x68,
+0x30,
+0xbc,
+0x11,
+0xe9,
+0x08,
+0x80,
+0x93,
+0x7b,
+0x5f,
+0xb2,
+0xaa,
+0xa5,
+0x93,
+0x42,
+0x2e,
+0x49,
+0x36,
+0xc7,
+0x5b,
+0xde,
+0xd4,
+0x55,
+0xda,
+0x61,
+0xb4,
+0xe4,
+0x58,
+0xaa,
+0x0b,
+0xed,
+0x37,
+0xb3,
+0x8a,
+0x91,
+0x14,
+0x87,
+0x09,
+0x69,
+0x5b,
+0x0e,
+0x49,
+0x96,
+0x1c,
+0x7c,
+0xf4,
+0xcb,
+0x8a,
+0xd9,
+0xe1,
+0xc0,
+0x85,
+0xa7,
+0xed,
+0x5e,
+0xf4,
+0x50,
+0x67,
+0xcf,
+0x49,
+0xe7,
+0x4e,
+0xee,
+0x48,
+0x9f,
+0x5c,
+0x49,
+0x89,
+0x9d,
+0x25,
+0xec,
+0xdf,
+0xa9,
+0x8a,
+0x54,
+0x1b,
+0xd4,
+0x96,
+0xeb,
+0x4a,
+0x47,
+0x50,
+0x3d,
+0xe2,
+0x4c,
+0xe3,
+0xad,
+0x13,
+0x25,
+0xef,
+0x2f,
+0xc5,
+0x30,
+0x39,
+0x56,
+0x57,
+0xf6,
+0x96,
+0xab,
+0xa1,
+0xa3,
+0x8a,
+0x22,
+0x14,
+0x00,
+0x35,
+0x43,
+0x0a,
+0xc5,
+0x6e,
+0xec,
+0x00,
+0xe1,
+0x88,
+0xd3,
+0xa1,
+0x02,
+0x6b,
+0x1a,
+0x05,
+0x3c,
+0x33,
+0xe6,
+0x0d,
+0x6c,
+0xd0,
+0xfe,
+0x08,
+0x3c,
+0x89,
+0x55,
+0x0e,
+0x47,
+0x09,
+0x6e,
+0xeb,
+0x07,
+0x5c,
+0x23,
+0x16,
+0xc0,
+0x3c,
+0x07,
+0x5f,
+0x2e,
+0x23,
+0x0e,
+0x7c,
+0x86,
+0x7d,
+0xc4,
+0x8f,
+0xdf,
+0x22,
+0x38,
+0xb0,
+0xe1,
+0xa3,
+0x88,
+0x86,
+0xd0,
+0x7e,
+0xad,
+0x90,
+0x1c,
+0x1f,
+0x10,
+0xb1,
+0x2f,
+0x2d,
+0xdb,
+0x96,
+0x34,
+0xf6,
+0x09,
+0x06,
+0x0b,
+0x65,
+0x33,
+0xb7,
+0x6e,
+0x13,
+0xed,
+0x0f,
+0x0e,
+0x45,
+0x77,
+0xfd,
+0xaa,
+0x09,
+0x28,
+0x4d,
+0x92,
+0x11,
+0x2f,
+0xf5,
+0x69,
+0xe8,
+0xb6,
+0x6c,
+0x0a,
+0x8b,
+0x5e,
+0xa5,
+0x7d,
+0x3d,
+0xa8,
+0x12,
+0x2c,
+0xf8,
+0x56,
+0xe2,
+0xd0,
+0xa2,
+0xa8,
+0x4a,
+0x43,
+0x35,
+0x3c,
+0xe6,
+0x0d,
+0x44,
+0xc1,
+0xaf,
+0xd8,
+0x80,
+0xe5,
+0xb8,
+0x30,
+0x5b,
+0x9a,
+0xf8,
+0xc8,
+0x70,
+0x36,
+0x05,
+0xd7,
+0x6a,
+0x04,
+0xf0,
+0x40,
+0x47,
+0xb2,
+0x39,
+0xf5,
+0x64,
+0x26,
+0x16,
+0xeb,
+0x26,
+0x30,
+0x9b,
+0xdd,
+0x18,
+0xe0,
+0x11,
+0x03,
+0xb1,
+0x4b,
+0x67,
+0x03,
+0x2f,
+0x41,
+0xb6,
+0xc9,
+0x2f,
+0xd3,
+0x24,
+0x85,
+0xba,
+0xe3,
+0x90,
+0xe8,
+0x31,
+0xf5,
+0xbb,
+0x1a,
+0x81,
+0xe1,
+0xa0,
+0x45,
+0xe8,
+0x04,
+0x54,
+0x2e,
+0x6b,
+0xb8,
+0xdb,
+0x66,
+0x4b,
+0xc3,
+0x94,
+0x22,
+0xc1,
+0x68,
+0xb2,
+0xce,
+0xbc,
+0xce,
+0x3c,
+0xcf,
+0xbf,
+0x7b,
+0xc5,
+0x2d,
+0x64,
+0xee,
+0x54,
+0x6d,
+0xf3,
+0x4d,
+0x92,
+0x97,
+0xcc,
+0x81,
+0x6b,
+0x43,
+0xed,
+0x18,
+0x70,
+0xce,
+0x63,
+0x58,
+0xfe,
+0x7c,
+0x6c,
+0x8b,
+0x30,
+0x27,
+0xb8,
+0xe0,
+0xf4,
+0xfd,
+0xe5,
+0x0c,
+0x47,
+0xa1,
+0xa1,
+0x60,
+0x76,
+0xf4,
+0xa3,
+0xba,
+0xef,
+0x74,
+0xd6,
+0x96,
+0xb9,
+0x83,
+0x9c,
+0x02,
+0xb7,
+0x12,
+0x66,
+0x16,
+0x92,
+0xd7,
+0xe0,
+0x3f,
+0x9a,
+0xcb,
+0x93,
+0xc6,
+0x2a,
+0xb9,
+0xcb,
+0xe9,
+0x0d,
+0x04,
+0xf0,
+0x80,
+0xc1,
+0xa5,
+0x26,
+0xf1,
+0x9a,
+0xaf,
+0x65,
+0xaa,
+0xd1,
+0xbc,
+0x02,
+0xdb,
+0x44,
+0xdf,
+0xba,
+0xb3,
+0x7f,
+0xc4,
+0x46,
+0xc0,
+0x56,
+0x26,
+0x24,
+0x58,
+0x18,
+0x03,
+0x69,
+0x34,
+0x86,
+0x97,
+0x2d,
+0x2a,
+0x55,
+0x77,
+0x8c,
+0x8d,
+0xc5,
+0xd7,
+0xbc,
+0x9a,
+0xc9,
+0xfa,
+0x3c,
+0x7d,
+0x76,
+0xc9,
+0x1e,
+0xfc,
+0x38,
+0x2c,
+0x39,
+0x88,
+0x41,
+0x1b,
+0xf8,
+0xdd,
+0x96,
+0xa1,
+0x2e,
+0x4c,
+0x7b,
+0xab,
+0x84,
+0xb3,
+0xd4,
+0x8a,
+0xb9,
+0x9a,
+0x7d,
+0x1d,
+0xa6,
+0xe1,
+0x4e,
+0x90,
+0x8f,
+0xd6,
+0xb5,
+0x2f,
+0xfd,
+0xf1,
+0x91,
+0x3b,
+0xfa,
+0x9a,
+0x9a,
+0x0f,
+0x7e,
+0x0b,
+0x2b,
+0x95,
+0x17,
+0xbc,
+0x5e,
+0xb6,
+0xe8,
+0xd0,
+0xca,
+0x4a,
+0x45,
+0xf6,
+0xa5,
+0x34,
+0xcd,
+0xb1,
+0x51,
+0xd7,
+0x58,
+0x1d,
+0x54,
+0xb0,
+0xad,
+0x33,
+0x9b,
+0x87,
+0x2b,
+0x4e,
+0xce,
+0x80,
+0x0c,
+0x63,
+0x8d,
+0x1c,
+0x22,
+0x55,
+0x95,
+0xc1,
+0x8c,
+0x9c,
+0x73,
+0xfd,
+0x3c,
+0x8d,
+0x28,
+0x04,
+0xe8,
+0x41,
+0x69,
+0x82,
+0x48,
+0xb2,
+0xa8,
+0xd8,
+0x89,
+0x04,
+0x36,
+0xff,
+0xd3,
+0x2c,
+0x9b,
+0x22,
+0x74,
+0x39,
+0xf7,
+0x0e,
+0x32,
+0xff,
+0x64,
+0x01,
+0x54,
+0x55,
+0x63,
+0xc7,
+0xd9,
+0x38,
+0xf5,
+0x6a,
+0xdd,
+0xf0,
+0x1e,
+0xe9,
+0x33,
+0x8f,
+0x11,
+0x67,
+0x00,
+0x21,
+0x38,
+0x22,
+0xd9,
+0xa6,
+0x29,
+0xc1,
+0x7b,
+0x8d,
+0xb7,
+0x31,
+0x80,
+0xd4,
+0xbc,
+0x3c,
+0xec,
+0x37,
+0x5e,
+0x8f,
+0x05,
+0x77,
+0x36,
+0x4b,
+0x7e,
+0xc2,
+0xd6,
+0x3a,
+0x73,
+0x4f,
+0x67,
+0x5b,
+0x73,
+0xca,
+0x60,
+0x73,
+0xff,
+0x5b,
+0x6c,
+0x82,
+0x7e,
+0x62,
+0x45,
+0xb5,
+0x0a,
+0xfe,
+0x1b,
+0xad,
+0x90,
+0xc7,
+0x48,
+0xe8,
+0x2b,
+0xf1,
+0xa4,
+0x7d,
+0x78,
+0x6b,
+0xcd,
+0x57,
+0x54,
+0x92,
+0xc0,
+0x57,
+0x91,
+0x4f,
+0xc0,
+0xc6,
+0x0a,
+0x23,
+0x9d,
+0x86,
+0xac,
+0x2c,
+0x93,
+0xed,
+0x88,
+0x6a,
+0x6c,
+0xe6,
+0xfa,
+0x66,
+0x1c,
+0xbf,
+0xc8,
+0x6b,
+0xb5,
+0xaa,
+0xda,
+0xf6,
+0xff,
+0xed,
+0x57,
+0xff,
+0x2e,
+0xa4,
+0xb6,
+0xb9,
+0x81,
+0x78,
+0xfb,
+0x43,
+0x5e,
+0x63,
+0x1d,
+0x9a,
+0xcb,
+0x13,
+0x61,
+0x2c,
+0x0c,
+0x76,
+0x01,
+0x5e,
+0x1e,
+0xc4,
+0xf6,
+0x1d,
+0xd9,
+0x2c,
+0xd6,
+0xbb,
+0x46,
+0x18,
+0xcb,
+0x94,
+0x21,
+0x57,
+0xb3,
+0x30,
+0x60,
+0x41,
+0xad,
+0xaa,
+0x09,
+0x80,
+0x6e,
+0x13,
+0xa3,
+0xf8,
+0x35,
+0x78,
+0xb9,
+0x90,
+0xa7,
+0x3e,
+0xb1,
+0x58,
+0x03,
+0xd4,
+0x62,
+0x43,
+0xda,
+0x3c,
+0x14,
+0x9f,
+0xc9,
+0xfa,
+0xd0,
+0x2f,
+0x5a,
+0xc0,
+0x1d,
+0x6e,
+0xb1,
+0x69,
+0xb0,
+0x64,
+0xb0,
+0x14,
+0xa1,
+0x39,
+0x8b,
+0x85,
+0xcd,
+0x37,
+0xaa,
+0x91,
+0x13,
+0xbd,
+0xcc,
+0xd7,
+0x77,
+0x1f,
+0x30,
+0x83,
+0x74,
+0xac,
+0xe1,
+0x40,
+0xff,
+0xaa,
+0xd3,
+0x3f,
+0x0a,
+0x2c,
+0xb4,
+0x25,
+0xa0,
+0x60,
+0x5c,
+0x30,
+0xb5,
+0xce,
+0x35,
+0xe2,
+0xcc,
+0x0d,
+0x89,
+0x7d,
+0x5d,
+0x7b,
+0x23,
+0xc8,
+0xb7,
+0x6c,
+0x10,
+0x93,
+0x16,
+0x6a,
+0x49,
+0xe3,
+0xe7,
+0x24,
+0x91,
+0xa2,
+0x2c,
+0x3c,
+0x78,
+0xb9,
+0x69,
+0xc3,
+0xb9,
+0xed,
+0xeb,
+0xed,
+0x51,
+0x67,
+0x42,
+0x00,
+0x50,
+0xa6,
+0x20,
+0x5f,
+0xcd,
+0x6e,
+0x9a,
+0x65,
+0x61,
+0x1f,
+0x9f,
+0xaf,
+0x69,
+0xf4,
+0xbb,
+0x57,
+0x62,
+0x4d,
+0xf1,
+0xcf,
+0xaf,
+0x89,
+0x38,
+0xe2,
+0x4d,
+0x7b,
+0x24,
+0x24,
+0x5e,
+0x23,
+0x0f,
+0x1b,
+0x73,
+0x2f,
+0x1e,
+0x2d,
+0x2a,
+0xf1,
+0x44,
+0x2f,
+0xe2,
+0x5d,
+0x07,
+0x48,
+0xaf,
+0xfd,
+0x42,
+0xf8,
+0x7b,
+0x1b,
+0x4e,
+0xe5,
+0xa6,
+0xc0,
+0xd6,
+0x66,
+0x15,
+0x4c,
+0xe2,
+0x39,
+0x44,
+0x95,
+0xf4,
+0x41,
+0x55,
+0x53,
+0x9d,
+0xd9,
+0x70,
+0xc1,
+0x58,
+0x1b,
+0x18,
+0x23,
+0x9d,
+0x06,
+0x11,
+0x60,
+0x43,
+0x66,
+0xa7,
+0x16,
+0xaf,
+0x5b,
+0x6c,
+0x27,
+0x89,
+0xbd,
+0xda,
+0x98,
+0x43,
+0xe4,
+0x41,
+0xde,
+0x92,
+0x78,
+0x22,
+0x74,
+0x40,
+0x6a,
+0xce,
+0x22,
+0x8c,
+0x67,
+0x5c,
+0xe1,
+0xba,
+0xe2,
+0x07,
+0x31,
+0xc5,
+0x21,
+0xa0,
+0x2f,
+0x02,
+0x04,
+0x10,
+0x74,
+0x36,
+0x43,
+0x33,
+0x9c,
+0xe7,
+0xe5,
+0xca,
+0x98,
+0xa5,
+0x14,
+0x0e,
+0x7d,
+0x5b,
+0x1c,
+0xe6,
+0x9a,
+0x11,
+0x81,
+0xe7,
+0x86,
+0x46,
+0x2c,
+0x2c,
+0x6d,
+0xc9,
+0x3f,
+0x1f,
+0xc8,
+0x42,
+0x4e,
+0xd0,
+0xf0,
+0x9d,
+0x5b,
+0x1b,
+0xb2,
+0x8c,
+0x84,
+0x4f,
+0x63,
+0xe0,
+0xd7,
+0x5d,
+0xd7,
+0xea,
+0x2f,
+0xea,
+0x8f,
+0x90,
+0x71,
+0x3b,
+0x0f,
+0xd9,
+0x88,
+0xbc,
+0x09,
+0xd8,
+0x02,
+0xe0,
+0xa3,
+0x30,
+0x57,
+0x5c,
+0x5d,
+0xa9,
+0xaf,
+0xb9,
+0x06,
+0xe4,
+0x62,
+0x67,
+0x1a,
+0x33,
+0xe3,
+0xb1,
+0x8d,
+0x70,
+0x01,
+0x69,
+0x14,
+0x1f,
+0x1a,
+0x08,
+0x5d,
+0x61,
+0x03,
+0x84,
+0xc9,
+0x4e,
+0xd8,
+0x6b,
+0xaf,
+0x26,
+0xe5,
+0x3f,
+0x3a,
+0x91,
+0x1b,
+0x2e,
+0xcd,
+0x27,
+0x26,
+0xb1,
+0x5b,
+0x20,
+0xc6,
+0xa2,
+0xec,
+0x30,
+0xfa,
+0xb3,
+0xa8,
+0xd0,
+0x63,
+0x73,
+0x84,
+0x0b,
+0x7d,
+0x02,
+0xa4,
+0x96,
+0x17,
+0xd4,
+0x2a,
+0x51,
+0x1f,
+0xb2,
+0x94,
+0xb5,
+0x94,
+0xb1,
+0x61,
+0xda,
+0x61,
+0xe9,
+0xdc,
+0x3f,
+0xf6,
+0x81,
+0x7a,
+0x36,
+0xbf,
+0x97,
+0xce,
+0x18,
+0xd8,
+0x38,
+0xb4,
+0x67,
+0xc4,
+0x4e,
+0x07,
+0x49,
+0x7f,
+0x21,
+0xf8,
+0xc1,
+0xec,
+0xa7,
+0x91,
+0xa5,
+0x92,
+0x21,
+0xcb,
+0xe3,
+0xd7,
+0xad,
+0x98,
+0xc5,
+0x6c,
+0x3e,
+0x1e,
+0x92,
+0x95,
+0x42,
+0x37,
+0xc6,
+0xf1,
+0xab,
+0xc6,
+0xdd,
+0x68,
+0x32,
+0xca,
+0x22,
+0x3a,
+0xa4,
+0x0e,
+0x62,
+0xd9,
+0x56,
+0xd0,
+0x0b,
+0x56,
+0x68,
+0x0e,
+0x98,
+0xea,
+0x08,
+0x94,
+0xe1,
+0x0d,
+0x51,
+0xfe,
+0x1e,
+0x91,
+0xdd,
+0xc1,
+0x38,
+0xab,
+0xe5,
+0x84,
+0x17,
+0x70,
+0x12,
+0x57,
+0xb8,
+0xc0,
+0x08,
+0x01,
+0xdf,
+0xa6,
+0xfb,
+0x30,
+0x5c,
+0xec,
+0x8a,
+0x75,
+0x47,
+0x67,
+0x01,
+0x39,
+0x22,
+0x85,
+0x6b,
+0x80,
+0xaf,
+0x42,
+0x7d,
+0xee,
+0x0a,
+0x69,
+0xbf,
+0xfb,
+0xca,
+0xec,
+0xc4,
+0x2a,
+0x5f,
+0x8d,
+0xfc,
+0x9b,
+0x56,
+0x91,
+0x89,
+0x14,
+0xb3,
+0xfe,
+0x40,
+0xcf,
+0x37,
+0xdd,
+0x60,
+0x54,
+0x9b,
+0xdd,
+0x48,
+0xcc,
+0x0d,
+0x16,
+0x47,
+0x71,
+0x5d,
+0x6a,
+0x6d,
+0x1a,
+0xf0,
+0xc8,
+0x80,
+0x5a,
+0xca,
+0xfa,
+0x93,
+0x18,
+0xd5,
+0xa8,
+0x73,
+0x6a,
+0x01,
+0x5d,
+0x16,
+0x9c,
+0x1b,
+0xc3,
+0xde,
+0xb7,
+0x2f,
+0x2a,
+0xbd,
+0xaa,
+0x49,
+0x8f,
+0xe4,
+0xe7,
+0x63,
+0x7e,
+0x68,
+0x15,
+0x4e,
+0xf7,
+0x10,
+0x54,
+0xd5,
+0x0c,
+0x0a,
+0x04,
+0x50,
+0x11,
+0x90,
+0xde,
+0x3e,
+0x15,
+0xfc,
+0x01,
+0x06,
+0xb1,
+0x7c,
+0x40,
+0x77,
+0xa6,
+0x42,
+0x93,
+0xd6,
+0x94,
+0x0a,
+0xca,
+0xfc,
+0xd8,
+0xce,
+0x5e,
+0x77,
+0x3e,
+0x98,
+0x6b,
+0x8a,
+0xf5,
+0x64,
+0x05,
+0x04,
+0x63,
+0xde,
+0x75,
+0x77,
+0xe5,
+0x1b,
+0x91,
+0x37,
+0xca,
+0xdb,
+0x06,
+0x2d,
+0x2f,
+0x28,
+0x4f,
+0x26,
+0xc2,
+0xa2,
+0x51,
+0x27,
+0xce,
+0xff,
+0x0e,
+0x7a,
+0x68,
+0x8f,
+0xc5,
+0x33,
+0x49,
+0xbf,
+0x21,
+0x92,
+0x2a,
+0x77,
+0x06,
+0x21,
+0xfa,
+0x96,
+0x12,
+0xdf,
+0x06,
+0xc6,
+0xec,
+0x98,
+0xc6,
+0x94,
+0xed,
+0xf7,
+0xd5,
+0x88,
+0xb1,
+0x6c,
+0xa2,
+0x89,
+0x0d,
+0xf8,
+0xf8,
+0x75,
+0xa3,
+0x7c,
+0xfc,
+0x81,
+0xb4,
+0x7c,
+0x44,
+0x43,
+0x28,
+0x69,
+0x5b,
+0x32,
+0x21,
+0x5d,
+0xc6,
+0x17,
+0xd9,
+0x53,
+0xc9,
+0x43,
+0x14,
+0xf4,
+0x00,
+0x6e,
+0x7d,
+0xe6,
+0x13,
+0xc2,
+0x88,
+0xc5,
+0x6a,
+0x7b,
+0x87,
+0xe4,
+0x75,
+0xdb,
+0x62,
+0xd4,
+0xd8,
+0x2a,
+0xeb,
+0x6c,
+0xf0,
+0x3e,
+0x09,
+0xaf,
+0x63,
+0x4b,
+0x8f,
+0xc5,
+0x92,
+0xb8,
+0x3f,
+0x28,
+0x4b,
+0xb5,
+0x20,
+0x55,
+0x69,
+0xec,
+0x22,
+0xeb,
+0x24,
+0x2b,
+0x20,
+0xc0,
+0x10,
+0x8c,
+0xe5,
+0x0e,
+0x27,
+0xcc,
+0x14,
+0x94,
+0xeb,
+0x07,
+0x01,
+0x70,
+0x73,
+0x35,
+0x5a,
+0xde,
+0xc7,
+0x1f,
+0xe4,
+0x0c,
+0x49,
+0xc6,
+0x10,
+0x75,
+0x74,
+0x25,
+0xc8,
+0x17,
+0x06,
+0x80,
+0xd7,
+0x0f,
+0xfd,
+0x80,
+0x3c,
+0x0c,
+0xe9,
+0xea,
+0x4e,
+0x6e,
+0x29,
+0xe2,
+0xeb,
+0x2c,
+0x42,
+0xa6,
+0xea,
+0xdc,
+0xf6,
+0x6a,
+0x3c,
+0x01,
+0x33,
+0x6a,
+0x96,
+0xca,
+0x5f,
+0xd6,
+0x13,
+0xbc,
+0x89,
+0xdf,
+0xaf,
+0x51,
+0x8d,
+0x83,
+0x57,
+0xe7,
+0x66,
+0xc1,
+0x0b,
+0x7a,
+0xee,
+0xff,
+0xee,
+0xd4,
+0x8a,
+0xd9,
+0xe6,
+0xca,
+0xb2,
+0x34,
+0xc2,
+0x43,
+0x73,
+0x5f,
+0x42,
+0x22,
+0xdf,
+0xcf,
+0xbb,
+0x2b,
+0x7b,
+0x41,
+0xb5,
+0xd6,
+0xa6,
+0x46,
+0xd3,
+0xb0,
+0x9b,
+0xf1,
+0x74,
+0xaa,
+0x3d,
+0x6f,
+0xc7,
+0x76,
+0x9a,
+0xde,
+0x73,
+0x7f,
+0x0d,
+0xc8,
+0xf5,
+0x0d,
+0xce,
+0xbd,
+0xa9,
+0xa0,
+0x2f,
+0x13,
+0x1b,
+0x92,
+0xc3,
+0x2a,
+0x03,
+0x39,
+0x94,
+0x51,
+0x42,
+0xbd,
+0xbc,
+0x26,
+0x31,
+0xa9,
+0x62,
+0x7c,
+0xc4,
+0xe6,
+0x23,
+0x11,
+0xe5,
+0x4c,
+0xb3,
+0x00,
+0xea,
+0x93,
+0x3e,
+0xea,
+0x43,
+0x11,
+0xce,
+0x67,
+0xa2,
+0x77,
+0x0d,
+0x82,
+0x8a,
+0x27,
+0xcb,
+0xba,
+0x7d,
+0x94,
+0xd5,
+0x4f,
+0x35,
+0xcf,
+0xf5,
+0x37,
+0x9f,
+0x60,
+0xa0,
+0xcd,
+0xf4,
+0x88,
+0xbb,
+0xcd,
+0xb4,
+0xe5,
+0x03,
+0x04,
+0x8f,
+0x04,
+0xef,
+0x83,
+0xf7,
+0x82,
+0x7d,
+0x16,
+0x02,
+0xfb,
+0xef,
+0x83,
+0x1a,
+0x26,
+0xb4,
+0x8a,
+0xd5,
+0x18,
+0xcd,
+0x05,
+0xaf,
+0x6c,
+0xd3,
+0xa8,
+0x24,
+0xf3,
+0xc1,
+0x34,
+0xf3,
+0xa0,
+0x1a,
+0x38,
+0x81,
+0x1d,
+0x85,
+0x53,
+0x47,
+0xee,
+0xc4,
+0x88,
+0x50,
+0x18,
+0xb6,
+0x7c,
+0x0e,
+0xbc,
+0xa1,
+0xf5,
+0x74,
+0x56,
+0xf1,
+0xdc,
+0x81,
+0xb8,
+0x90,
+0x78,
+0x28,
+0x41,
+0x4b,
+0xea,
+0x83,
+0x56,
+0x6b,
+0x64,
+0xf7,
+0x18,
+0xe3,
+0xc7,
+0x3b,
+0xe2,
+0xfd,
+0xdd,
+0xc4,
+0xd1,
+0xc1,
+0xa5,
+0x4b,
+0x40,
+0x8b,
+0xef,
+0x67,
+0x87,
+0x5f,
+0x26,
+0x42,
+0x61,
+0xb9,
+0xb6,
+0x7a,
+0x44,
+0x3d,
+0xc7,
+0x17,
+0x22,
+0xa1,
+0x6b,
+0x92,
+0x8f,
+0x70,
+0x9a,
+0x5f,
+0x6f,
+0x2a,
+0x3a,
+0xaa,
+0xcf,
+0x82,
+0x96,
+0x44,
+0x13,
+0xc6,
+0x29,
+0xdf,
+0x45,
+0x5d,
+0xc0,
+0x6a,
+0x0e,
+0x3e,
+0x84,
+0x23,
+0xb5,
+0x77,
+0x4f,
+0x92,
+0x92,
+0xb0,
+0x24,
+0x7e,
+0x9a,
+0x94,
+0xdb,
+0xa9,
+0xb2,
+0x8d,
+0x24,
+0xe7,
+0x32,
+0xa9,
+0x18,
+0x28,
+0xcb,
+0xfa,
+0x66,
+0xfe,
+0xde,
+0xc2,
+0xde,
+0x49,
+0x22,
+0x36,
+0x5b,
+0x63,
+0xc8,
+0xeb,
+0x38,
+0x33,
+0x76,
+0x5d,
+0xe1,
+0xb8,
+0xd4,
+0xb6,
+0x94,
+0xb0,
+0x4a,
+0x13,
+0xab,
+0x0d,
+0x80,
+0xb6,
+0x9c,
+0xf8,
+0xb9,
+0xb3,
+0x9b,
+0x8c,
+0x96,
+0x5b,
+0x07,
+0x04,
+0x0c,
+0x6f,
+0xc1,
+0x69,
+0x15,
+0x86,
+0xc5,
+0x55,
+0xf4,
+0x9b,
+0xdb,
+0x67,
+0xbf,
+0x47,
+0xa0,
+0xa5,
+0xe5,
+0x00,
+0x97,
+0x65,
+0x1b,
+0xfd,
+0x0a,
+0xcb,
+0x7c,
+0x3b,
+0xde,
+0x30,
+0x76,
+0x80,
+0x09,
+0xfb,
+0xcd,
+0xb7,
+0xbb,
+0x0f,
+0x85,
+0xb8,
+0x2b,
+0x62,
+0xe1,
+0xf8,
+0xd0,
+0x91,
+0x1e,
+0x03,
+0x1a,
+0x76,
+0x41,
+0x4a,
+0xe6,
+0x89,
+0x49,
+0xa3,
+0x9d,
+0xa6,
+0xd8,
+0xb7,
+0x74,
+0xc5,
+0xa1,
+0x90,
+0xf7,
+0xe9,
+0x4f,
+0xd0,
+0x6f,
+0xcf,
+0x92,
+0x56,
+0xdf,
+0x4f,
+0xb9,
+0x1e,
+0x5c,
+0x0f,
+0x2c,
+0x78,
+0x0e,
+0x6f,
+0x62,
+0x3e,
+0x81,
+0x64,
+0x45,
+0x30,
+0x43,
+0x2b,
+0xa9,
+0x47,
+0x09,
+0x30,
+0xf8,
+0xb9,
+0xa8,
+0xe4,
+0x25,
+0x1b,
+0x27,
+0xe6,
+0x38,
+0xce,
+0x84,
+0x63,
+0x7c,
+0xa6,
+0x67,
+0xb2,
+0x2e,
+0xce,
+0xce,
+0xac,
+0x5c,
+0x88,
+0xc0,
+0x78,
+0x87,
+0x97,
+0x82,
+0xc8,
+0xad,
+0x5d,
+0x28,
+0xb8,
+0x4a,
+0x83,
+0x41,
+0xbc,
+0xec,
+0x63,
+0xfa,
+0x61,
+0x29,
+0x48,
+0xb2,
+0xa7,
+0x02,
+0x38,
+0xc8,
+0x6b,
+0xac,
+0x49,
+0x3d,
+0x0d,
+0x6a,
+0x2c,
+0x0b,
+0x70,
+0xfe,
+0x51,
+0xd9,
+0x89,
+0x00,
+0x3b,
+0xdc,
+0x45,
+0x0f,
+0x25,
+0xe5,
+0xaa,
+0xee,
+0x8c,
+0x55,
+0xd0,
+0x22,
+0x4c,
+0x7f,
+0x9a,
+0xb3,
+0x7e,
+0xce,
+0x74,
+0xa2,
+0x33,
+0x39,
+0x0a,
+0x9c,
+0xe7,
+0x2e,
+0x96,
+0x6e,
+0x91,
+0x2e,
+0x2c,
+0xcb,
+0xc6,
+0x03,
+0x29,
+0x08,
+0x01,
+0xaa,
+0xfc,
+0xef,
+0x20,
+0x9d,
+0xcf,
+0xa1,
+0xcf,
+0x8a,
+0x91,
+0x64,
+0x4b,
+0x4f,
+0xbf,
+0xcb,
+0x3b,
+0x42,
+0xe7,
+0x14,
+0xce,
+0x1e,
+0xb0,
+0x05,
+0xaa,
+0x30,
+0xc2,
+0x06,
+0xcd,
+0x1c,
+0xba,
+0x5d,
+0xd7,
+0x4b,
+0x32,
+0x87,
+0xc1,
+0xb9,
+0x43,
+0x20,
+0xc2,
+0xbc,
+0x32,
+0x80,
+0x1c,
+0xaa,
+0x31,
+0x7d,
+0xbb,
+0xac,
+0x6f,
+0x30,
+0x24,
+0xbf,
+0x02,
+0xdc,
+0xfc,
+0xb0,
+0xbb,
+0x9f,
+0x31,
+0xb1,
+0x3a,
+0xad,
+0x1e,
+0x3d,
+0xdf,
+0xa5,
+0x4b,
+0x17,
+0x42,
+0x8d,
+0x45,
+0x6f,
+0xb9,
+0x4f,
+0x36,
+0xcd,
+0x92,
+0xbc,
+0x4e,
+0x65,
+0xf3,
+0x0b,
+0x6b,
+0x58,
+0x92,
+0xfd,
+0x07,
+0xa5,
+0x18,
+0x69,
+0x32,
+0x8d,
+0xea,
+0x01,
+0xe1,
+0x4b,
+0xcd,
+0x6a,
+0x83,
+0x0d,
+0x31,
+0x58,
+0x11,
+0x78,
+0x1e,
+0x43,
+0xa9,
+0xdb,
+0xf0,
+0x4d,
+0xe1,
+0xb2,
+0x3e,
+0x69,
+0xb2,
+0xc3,
+0x25,
+0x3b,
+0x9b,
+0x68,
+0xc9,
+0xa0,
+0x31,
+0x91,
+0x48,
+0xe6,
+0xb8,
+0x1d,
+0xc1,
+0x82,
+0x86,
+0xfb,
+0x41,
+0xc7,
+0xfb,
+0x89,
+0xf1,
+0x87,
+0x48,
+0x21,
+0xbf,
+0x91,
+0x49,
+0x2a,
+0xc1,
+0xb0,
+0xa5,
+0x65,
+0x96,
+0x96,
+0xbf,
+0x3a,
+0xe7,
+0x94,
+0x7f,
+0xbf,
+0x80,
+0xbb,
+0x88,
+0xcd,
+0x5b,
+0x50,
+0x9b,
+0xb7,
+0xfd,
+0xb6,
+0x48,
+0xbc,
+0x3a,
+0xc7,
+0x22,
+0xb8,
+0xcb,
+0x76,
+0x5d,
+0x85,
+0x63,
+0x0e,
+0x70,
+0x5c,
+0x65,
+0xce,
+0x6d,
+0xb9,
+0x70,
+0x94,
+0xd9,
+0xaf,
+0x3c,
+0x42,
+0x35,
+0x92,
+0x6c,
+0xa1,
+0x82,
+0x02,
+0xe8,
+0x7e,
+0x9f,
+0xfc,
+0xbd,
+0x44,
+0x68,
+0x26,
+0xb8,
+0xa0,
+0xb0,
+0x23,
+0x83,
+0x42,
+0xe7,
+0x6a,
+0x11,
+0x2f,
+0xc6,
+0x73,
+0xb9,
+0xb8,
+0x37,
+0xaa,
+0x5b,
+0x2c,
+0x1e,
+0xfa,
+0x9b,
+0xbc,
+0x17,
+0x89,
+0xe4,
+0x74,
+0x61,
+0x37,
+0x15,
+0xb0,
+0x8f,
+0xd8,
+0xd8,
+0x03,
+0x1e,
+0xa6,
+0xbd,
+0x23,
+0x02,
+0x69,
+0x00,
+0x4d,
+0x4c,
+0x57,
+0x93,
+0xf0,
+0x45,
+0xb8,
+0x49,
+0xfa,
+0xe3,
+0xee,
+0x66,
+0x74,
+0x8b,
+0xc0,
+0x6e,
+0x54,
+0x7c,
+0xf5,
+0x05,
+0xea,
+0xad,
+0x85,
+0xee,
+0xd6,
+0x5b,
+0xc1,
+0x11,
+0xdc,
+0x16,
+0xb5,
+0x31,
+0x49,
+0xe4,
+0xa3,
+0x84,
+0x9b,
+0x7d,
+0x2d,
+0x46,
+0x7b,
+0x2d,
+0x70,
+0x49,
+0x4e,
+0x06,
+0xe8,
+0x07,
+0x26,
+0xba,
+0xc1,
+0xa2,
+0x0d,
+0x29,
+0x63,
+0xa6,
+0x15,
+0x48,
+0xb8,
+0x02,
+0xa6,
+0x03,
+0x6c,
+0x09,
+0xa1,
+0xdd,
+0x6f,
+0x2d,
+0x1f,
+0x1f,
+0xda,
+0xbd,
+0x00,
+0x96,
+0x23,
+0x15,
+0x14,
+0x47,
+0xe1,
+0x22,
+0xf5,
+0x25,
+0xab,
+0x24,
+0x28,
+0x6b,
+0x15,
+0x37,
+0x32,
+0xd1,
+0x92,
+0xc3,
+0x01,
+0x48,
+0xdd,
+0x78,
+0x1b,
+0x4b,
+0xd1,
+0xab,
+0xb6,
+0x35,
+0xf5,
+0xba,
+0x00,
+0x1c,
+0xf7,
+0x7b,
+0xad,
+0x1b,
+0x29,
+0x05,
+0x47,
+0x75,
+0xb0,
+0x05,
+0xb2,
+0xb7,
+0xb0,
+0xe5,
+0xcb,
+0x17,
+0xa9,
+0xb3,
+0x0f,
+0x7f,
+0x08,
+0x0f,
+0x5a,
+0x23,
+0xfd,
+0x20,
+0xed,
+0x50,
+0xd1,
+0x9c,
+0xf6,
+0x23,
+0x81,
+0x6d,
+0xa0,
+0x42,
+0xce,
+0x39,
+0x4f,
+0x6f,
+0x5c,
+0x51,
+0xde,
+0xb2,
+0x69,
+0x8f,
+0x50,
+0x44,
+0xa8,
+0x43,
+0x4e,
+0xa5,
+0x31,
+0xb9,
+0x96,
+0x02,
+0xce,
+0x84,
+0x77,
+0xe5,
+0x13,
+0x14,
+0x6b,
+0x20,
+0xdd,
+0x1c,
+0x4d,
+0xeb,
+0x1b,
+0x86,
+0x17,
+0x2f,
+0xd1,
+0x3c,
+0x68,
+0x4e,
+0x57,
+0xca,
+0x4d,
+0x4f,
+0x0b,
+0x37,
+0xb0,
+0x0e,
+0x93,
+0xee,
+0x84,
+0x78,
+0xe7,
+0xeb,
+0x33,
+0xeb,
+0x9a,
+0x68,
+0x9b,
+0x0d,
+0x39,
+0x91,
+0xa3,
+0x51,
+0xe6,
+0x94,
+0xe7,
+0xab,
+0x9d,
+0x18,
+0x9d,
+0x05,
+0xcb,
+0x86,
+0x3f,
+0x12,
+0x7d,
+0xf4,
+0x31,
+0x05,
+0xd9,
+0x34,
+0x14,
+0xd9,
+0xf0,
+0xf4,
+0x38,
+0x5c,
+0x37,
+0xf4,
+0x80,
+0xd4,
+0x29,
+0x7d,
+0x08,
+0x31,
+0x3a,
+0x11,
+0x9c,
+0x38,
+0xb5,
+0xcc,
+0x0b,
+0x0b,
+0x2c,
+0xfb,
+0x87,
+0x31,
+0xb5,
+0x16,
+0x0e,
+0xb6,
+0x6f,
+0x73,
+0x73,
+0x76,
+0x4b,
+0x39,
+0x39,
+0x4f,
+0xf7,
+0xf4,
+0x3c,
+0xc2,
+0x96,
+0x2a,
+0x2d,
+0xb7,
+0x37,
+0xbb,
+0x06,
+0x5f,
+0xd9,
+0xe2,
+0xb9,
+0xe6,
+0x4f,
+0xb5,
+0x25,
+0x58,
+0x22,
+0x18,
+0x95,
+0x84,
+0x8c,
+0x60,
+0x71,
+0xba,
+0x4c,
+0xca,
+0xe8,
+0xc7,
+0x03,
+0x96,
+0x08,
+0x5c,
+0x69,
+0x1d,
+0xbb,
+0x40,
+0x3a,
+0x0a,
+0x97,
+0x5a,
+0x77,
+0x53,
+0xdd,
+0x18,
+0x4b,
+0x2c,
+0x35,
+0xc9,
+0x6d,
+0x01,
+0x2b,
+0xc4,
+0xcc,
+0x45,
+0x47,
+0xe1,
+0xd1,
+0xdb,
+0xe7,
+0x6c,
+0x61,
+0x05,
+0xb6,
+0x08,
+0xcf,
+0x67,
+0xde,
+0x38,
+0x0d,
+0xe6,
+0x46,
+0xaa,
+0x66,
+0x73,
+0x35,
+0x81,
+0x90,
+0xe6,
+0x78,
+0x46,
+0xda,
+0x8d,
+0xa9,
+0x44,
+0xd8,
+0x13,
+0xce,
+0x8b,
+0x4d,
+0x75,
+0xa7,
+0xc0,
+0x61,
+0x9b,
+0x10,
+0x24,
+0xb6,
+0xdf,
+0xcf,
+0x82,
+0x5d,
+0x84,
+0x77,
+0x14,
+0xe4,
+0xe9,
+0xe6,
+0x58,
+0x49,
+0x35,
+0x06,
+0xc9,
+0x06,
+0x0a,
+0x32,
+0x2a,
+0x90,
+0x62,
+0xa5,
+0xd9,
+0xde,
+0x3d,
+0x38,
+0xdc,
+0x6b,
+0x80,
+0x3a,
+0x2e,
+0xd6,
+0x9b,
+0x0f,
+0x2a,
+0xa7,
+0xc1,
+0x47,
+0xe7,
+0x37,
+0xba,
+0x32,
+0xaa,
+0xff,
+0x78,
+0x6b,
+0x88,
+0x97,
+0x6f,
+0xa5,
+0x49,
+0xc5,
+0x83,
+0xc5,
+0x9f,
+0x4c,
+0x87,
+0xa5,
+0x67,
+0xd8,
+0x5b,
+0x81,
+0xf4,
+0x2e,
+0x15,
+0xb4,
+0x71,
+0xe3,
+0x39,
+0x2f,
+0x39,
+0xc6,
+0xba,
+0xdd,
+0xcb,
+0x0e,
+0xb1,
+0x70,
+0xd5,
+0x69,
+0xc3,
+0xd4,
+0x18,
+0x3d,
+0x2c,
+0xc4,
+0x2d,
+0x6e,
+0xf1,
+0x04,
+0x88,
+0x13,
+0x0f,
+0xba,
+0x49,
+0x2b,
+0x47,
+0x5b,
+0x44,
+0x5d,
+0xeb,
+0x6d,
+0x2e,
+0x38,
+0x77,
+0xd3,
+0xf5,
+0xdb,
+0x54,
+0x1d,
+0xac,
+0x4e,
+0xb6,
+0x7d,
+0xa1,
+0xe4,
+0xd7,
+0x01,
+0xbb,
+0x16,
+0xec,
+0x4e,
+0xef,
+0xe9,
+0xa2,
+0x6d,
+0x4a,
+0x6c,
+0x57,
+0xd5,
+0x10,
+0x7e,
+0x5d,
+0xb7,
+0x18,
+0x94,
+0x70,
+0x9a,
+0xd5,
+0x92,
+0xaf,
+0x84,
+0x74,
+0x69,
+0xb1,
+0x50,
+0x98,
+0xa2,
+0x1d,
+0x0e,
+0xa5,
+0x0b,
+0x86,
+0x1b,
+0xba,
+0x10,
+0xa3,
+0x70,
+0x61,
+0x81,
+0xba,
+0xc8,
+0xdd,
+0x3f,
+0x4f,
+0x76,
+0x07,
+0x43,
+0x05,
+0x2a,
+0x76,
+0xad,
+0x49,
+0xaa,
+0x3b,
+0x3c,
+0x71,
+0x5d,
+0xe0,
+0xc6,
+0xcf,
+0x91,
+0x94,
+0x65,
+0x3e,
+0x7d,
+0x95,
+0x6f,
+0x43,
+0x33,
+0xbf,
+0x24,
+0x85,
+0x18,
+0xbf,
+0x81,
+0xa3,
+0xfc,
+0x2a,
+0x4a,
+0xf4,
+0x94,
+0x21,
+0x4f,
+0x1f,
+0x81,
+0x4c,
+0x6e,
+0xa2,
+0x24,
+0x25,
+0x92,
+0x7a,
+0x89,
+0xdb,
+0x06,
+0x8a,
+0xe5,
+0xbf,
+0x20,
+0xd3,
+0x0c,
+0xc1,
+0x89,
+0xc4,
+0x09,
+0x82,
+0xbf,
+0x5a,
+0xff,
+0x30,
+0x4d,
+0xe4,
+0xf5,
+0x47,
+0x48,
+0xcf,
+0x38,
+0x36,
+0xc7,
+0x69,
+0x62,
+0x4f,
+0x8d,
+0x40,
+0x35,
+0xa2,
+0xd7,
+0xcb,
+0xec,
+0x82,
+0xbc,
+0x65,
+0xb1,
+0x70,
+0x26,
+0x8b,
+0x03,
+0xee,
+0x4c,
+0x45,
+0xad,
+0xc5,
+0x45,
+0xc7,
+0x2e,
+0xaa,
+0x2f,
+0x8c,
+0x61,
+0xf5,
+0xb8,
+0xc1,
+0x96,
+0xa3,
+0xbc,
+0x15,
+0xaa,
+0x4b,
+0x5c,
+0x29,
+0x62,
+0x5c,
+0x22,
+0x6f,
+0x7d,
+0x61,
+0x20,
+0xe0,
+0xa1,
+0xf8,
+0x5e,
+0x98,
+0xc9,
+0x74,
+0x07,
+0x1f,
+0xc2,
+0xab,
+0x11,
+0x1b,
+0xe6,
+0x2e,
+0x84,
+0x08,
+0xde,
+0x58,
+0x51,
+0x0e,
+0x1b,
+0x5c,
+0x66,
+0xab,
+0x36,
+0x07,
+0xd5,
+0x85,
+0xa4,
+0x3b,
+0x5c,
+0x89,
+0x98,
+0x7d,
+0xfd,
+0xa7,
+0xe2,
+0x08,
+0x4d,
+0xfb,
+0xa3,
+0x57,
+0xc0,
+0x47,
+0x18,
+0xf8,
+0x1a,
+0x33,
+0x8a,
+0x87,
+0xba,
+0x90,
+0x9d,
+0x44,
+0xdd,
+0x0e,
+0x50,
+0xa8,
+0xe4,
+0xa8,
+0x95,
+0x34,
+0x32,
+0xe1,
+0xcb,
+0x68,
+0x96,
+0x12,
+0x37,
+0x7d,
+0xab,
+0xf7,
+0x0d,
+0x1b,
+0x40,
+0x43,
+0x9d,
+0x7d,
+0xb4,
+0xbd,
+0xbd,
+0x21,
+0x34,
+0x9a,
+0xe9,
+0x77,
+0x9a,
+0x0c,
+0x08,
+0xde,
+0x71,
+0x80,
+0x56,
+0x4d,
+0xb4,
+0x2f,
+0x85,
+0xe4,
+0x43,
+0xd5,
+0x7d,
+0xc1,
+0xc5,
+0x0f,
+0x18,
+0xea,
+0x45,
+0x61,
+0x4c,
+0x63,
+0x56,
+0x8b,
+0x09
+};
+
+
+
+/* -----------EXPECTED OUTPUTS----------- */
+uint8_t exp_outp_0 [] = {
+0x2d,
+0x17,
+0x07,
+0x03,
+0x05,
+0x21,
+0x05,
+0x07,
+0x00,
+0x13,
+0x05,
+0x0b,
+0x8a,
+0x27,
+0x00,
+0x03,
+0x07,
+0x08,
+0x07,
+0x07,
+0x05,
+0x00,
+0x05,
+0x27,
+0x04,
+0x07,
+0x07,
+0x06,
+0x07,
+0x21,
+0x07,
+0x07,
+0x12,
+0x0b,
+0x07,
+0x07,
+0x00,
+0x4c,
+0x27,
+0x07,
+0x08,
+0x08,
+0x8a,
+0x06,
+0x0c,
+0x87,
+0x05,
+0x05,
+0x17,
+0x07,
+0x05,
+0x07,
+0x00,
+0x07,
+0x56,
+0x07,
+0x05,
+0x04,
+0x4c,
+0x08,
+0x07,
+0x03,
+0x04,
+0x27,
+0x04,
+0x07,
+0x07,
+0x04,
+0x07,
+0x08,
+0x04,
+0x08,
+0x07,
+0x4c,
+0x8b,
+0x04,
+0x07,
+0x07,
+0x06,
+0x84,
+0x04,
+0x07,
+0x07,
+0x05,
+0x07,
+0x4c,
+0x04,
+0x04,
+0x07,
+0x07,
+0x07,
+0x0b,
+0x00,
+0x07,
+0x00,
+0x08,
+0x07,
+0x07,
+0x03,
+0x07,
+0x07,
+0x07,
+0x07,
+0x05,
+0x07,
+0x0e,
+0x07,
+0x07,
+0x05,
+0x00,
+0x07,
+0x00,
+0x07,
+0x05,
+0x07,
+0x07,
+0x4e,
+0x0a,
+0x9a,
+0x00,
+0x00,
+0x22,
+0x0b,
+0x0e,
+0x07,
+0x21,
+0x07,
+0x17,
+0x07,
+0x05,
+0x0d,
+0x07,
+0x07,
+0x01,
+0x04,
+0x07,
+0x07,
+0x07,
+0x05,
+0x07,
+0x07,
+0x03,
+0x07,
+0x05,
+0x21,
+0x07,
+0x8b,
+0x0c,
+0x07,
+0x05,
+0x06,
+0x07,
+0x04,
+0x21,
+0x08,
+0x06,
+0x07,
+0x03,
+0x04,
+0x00,
+0x03,
+0x0a,
+0x03,
+0x04,
+0x04,
+0x0b,
+0x07,
+0x16,
+0x03,
+0x05,
+0x04,
+0x00,
+0x03,
+0x21,
+0x0a,
+0x07,
+0x07,
+0x8a,
+0x07,
+0x00,
+0x07,
+0x07,
+0x07,
+0x07,
+0x16,
+0x05,
+0x05,
+0x00,
+0x07,
+0x07,
+0x01,
+0x00,
+0x2d,
+0x08,
+0x0f,
+0x07,
+0x0d,
+0x07,
+0x07,
+0x07,
+0x8b,
+0x23,
+0x00,
+0x00,
+0x05,
+0x0d,
+0x08,
+0x87,
+0x08,
+0x08,
+0x04,
+0x07,
+0x04,
+0x24,
+0xcd,
+0x4c,
+0x0a,
+0x07,
+0x07,
+0x07,
+0x07,
+0x04,
+0x03,
+0x08,
+0x07,
+0x07,
+0x07,
+0x0b,
+0x04,
+0x00,
+0x04,
+0x00,
+0x03,
+0x07,
+0x07,
+0x00,
+0x85,
+0x07,
+0x03,
+0x07,
+0x00,
+0x07,
+0x03,
+0x04,
+0x02,
+0x07,
+0x07,
+0x05,
+0x21,
+0x08,
+0x07,
+0x07,
+0x00,
+0x21,
+0x04,
+0x07,
+0x05,
+0x00,
+0x07,
+0x27,
+0x07,
+0x08,
+0x4c,
+0x07,
+0x01,
+0x03,
+0x01,
+0x0b,
+0x03,
+0x00,
+0x03,
+0x00,
+0x07,
+0x05,
+0x05,
+0x80,
+0x21,
+0x8b,
+0x07,
+0x07,
+0x07,
+0x07,
+0x0c,
+0x05,
+0x03,
+0x07,
+0x07,
+0x07,
+0x07,
+0x07,
+0x4c,
+0x8a,
+0x21,
+0x07,
+0x07,
+0x07,
+0x00,
+0x07,
+0x00,
+0x04,
+0x07,
+0x00,
+0x8a,
+0x03,
+0x04,
+0x07,
+0x03,
+0x0c,
+0x07,
+0x07,
+0x04,
+0x04,
+0x0a,
+0x08,
+0x07,
+0x00,
+0x09,
+0x03,
+0x03,
+0x00,
+0x07,
+0xcd,
+0x0c,
+0x07,
+0x4c,
+0x00,
+0x07,
+0x00,
+0x00,
+0x07,
+0x21,
+0x5e,
+0x07,
+0x07,
+0x08,
+0x22,
+0x03,
+0x07,
+0x07,
+0x21,
+0x40,
+0x07,
+0xc9,
+0x0c,
+0x04,
+0x07,
+0x0e,
+0x07,
+0x07,
+0x05,
+0x07,
+0x05,
+0x05,
+0x05,
+0x01,
+0xcd,
+0x00,
+0x0c,
+0x22,
+0x8a,
+0x08,
+0x03,
+0x00,
+0x16,
+0x8b,
+0x21,
+0x17,
+0x0d,
+0x5e,
+0x07,
+0x03,
+0x06,
+0x27,
+0x27,
+0x07,
+0x8d,
+0x00,
+0x07,
+0x07,
+0x07,
+0x01,
+0x22,
+0x8b,
+0x00,
+0x07,
+0x07,
+0x03,
+0x21,
+0x0a,
+0x01,
+0x4d,
+0x04,
+0x07,
+0x07,
+0x18,
+0x17,
+0x08,
+0x0b,
+0x03,
+0x8b,
+0x07,
+0x05,
+0x07,
+0x07,
+0x07,
+0x07,
+0x08,
+0x47,
+0xa9,
+0x07,
+0x04,
+0x07,
+0x05,
+0x07,
+0x00,
+0x05,
+0x07,
+0x8b,
+0x07,
+0x07,
+0x00,
+0x07,
+0x06,
+0x0b,
+0x05,
+0x0a,
+0x07,
+0x07,
+0x8b,
+0x08,
+0x07,
+0x07,
+0x07,
+0x07,
+0x04,
+0x21,
+0x8b,
+0x04,
+0x08,
+0x08,
+0x16,
+0x03,
+0x05,
+0x03,
+0x07,
+0x27,
+0x0b,
+0x16,
+0x00,
+0x17,
+0x07,
+0x07,
+0x05,
+0x04,
+0x07,
+0x07,
+0x07,
+0x05,
+0x07,
+0x8a,
+0x05,
+0x06,
+0x16,
+0x07,
+0x00,
+0x07,
+0x07,
+0x0c,
+0x08,
+0x06,
+0x2b,
+0x06,
+0x07,
+0x05,
+0x0f,
+0x0c,
+0x22,
+0x0c,
+0x08,
+0x05,
+0x07,
+0x04,
+0x07,
+0x01,
+0x07,
+0x03,
+0x03,
+0x22,
+0x07,
+0x07,
+0x0b,
+0x04,
+0x00,
+0x05,
+0x07,
+0x8a,
+0x03,
+0x07,
+0x00,
+0x07,
+0x08,
+0x04,
+0x07,
+0x07,
+0x06,
+0x0c,
+0x8a,
+0x03,
+0x05,
+0x04,
+0x07,
+0x08,
+0x04,
+0x0a,
+0x07,
+0x00,
+0x07,
+0x00,
+0x08,
+0x00,
+0x04,
+0x01,
+0x07,
+0x21,
+0x08,
+0x00,
+0x07,
+0x08,
+0x07,
+0x07,
+0x07,
+0x07,
+0x0e,
+0x08,
+0x08,
+0x0b,
+0x07,
+0x04,
+0x07,
+0x03,
+0x07,
+0x00,
+0x44,
+0x07,
+0x00,
+0x07,
+0xcd,
+0x4d,
+0x07,
+0x05,
+0x05,
+0x07,
+0x07,
+0x08,
+0x00,
+0x05,
+0x04,
+0x0a,
+0x08,
+0x8a,
+0x07,
+0x07,
+0x03,
+0x4d,
+0x0e,
+0x05,
+0x04,
+0x4e,
+0x05,
+0x05,
+0x07,
+0x0b,
+0x07,
+0x08,
+0x00,
+0x01,
+0x0c,
+0x06,
+0x07,
+0x07,
+0x07,
+0x04,
+0x03,
+0x07,
+0x05,
+0x47,
+0x08,
+0x07,
+0x45,
+0x01,
+0x07,
+0x07,
+0x07,
+0x07,
+0x03,
+0x16,
+0x02,
+0x07,
+0x21,
+0x07,
+0x04,
+0x03,
+0x07,
+0x07,
+0x0a,
+0x07,
+0x08,
+0x08,
+0x03,
+0x08,
+0x8b,
+0x0b,
+0x05,
+0x05,
+0x00,
+0x07,
+0x07,
+0x07,
+0x07,
+0x05,
+0x07,
+0x17,
+0x07,
+0x07,
+0x07,
+0x4c,
+0x07,
+0x07,
+0x00,
+0x3d,
+0x07,
+0x87,
+0x03,
+0x00,
+0x00,
+0x08,
+0x04,
+0x07,
+0x21,
+0x05,
+0x21,
+0x07,
+0x03,
+0x05,
+0x07,
+0x04,
+0x49,
+0x04,
+0x04,
+0x05,
+0x07,
+0x07,
+0x49,
+0x0c,
+0x07,
+0x4c,
+0x07,
+0x03,
+0x8a,
+0x07,
+0x04,
+0x0e,
+0x0e,
+0x07,
+0x87,
+0x8b,
+0x04,
+0x04,
+0x07,
+0x04,
+0x07,
+0x08,
+0x08,
+0x04,
+0x04,
+0x0b,
+0x0e,
+0x07,
+0x07,
+0x08,
+0x4e,
+0x07,
+0x0b,
+0x07,
+0x07,
+0x05,
+0x07,
+0x0b,
+0x07,
+0x13,
+0x08,
+0x00,
+0x07,
+0x06,
+0x07,
+0x08,
+0x04,
+0x01,
+0x00,
+0x07,
+0x0a,
+0x03,
+0x07,
+0x05,
+0x47,
+0x8b,
+0x07,
+0x4a,
+0x07,
+0x00,
+0x07,
+0x07,
+0x03,
+0x4c,
+0x0a,
+0x03,
+0x05,
+0x8c,
+0x0b,
+0x07,
+0x03,
+0x0a,
+0x03,
+0x8a,
+0x00,
+0x8b,
+0x21,
+0x07,
+0x00,
+0x13,
+0x21,
+0x07,
+0x0c,
+0x08,
+0x0f,
+0x0f,
+0x07,
+0x1b,
+0x0c,
+0x07,
+0x07,
+0x07,
+0x00,
+0x07,
+0x01,
+0x04,
+0x07,
+0x04,
+0x03,
+0x07,
+0x07,
+0x00,
+0x07,
+0x10,
+0x04,
+0x03,
+0x04,
+0x00,
+0x0b,
+0x08,
+0x00,
+0x06,
+0x07,
+0x00,
+0x27,
+0x8b,
+0x04,
+0x03,
+0x00,
+0x8b,
+0x00,
+0x08,
+0x08,
+0x05,
+0x08,
+0x21,
+0x07,
+0x05,
+0x03,
+0x05,
+0x07,
+0x05,
+0x07,
+0x07,
+0x07,
+0x07,
+0x07,
+0x21,
+0x07,
+0x05,
+0x0a,
+0x0b,
+0x07,
+0x04,
+0x08,
+0x8b,
+0x16,
+0x07,
+0x07,
+0x4e,
+0x04,
+0x0a,
+0x17,
+0x05,
+0x8a,
+0x08,
+0x07,
+0x05,
+0x8b,
+0x07,
+0x03,
+0x05,
+0x05,
+0x07,
+0x07,
+0x07,
+0x07,
+0x07,
+0x08,
+0x08,
+0x05,
+0x0b,
+0x08,
+0x07,
+0x87,
+0x21,
+0x00,
+0x07,
+0x07,
+0x0a,
+0x00,
+0x01,
+0x07,
+0x03,
+0x07,
+0x4d,
+0x00,
+0x00,
+0x5e,
+0x04,
+0x02,
+0x04,
+0x0b,
+0x03,
+0x07,
+0x00,
+0x03,
+0x00,
+0x07,
+0x07,
+0x0e,
+0x22,
+0x07,
+0x07,
+0x07,
+0x04,
+0x16,
+0xa9,
+0x8a,
+0x27,
+0x4d,
+0x07,
+0x04,
+0x07,
+0x21,
+0x07,
+0x07,
+0x0e,
+0x05,
+0x07,
+0x07,
+0x08,
+0x0f,
+0x0c,
+0x08,
+0x8a,
+0x08,
+0x40,
+0x00,
+0x0a,
+0x08,
+0x07,
+0x8a,
+0x0e,
+0x05,
+0x05,
+0x0a,
+0x03,
+0xa2,
+0x05,
+0x07,
+0x07,
+0x08,
+0x07,
+0x03,
+0x07,
+0x04,
+0x07,
+0x07,
+0x07,
+0x07,
+0x2d,
+0x03,
+0x17,
+0x07,
+0x8b,
+0x4f,
+0x21,
+0x04,
+0x22,
+0x0b,
+0x07,
+0x4c,
+0x08,
+0x07,
+0x08,
+0x07,
+0x00,
+0x00,
+0x00,
+0x06,
+0x07,
+0x08,
+0x07,
+0x08,
+0x16,
+0x04,
+0x08,
+0x05,
+0x05,
+0x07,
+0x07,
+0x07,
+0x29,
+0x06,
+0x07,
+0x04,
+0x0d,
+0x07,
+0x07,
+0x4e,
+0x07,
+0x0d,
+0x05,
+0x00,
+0x08,
+0x07,
+0x07,
+0x07,
+0x00,
+0x06,
+0x06,
+0x07,
+0x07,
+0x05,
+0x06,
+0x27,
+0x07,
+0x07,
+0x01,
+0x07,
+0x03,
+0x00,
+0x07,
+0x8a,
+0x04,
+0x05,
+0x00,
+0x07,
+0x04,
+0x05,
+0x07,
+0x01,
+0x05,
+0x04,
+0x05,
+0x03,
+0x07,
+0x07,
+0x07,
+0x07,
+0x0b,
+0x07,
+0x06,
+0x00,
+0x03,
+0x03,
+0x23,
+0x05,
+0x04,
+0x0a,
+0x00,
+0x07,
+0x07,
+0x07,
+0x8b,
+0x21,
+0x0c,
+0x07,
+0x8b,
+0x07,
+0x04,
+0x8a,
+0x07,
+0x07,
+0x03,
+0x03,
+0x21,
+0x88,
+0x4d,
+0x07,
+0x05,
+0x07,
+0x05,
+0x07,
+0x07,
+0x07,
+0x07,
+0x47,
+0x16,
+0x07,
+0x08,
+0x07,
+0x8b,
+0x07,
+0x03,
+0x00,
+0x47,
+0x06,
+0x47,
+0x07,
+0x07,
+0x01,
+0x8a,
+0x00,
+0x05,
+0x27,
+0x03,
+0x9a,
+0x02,
+0x16,
+0x08,
+0x04,
+0x07,
+0x07,
+0x07,
+0x0b,
+0x8a,
+0x00,
+0x00,
+0x01,
+0x08,
+0x4c,
+0x0c,
+0x0a,
+0xcc,
+0x4c,
+0x07,
+0x07,
+0x07,
+0x04,
+0x07,
+0x8a,
+0x16,
+0x07,
+0x0b,
+0x16,
+0x07,
+0x03,
+0x07,
+0x03,
+0x4c,
+0x07,
+0x06,
+0x07,
+0x4c,
+0x07,
+0x07,
+0x08,
+0x4d,
+0x00,
+0x0a,
+0x0c,
+0x03,
+0x07,
+0x07,
+0x05,
+0x03,
+0x8b,
+0x00,
+0x07,
+0x07,
+0x07,
+0x07,
+0x07,
+0x05,
+0xcd,
+0x00,
+0x07,
+0x0a,
+0x03,
+0x21,
+0x8b,
+0x04,
+0x00,
+0x00,
+0x25,
+0x04,
+0x07,
+0x8a,
+0x04,
+0x21,
+0x07,
+0x00,
+0x00,
+0x00,
+0x07,
+0x08,
+0x07,
+0x07,
+0x07,
+0x0c,
+0x04,
+0x07,
+0x07,
+0x08,
+0x21,
+0x07,
+0x08,
+0x07,
+0x05,
+0x0a,
+0x07,
+0x03,
+0x04,
+0x04,
+0x04,
+0x08,
+0x05,
+0x07,
+0x03,
+0x05,
+0x87,
+0x05,
+0x5c,
+0x07,
+0x0b,
+0x03,
+0x09,
+0x07,
+0x8b,
+0x07,
+0x07,
+0x07,
+0x00,
+0x07,
+0x04,
+0x4c,
+0x07,
+0x4c,
+0x07,
+0x4c,
+0x8b,
+0x00,
+0x00,
+0x2d,
+0x4c,
+0x21,
+0x07,
+0x0a,
+0x07,
+0x16,
+0x04,
+0x04,
+0x07,
+0x08,
+0x03,
+0x07,
+0x4c,
+0x03,
+0x03,
+0x27,
+0x07,
+0x07,
+0x04,
+0x08,
+0x04,
+0x04,
+0x21,
+0x20,
+0x06,
+0x04,
+0x07,
+0x07,
+0x07,
+0x03,
+0x03,
+0x03,
+0x07,
+0x07,
+0x07,
+0x03,
+0x08,
+0x07,
+0x00,
+0x03,
+0x08,
+0x04,
+0x07,
+0x07,
+0x03,
+0x07,
+0x05,
+0xa5,
+0x03,
+0x05,
+0x07,
+0x07,
+0x8b,
+0x8b,
+0xcd,
+0x07,
+0x03,
+0x5e,
+0x48,
+0x07,
+0x0b,
+0x00,
+0x4e,
+0x0c,
+0x07,
+0x0a,
+0x07,
+0x07,
+0x07,
+0x07,
+0x03,
+0x07,
+0x0a,
+0x07,
+0x04,
+0x07,
+0x29,
+0x2d,
+0x05,
+0x07,
+0x23,
+0x04,
+0x8b,
+0x17,
+0x05,
+0x07,
+0x07,
+0x07,
+0x06,
+0x0d,
+0x07,
+0x07,
+0x04,
+0x07,
+0x00,
+0x07,
+0x00,
+0x27,
+0x07,
+0x16,
+0x05,
+0x47,
+0x08,
+0x07,
+0x04,
+0x5e,
+0x07,
+0x07,
+0x23,
+0x00,
+0x08,
+0x5e,
+0x0f,
+0x0c,
+0x07,
+0x04,
+0x05,
+0x07,
+0x07,
+0x16,
+0x07,
+0x07,
+0x07,
+0x2d,
+0x07,
+0x07,
+0x8a,
+0x07,
+0x08,
+0x9b,
+0x00,
+0x07,
+0x07,
+0x8b,
+0x27,
+0x07,
+0x21,
+0x0b,
+0x16,
+0x17,
+0x05,
+0x07,
+0x0b,
+0x05,
+0x01,
+0x04,
+0x07,
+0x07,
+0x08,
+0x04,
+0x07,
+0x07,
+0x07,
+0x8b,
+0x0c,
+0x0a,
+0x8b,
+0x03,
+0x00,
+0x27,
+0x0a,
+0xc9,
+0x07,
+0x07,
+0x07,
+0x08,
+0x08,
+0x07,
+0x01,
+0x05,
+0x07,
+0x05,
+0x07,
+0x16,
+0x04,
+0x07,
+0x21,
+0x06,
+0x03,
+0x07,
+0x07,
+0x06,
+0x07,
+0x0f,
+0x0e,
+0x07,
+0x08,
+0x08,
+0x00,
+0x8a,
+0x00,
+0x05,
+0x07,
+0x03,
+0x07,
+0x08,
+0x8b,
+0x05,
+0x03,
+0x8a,
+0x0b,
+0x14,
+0x08,
+0x0d,
+0x08,
+0x8a,
+0x07,
+0x0c,
+0x08,
+0x07,
+0x07,
+0x05,
+0x08,
+0x07,
+0x07,
+0x07,
+0x04,
+0x4e,
+0x4c,
+0x84,
+0x00,
+0x07,
+0x07,
+0x00,
+0x04,
+0x21,
+0x07,
+0x8a,
+0x03,
+0x07,
+0x00,
+0x03,
+0x07,
+0x00,
+0x05,
+0x07,
+0x8a,
+0x07,
+0x04,
+0x07,
+0x8a,
+0x8b,
+0x4c,
+0x05,
+0x00,
+0x04,
+0x07,
+0x05,
+0x0b,
+0x07,
+0x07,
+0x0c,
+0x01,
+0x07,
+0x17,
+0x04,
+0x03,
+0x22,
+0x07,
+0x21,
+0x08,
+0x05,
+0x07,
+0x4e,
+0x06,
+0x07,
+0x08,
+0x03,
+0x08,
+0x29,
+0x07,
+0x4c,
+0x01,
+0x03,
+0x0b,
+0x0b,
+0x0e,
+0x18,
+0x07,
+0x0b,
+0x07,
+0x08,
+0x03,
+0x07,
+0x85,
+0x07,
+0x17,
+0x21,
+0x83,
+0x07,
+0x07,
+0x4c,
+0x07,
+0x00,
+0x00,
+0x07,
+0x03,
+0x07,
+0x0a,
+0x03,
+0x00,
+0x08,
+0x04,
+0x00,
+0x07,
+0x16,
+0x02,
+0x08,
+0x0c,
+0x08,
+0x07,
+0x02,
+0x07,
+0x01,
+0x00,
+0x21,
+0x07,
+0x00,
+0x07,
+0x00,
+0x06,
+0x07,
+0x03,
+0x16,
+0x0b,
+0x06,
+0x00,
+0x00,
+0x00,
+0x05,
+0x07,
+0x07,
+0x04,
+0x07,
+0x03,
+0x01,
+0x04,
+0x07,
+0x07,
+0x23,
+0x00,
+0x04,
+0x06,
+0x06,
+0x07,
+0x05,
+0x0d,
+0x07,
+0x8b,
+0x08,
+0x05,
+0x0f,
+0x07,
+0x08,
+0x05,
+0x07,
+0x04,
+0x00,
+0x05,
+0x4c,
+0x07,
+0x07,
+0x01,
+0x47,
+0x0a,
+0x4c,
+0x07,
+0x03,
+0x2d,
+0x00,
+0x04,
+0x07,
+0x03,
+0x05,
+0x00,
+0x07,
+0x0f,
+0x04,
+0x07,
+0x00,
+0x10,
+0x07,
+0x07,
+0x08,
+0x08,
+0x05,
+0x07,
+0x00,
+0x0c,
+0x07,
+0x03,
+0x03,
+0x08,
+0x07,
+0x07,
+0x00,
+0x07,
+0x07,
+0x05,
+0x4c,
+0x07,
+0x08,
+0x07,
+0x0f,
+0x16,
+0x8a,
+0x07,
+0x00,
+0x07,
+0x07,
+0x07,
+0x07,
+0x07,
+0x17,
+0x0b,
+0x07,
+0x08,
+0x0c,
+0x07,
+0x07,
+0x0a,
+0x07,
+0x08,
+0x07,
+0x07,
+0x04,
+0x04,
+0x4e,
+0x05,
+0x07,
+0x03,
+0x05,
+0x07,
+0x03,
+0x17,
+0x07,
+0x07,
+0x07,
+0x07,
+0x83,
+0x00,
+0x0a,
+0x05,
+0x08,
+0x8a,
+0x07,
+0x0e,
+0x08,
+0x0d,
+0x04,
+0x44,
+0x07,
+0x07,
+0x4c,
+0x0a,
+0x05,
+0x05,
+0x4c,
+0x07,
+0x05,
+0x07,
+0x45,
+0x05,
+0x07,
+0x07,
+0x8a,
+0x04,
+0x0a,
+0x08,
+0x07,
+0x0a,
+0x8b,
+0x03,
+0x03,
+0x07,
+0x08,
+0x07,
+0x05,
+0x4c,
+0x45,
+0x07,
+0x05,
+0x04,
+0x04,
+0x07,
+0x08,
+0x00,
+0x07,
+0x01,
+0x00,
+0x0c,
+0x05,
+0x0a,
+0x0a,
+0x07,
+0x07,
+0x8a,
+0x07,
+0x06,
+0x00,
+0x8b,
+0x07,
+0x16,
+0x87,
+0x0c,
+0x07,
+0x06,
+0x04,
+0x04,
+0x08,
+0x07,
+0x0a,
+0x07,
+0x04,
+0x2d,
+0x05,
+0x08,
+0x12,
+0xa9,
+0x0b,
+0x07,
+0x00,
+0x07,
+0x21,
+0x43,
+0x00,
+0x8b,
+0x0f,
+0x06,
+0x03,
+0x06,
+0x05,
+0x04,
+0x08,
+0x06,
+0x07,
+0x07,
+0x00,
+0x03,
+0x8a,
+0x0c,
+0x05,
+0x07,
+0x07,
+0x05,
+0x4c,
+0x07,
+0x07,
+0x07,
+0x05,
+0x07,
+0x03,
+0x47,
+0x8b,
+0x27,
+0x04,
+0x07,
+0x07,
+0x03,
+0x07,
+0x21,
+0x05,
+0x07,
+0x03,
+0x07,
+0x07,
+0x08,
+0x05,
+0x4e,
+0x07,
+0x07,
+0x07,
+0x0f,
+0x07,
+0x16,
+0x08,
+0x07,
+0x01,
+0x07,
+0x07,
+0x05,
+0x07,
+0x00,
+0x07,
+0x08,
+0x01,
+0x0b,
+0x07,
+0x01,
+0x0b,
+0x8b,
+0x07,
+0x00,
+0x08,
+0x8a,
+0x08,
+0x0f,
+0x07,
+0x87,
+0x04,
+0x08,
+0xcd,
+0x03,
+0x04,
+0x07,
+0x07,
+0x08,
+0x00,
+0x08,
+0x16,
+0x05,
+0x0b,
+0x00,
+0x07,
+0x05,
+0x01,
+0x00,
+0x00,
+0x08,
+0x03,
+0x12,
+0x07,
+0x04,
+0x00,
+0x07,
+0x03,
+0x07,
+0x03,
+0x07,
+0x08,
+0x04,
+0x00,
+0x05,
+0x00,
+0x07,
+0x07,
+0x0b,
+0x00,
+0x0b,
+0x00,
+0x07,
+0x04,
+0x07,
+0x07,
+0x22,
+0x07,
+0x21,
+0x03,
+0x04,
+0x07,
+0x07,
+0x00,
+0x07,
+0x05,
+0x08,
+0xcd,
+0x03,
+0x07,
+0x07,
+0x08,
+0x06,
+0x00,
+0x0b,
+0x03,
+0x05,
+0x07,
+0x07,
+0x05,
+0x07,
+0x2d,
+0x05,
+0x04,
+0x07,
+0x07,
+0x04,
+0x08,
+0x08,
+0x07,
+0x07,
+0x08,
+0x08,
+0x01,
+0x08,
+0x05,
+0x05,
+0x07,
+0x05,
+0x8b,
+0x05,
+0x03,
+0x0a,
+0x08,
+0x07,
+0x5e,
+0x03,
+0x0d,
+0x21,
+0x8b,
+0x07,
+0x21,
+0x07,
+0x07,
+0x0c,
+0x07,
+0x00,
+0x08,
+0x25,
+0x05,
+0x07,
+0x5e,
+0x07,
+0xcd,
+0x07,
+0x04,
+0x05,
+0x17,
+0x08,
+0x07,
+0x07,
+0x08,
+0x07,
+0x08,
+0x00,
+0x0b,
+0x01,
+0x07,
+0x01,
+0x4d,
+0x01,
+0x07,
+0x05,
+0x05,
+0x08,
+0x07,
+0x07,
+0x16,
+0x0b,
+0x00,
+0x07,
+0x05,
+0x06,
+0x07,
+0x0e,
+0x29,
+0x9b,
+0x08,
+0x07,
+0x00,
+0x04,
+0x07,
+0x05,
+0x25,
+0x07,
+0x03,
+0x03,
+0x00,
+0x16,
+0x0a,
+0x01,
+0x08,
+0x06,
+0x07,
+0x05,
+0x07,
+0x03,
+0x05,
+0x07,
+0x0a,
+0x07,
+0x03,
+0x87,
+0x07,
+0x07,
+0x07,
+0x18,
+0x04,
+0x08,
+0x08,
+0x07,
+0x07,
+0x07,
+0x05,
+0x07,
+0x04,
+0x05,
+0x07,
+0x04,
+0x00,
+0x05,
+0x07,
+0x08,
+0x03,
+0x00,
+0x4c,
+0x07,
+0x08,
+0x1a,
+0x06,
+0x05,
+0x00,
+0x2d,
+0x01,
+0x07,
+0x0c,
+0x07,
+0x03,
+0x05,
+0x07,
+0x05,
+0x07,
+0x08,
+0x04,
+0x07,
+0x05,
+0x07,
+0x05,
+0x00,
+0x16,
+0x03,
+0x0a,
+0x07,
+0x8b,
+0x00,
+0x07,
+0x05,
+0x07,
+0x05,
+0x04,
+0x08,
+0x00,
+0x00,
+0x07,
+0x07,
+0x04,
+0x07,
+0x47,
+0x08,
+0x8b,
+0x0b,
+0x04,
+0x07,
+0x07,
+0x21,
+0x8a,
+0x07,
+0x03,
+0x08,
+0x8a,
+0x03,
+0x08,
+0x08,
+0x05,
+0x07,
+0x08,
+0x04,
+0x03,
+0x00,
+0x07,
+0x07,
+0x07,
+0x00,
+0x01,
+0x4d,
+0x8b,
+0x01,
+0x8b,
+0x00,
+0x01,
+0x8a,
+0x04,
+0x04,
+0x07,
+0x07,
+0x08,
+0x00,
+0x07,
+0x07,
+0x4c,
+0x07,
+0x07,
+0x07,
+0x03
+};
+
+
+
+L1_DATA uint32_t threshs_l1 [40] = {0};
+L1_DATA int32_t kappa_l1 [40] = {0};
+L1_DATA int32_t lambda_l1 [40] = {0};
+L1_DATA uint8_t inp_l1 [8192] = {0};
+L1_DATA uint8_t outp_l1 [2048] = {0};
+L1_DATA uint8_t im2col_l1 [IM2COL_DIM] = {0};
+
+#endif
diff --git a/rt_nn_tests/xptnn_maxpool/pulp_nn_kernels.h b/rt_nn_tests/xptnn_maxpool/pulp_nn_kernels.h
new file mode 100644
index 0000000..f4ec977
--- /dev/null
+++ b/rt_nn_tests/xptnn_maxpool/pulp_nn_kernels.h
@@ -0,0 +1,26528 @@
+/*
+ * pulp_nn_kernels.h
+ * Nazareno Bruschi <nazareno.bruschi@unibo.it>
+ *
+ * Copyright (C) 2019-2020 University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef __XPULPNN_KERNELS__
+#define __XPULPNN_KERNELS__
+
+void xpulp_nn_conv_u8_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_u2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u8_i2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_u2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i8_i2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_u2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u4_i2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_u2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i4_i2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_u2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_u2_i2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_u2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv_i2_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+#ifndef PROFILE
+                        uint8_t flag_batch_norm);
+#else
+  uint8_t flag_batch_norm,
+    uint32_t *im2col_cycles,
+    uint32_t *hotloop_prep_cycles,
+    uint32_t *hotloop_cycles,
+    uint32_t *requant_cycles,
+    uint32_t *hotloop_leftover_cycles,
+    uint32_t *matmul4x2_leftover_cycles);
+#endif
+                        
+
+void xpulp_nn_conv_i2_i2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_u2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u8_i2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_u2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i8_i2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_u2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u4_i2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_u2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i4_i2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_u2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_u2_i2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_u2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_conv1d_i2_i2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t dilation_x,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_i8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_i8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_i8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_i4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_i4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_i4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_i2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_i2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u8_i2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i8_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_i8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_i8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_i8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_i4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_i4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_i4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_i2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_i2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u4_i2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i4_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_i8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_i8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_i8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_i4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_i4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_i4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_i2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_i2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_u2_i2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_pointwise_i2_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u8_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i8_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u8_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i8_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u8_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i8_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u4_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i4_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u4_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i4_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u4_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i4_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u2_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i2_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u2_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i2_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_u2_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u8_i2_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_u2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i8_i2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u8_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i8_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u8_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i8_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u8_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i8_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u4_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i4_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u4_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i4_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u4_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i4_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u2_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i2_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u2_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i2_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_u2_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u4_i2_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_u2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i4_i2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u8_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i8_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i8_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u8_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i8_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i8_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u8_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i8_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i8_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u4_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i4_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i4_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u4_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i4_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i4_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u4_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i4_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i4_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u2_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i2_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i2_i8_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u2_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i2_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i2_i4_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_u2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+#ifndef PROFILE
+                        uint8_t flag_batch_norm);
+#else
+  uint8_t flag_batch_norm,
+    uint32_t *hotloop_prep_cycles,
+    uint32_t *hotloop_cycles,
+    uint32_t *requant_cycles,
+    uint32_t *hotloop_leftover_cycles);
+#endif
+
+uint8_t *xpulp_nn_matmul_u2_u2_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_u2_i2_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_u2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_matmul_i2_i2_i2_4x4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pOut2,
+                        int8_t *pOut3,
+                        int8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_i8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_i8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_i8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_i4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_i4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_i4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_i2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_i2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u8_i2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i8_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_i8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_i8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_i8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_i4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_i4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_i4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_i2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_i2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u4_i2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i4_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_i8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_i8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_i8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_i4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_i4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_i4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_i2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_i2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_u2_i2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_depthwise_i2_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int8_t *pWtBuffer,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_i32_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_i8_i32_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_u8_i32_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_i8_i32_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_u8_i32_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_i8_i32_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_u4_i32_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_i4_i32_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_u4_i32_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_i4_i32_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_u4_i32_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_i4_i32_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_u2_i32_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_i2_i32_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_u2_i32_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_i2_i32_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_u2_i32_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_i2_i32_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons);
+
+void xpulp_nn_linear_u8_u8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_i8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_u8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_i8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_u8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_i8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_u4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_i4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_u4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_i4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_u4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_i4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_u2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_i2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_u2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_i2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_u2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u8_i2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i8_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_u8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_i8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_u8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_i8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_u8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_i8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_u4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_i4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_u4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_i4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_u4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_i4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_u2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_i2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_u2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_i2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_u2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u4_i2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i4_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_u8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_i8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_u8_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_i8_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_u8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_i8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_u8_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_i8_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_u8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_i8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_u8_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_i8_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_u4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_i4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_u4_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_i4_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_u4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_i4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_u4_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_i4_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_u4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_i4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_u4_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_i4_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_u2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_i2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_u2_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_i2_i8(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_u2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_i2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_u2_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_i2_i4(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_u2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_u2_i2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_u2_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_linear_i2_i2_i2(
+                        int8_t *pIn,
+                        int8_t *pBias,
+                        int8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_vec,
+                        uint16_t num_o_neurons,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_maxpool_u8(
+                        uint8_t * pIn,
+                        uint8_t * pOut,
+                        uint16_t  dim_im_in_x,
+                        uint16_t  dim_im_in_y,
+                        uint16_t  ch_im_in,
+                        uint16_t  dim_im_out_x,
+                        uint16_t  dim_im_out_y,
+                        uint16_t  dim_kernel_x,
+                        uint16_t  dim_kernel_y,
+                        uint16_t  padding_t,
+                        uint16_t  padding_b,
+                        uint16_t  padding_l,
+                        uint16_t  padding_r,
+                        uint16_t  stride_x,
+                        uint16_t  stride_y);
+
+void xpulp_nn_maxpool_i8(
+                        int8_t * pIn,
+                        int8_t * pOut,
+                        uint16_t  dim_im_in_x,
+                        uint16_t  dim_im_in_y,
+                        uint16_t  ch_im_in,
+                        uint16_t  dim_im_out_x,
+                        uint16_t  dim_im_out_y,
+                        uint16_t  dim_kernel_x,
+                        uint16_t  dim_kernel_y,
+                        uint16_t  padding_t,
+                        uint16_t  padding_b,
+                        uint16_t  padding_l,
+                        uint16_t  padding_r,
+                        uint16_t  stride_x,
+                        uint16_t  stride_y);
+
+void xpulp_nn_maxpool_u4(
+                        uint8_t * pIn,
+                        uint8_t * pOut,
+                        uint16_t  dim_im_in_x,
+                        uint16_t  dim_im_in_y,
+                        uint16_t  ch_im_in,
+                        uint16_t  dim_im_out_x,
+                        uint16_t  dim_im_out_y,
+                        uint16_t  dim_kernel_x,
+                        uint16_t  dim_kernel_y,
+                        uint16_t  padding_t,
+                        uint16_t  padding_b,
+                        uint16_t  padding_l,
+                        uint16_t  padding_r,
+                        uint16_t  stride_x,
+                        uint16_t  stride_y);
+
+void xpulp_nn_maxpool_i4(
+                        int8_t * pIn,
+                        int8_t * pOut,
+                        uint16_t  dim_im_in_x,
+                        uint16_t  dim_im_in_y,
+                        uint16_t  ch_im_in,
+                        uint16_t  dim_im_out_x,
+                        uint16_t  dim_im_out_y,
+                        uint16_t  dim_kernel_x,
+                        uint16_t  dim_kernel_y,
+                        uint16_t  padding_t,
+                        uint16_t  padding_b,
+                        uint16_t  padding_l,
+                        uint16_t  padding_r,
+                        uint16_t  stride_x,
+                        uint16_t  stride_y);
+
+void xpulp_nn_maxpool_u2(
+                        uint8_t * pIn,
+                        uint8_t * pOut,
+                        uint16_t  dim_im_in_x,
+                        uint16_t  dim_im_in_y,
+                        uint16_t  ch_im_in,
+                        uint16_t  dim_im_out_x,
+                        uint16_t  dim_im_out_y,
+                        uint16_t  dim_kernel_x,
+                        uint16_t  dim_kernel_y,
+                        uint16_t  padding_t,
+                        uint16_t  padding_b,
+                        uint16_t  padding_l,
+                        uint16_t  padding_r,
+                        uint16_t  stride_x,
+                        uint16_t  stride_y);
+
+void xpulp_nn_maxpool_i2(
+                        int8_t * pIn,
+                        int8_t * pOut,
+                        uint16_t  dim_im_in_x,
+                        uint16_t  dim_im_in_y,
+                        uint16_t  ch_im_in,
+                        uint16_t  dim_im_out_x,
+                        uint16_t  dim_im_out_y,
+                        uint16_t  dim_kernel_x,
+                        uint16_t  dim_kernel_y,
+                        uint16_t  padding_t,
+                        uint16_t  padding_b,
+                        uint16_t  padding_l,
+                        uint16_t  padding_r,
+                        uint16_t  stride_x,
+                        uint16_t  stride_y);
+
+void xpulp_nn_avgpool_u8_u8(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u8_i8(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i8_u8(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i8_i8(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u8_u4(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u8_i4(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i8_u4(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i8_i4(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u8_u2(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u8_i2(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i8_u2(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i8_i2(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u4_u8(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u4_i8(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i4_u8(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i4_i8(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u4_u4(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u4_i4(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i4_u4(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i4_i4(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u4_u2(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u4_i2(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i4_u2(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i4_i2(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u2_u8(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u2_i8(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i2_u8(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i2_i8(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u2_u4(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u2_i4(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i2_u4(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i2_i4(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u2_u2(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u2_i2(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i2_u2(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i2_i2(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_add_u8_u8(
+                        uint8_t * pIn1,
+                        uint8_t * pIn2,
+                        uint8_t * pOut,
+                        uint16_t out_mult1,
+                        uint16_t out_mult2,
+                        uint16_t out_shift,
+                        uint16_t dim_im_in_x,
+                        uint16_t dim_im_in_y,
+                        uint16_t ch_im_in);
+
+void xpulp_nn_add_u8_u8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u4(
+                        uint8_t * pIn1,
+                        uint8_t * pIn2,
+                        uint8_t * pOut,
+                        uint16_t out_mult1,
+                        uint16_t out_mult2,
+                        uint16_t out_shift,
+                        uint16_t dim_im_in_x,
+                        uint16_t dim_im_in_y,
+                        uint16_t ch_im_in);
+
+void xpulp_nn_add_u8_u4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u2(
+                        uint8_t * pIn1,
+                        uint8_t * pIn2,
+                        uint8_t * pOut,
+                        uint16_t out_mult1,
+                        uint16_t out_mult2,
+                        uint16_t out_shift,
+                        uint16_t dim_im_in_x,
+                        uint16_t dim_im_in_y,
+                        uint16_t ch_im_in);
+
+void xpulp_nn_add_u8_u2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u4(
+                        uint8_t * pIn1,
+                        uint8_t * pIn2,
+                        uint8_t * pOut,
+                        uint16_t out_mult1,
+                        uint16_t out_mult2,
+                        uint16_t out_shift,
+                        uint16_t dim_im_in_x,
+                        uint16_t dim_im_in_y,
+                        uint16_t ch_im_in);
+
+void xpulp_nn_add_u4_u4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u2(
+                        uint8_t * pIn1,
+                        uint8_t * pIn2,
+                        uint8_t * pOut,
+                        uint16_t out_mult1,
+                        uint16_t out_mult2,
+                        uint16_t out_shift,
+                        uint16_t dim_im_in_x,
+                        uint16_t dim_im_in_y,
+                        uint16_t ch_im_in);
+
+void xpulp_nn_add_u4_u2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u2(
+                        uint8_t * pIn1,
+                        uint8_t * pIn2,
+                        uint8_t * pOut,
+                        uint16_t out_mult1,
+                        uint16_t out_mult2,
+                        uint16_t out_shift,
+                        uint16_t dim_im_in_x,
+                        uint16_t dim_im_in_y,
+                        uint16_t ch_im_in);
+
+void xpulp_nn_add_u2_u2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+
+
+#endif
diff --git a/rt_nn_tests/xptnn_maxpool/pulp_nn_mix_kernels.h b/rt_nn_tests/xptnn_maxpool/pulp_nn_mix_kernels.h
new file mode 100644
index 0000000..8b2a3c6
--- /dev/null
+++ b/rt_nn_tests/xptnn_maxpool/pulp_nn_mix_kernels.h
@@ -0,0 +1,7093 @@
+/*
+ * pulp_nn_kernels.h
+ * Nazareno Bruschi <nazareno.bruschi@unibo.it>
+ *
+ * Copyright (C) 2019-2020 University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef __XPULPNN-MIXED_KERNELS__
+#define __XPULPNN-MIXED_KERNELS__
+
+void xpulp_nn_mix_conv_u8_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u8_u2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u4_u2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u8_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u8_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u8_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u8_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u8_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u8_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u4_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u4_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u4_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u4_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u4_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u4_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u2_i8(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u2_i8_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u2_i4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u2_i4_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u2_i2(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_mix_conv_u2_u2_i2_4x4(
+                        uint8_t *pIn,
+                        uint8_t *pIm2ColBuffer,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t dim_in_x,
+                        uint16_t dim_in_y,
+                        uint16_t ch_in,
+                        uint16_t dim_out_x,
+                        uint16_t dim_out_y,
+                        uint16_t ch_out,
+                        uint16_t dim_kernel_x,
+                        uint16_t dim_kernel_y,
+                        uint16_t padding_y_top,
+                        uint16_t padding_y_bottom,
+                        uint16_t padding_x_left,
+                        uint16_t padding_x_right,
+                        uint16_t stride_x,
+                        uint16_t stride_y,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u8_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u8_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u8_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u4_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u4_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u4_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u2_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u2_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u8_u2_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u8_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u8_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u8_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u4_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u4_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u4_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u2_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u2_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u4_u2_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u8_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u8_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u8_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u8_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u8_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u8_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u4_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u4_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u4_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u4_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u4_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u4_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u2_i8(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u2_i8_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u2_i4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u2_i4_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u2_i2(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+uint8_t *xpulp_nn_mix_matmul_u2_u2_i2_4x4(
+                        uint8_t *pIn,
+                        int8_t *pBias,
+                        uint8_t *pOut,
+                        uint8_t *pOut2,
+                        uint8_t *pOut3,
+                        uint8_t *pOut4,
+                        int8_t *pWeight,
+                        int32_t *pKappa,
+                        int32_t *pLambda,
+                        uint16_t out_mul,
+                        uint16_t out_shift,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+                        uint8_t flag_relu,
+                        uint8_t flag_batchnorm);
+
+void xpulp_nn_avgpool_u8_u8(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u8_i8(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i8_u8(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i8_i8(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u8_u4(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u8_i4(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i8_u4(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i8_i4(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u8_u2(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u8_i2(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i8_u2(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i8_i2(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u4_u8(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u4_i8(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i4_u8(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i4_i8(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u4_u4(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u4_i4(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i4_u4(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i4_i4(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u4_u2(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u4_i2(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i4_u2(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i4_i2(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u2_u8(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u2_i8(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i2_u8(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i2_i8(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u2_u4(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u2_i4(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i2_u4(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i2_i4(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u2_u2(
+                       uint8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_u2_i2(
+                       uint8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i2_u2(
+                       int8_t * pIn,
+                       uint8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_avgpool_i2_i2(
+                       int8_t * pIn,
+                       int8_t * pOut,
+                       int32_t lambda,
+                       uint16_t out_shift,
+                       int32_t out_add,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       uint16_t dim_im_out_x,
+                       uint16_t dim_im_out_y,
+                       uint16_t dim_kernel_x,
+                       uint16_t dim_kernel_y,
+                       uint16_t padding_t,
+                       uint16_t padding_b,
+                       uint16_t padding_l,
+                       uint16_t padding_r,
+                       uint16_t stride_x,
+                       uint16_t stride_y,
+                       int flag_requant
+);
+
+void xpulp_nn_add_u8_u8(
+                        uint8_t * pIn1,
+                        uint8_t * pIn2,
+                        uint8_t * pOut,
+                        uint16_t out_mult1,
+                        uint16_t out_mult2,
+                        uint16_t out_shift,
+                        uint16_t dim_im_in_x,
+                        uint16_t dim_im_in_y,
+                        uint16_t ch_im_in);
+
+void xpulp_nn_add_u8_u8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u4(
+                        uint8_t * pIn1,
+                        uint8_t * pIn2,
+                        uint8_t * pOut,
+                        uint16_t out_mult1,
+                        uint16_t out_mult2,
+                        uint16_t out_shift,
+                        uint16_t dim_im_in_x,
+                        uint16_t dim_im_in_y,
+                        uint16_t ch_im_in);
+
+void xpulp_nn_add_u8_u4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u2(
+                        uint8_t * pIn1,
+                        uint8_t * pIn2,
+                        uint8_t * pOut,
+                        uint16_t out_mult1,
+                        uint16_t out_mult2,
+                        uint16_t out_shift,
+                        uint16_t dim_im_in_x,
+                        uint16_t dim_im_in_y,
+                        uint16_t ch_im_in);
+
+void xpulp_nn_add_u8_u2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_u2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u8_i2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_u2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i8_i2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u4(
+                        uint8_t * pIn1,
+                        uint8_t * pIn2,
+                        uint8_t * pOut,
+                        uint16_t out_mult1,
+                        uint16_t out_mult2,
+                        uint16_t out_shift,
+                        uint16_t dim_im_in_x,
+                        uint16_t dim_im_in_y,
+                        uint16_t ch_im_in);
+
+void xpulp_nn_add_u4_u4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u2(
+                        uint8_t * pIn1,
+                        uint8_t * pIn2,
+                        uint8_t * pOut,
+                        uint16_t out_mult1,
+                        uint16_t out_mult2,
+                        uint16_t out_shift,
+                        uint16_t dim_im_in_x,
+                        uint16_t dim_im_in_y,
+                        uint16_t ch_im_in);
+
+void xpulp_nn_add_u4_u2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_u2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u4_i2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_u2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i4_i2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i8_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i8_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i8_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i8_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i8_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i8_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i4_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i4_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i4_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i4_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i4_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i4_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u2(
+                        uint8_t * pIn1,
+                        uint8_t * pIn2,
+                        uint8_t * pOut,
+                        uint16_t out_mult1,
+                        uint16_t out_mult2,
+                        uint16_t out_shift,
+                        uint16_t dim_im_in_x,
+                        uint16_t dim_im_in_y,
+                        uint16_t ch_im_in);
+
+void xpulp_nn_add_u2_u2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_u2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_u2_i2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_u2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i2_u8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i2_u4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i2_u2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i2_i8(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i2_i4(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+void xpulp_nn_add_i2_i2_i2(
+                       uint8_t * pIn1,
+                       uint8_t * pIn2,
+                       uint8_t * pOut,
+                       int32_t in_mult1,
+                       int32_t in_add1,
+                       uint16_t in_shift1,
+                       int32_t in_mult2,
+                       int32_t in_add2,
+                       uint16_t in_shift2,
+                       int32_t out_mult,
+                       int32_t out_add,
+                       uint16_t out_shift,
+                       uint16_t dim_im_in_x,
+                       uint16_t dim_im_in_y,
+                       uint16_t ch_im_in,
+                       int      out_requant_flag);
+
+
+
+#endif
\ No newline at end of file
diff --git a/rt_nn_tests/xptnn_maxpool/pulp_nn_utils.h b/rt_nn_tests/xptnn_maxpool/pulp_nn_utils.h
new file mode 100644
index 0000000..44d2251
--- /dev/null
+++ b/rt_nn_tests/xptnn_maxpool/pulp_nn_utils.h
@@ -0,0 +1,2079 @@
+/*
+ * pulp_nn_utils.h
+ * Nazareno   Bruschi  <nazareno.bruschi@unibo.it>
+ * Alessandro Nadalini <alessandro.nadalini3@unibo.it>
+ * Georg Rutishauser   <georgr@iis.ee.ethz.ch>
+ *
+ * Copyright (C) 2019-2020 ETH Zurich & University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef __PULPNN_UTILS__
+#define __PULPNN_UTILS__
+
+#include <stdint.h>
+
+typedef   signed short v2s __attribute__((vector_size (4)));
+
+
+
+#define min(a,b)                                             ((a)<(b)?(a):(b))
+#define log2(x)                                              __builtin_pulp_fl1(x)
+#define PACK_INT2_SIZE(x)                                    ((x) >> 2)
+
+#define MemoryFence()                                        asm volatile("":::"memory")
+
+#define CHANS_DECOMPR(x)                                     (5*x >> 2) // equivalent to division by 0.8
+
+/* Functions for Compressed MAC */
+#define CompressedMAC(sum, ptr, config) asm volatile(                \
+    "pv.smlsdotsp.t %[shum], %[phtr], %[chonfig];"                    \
+    : [shum] "+r" (sum), [phtr] "+r" (ptr): [chonfig] "I" (config))
+
+#define CompressedMACUnsigned(sum, ptr, config) asm volatile(                   \
+    "pv.smlsdotsup.t %[shum], %[phtr], %[chonfig];" \
+    : [shum] "+r" (sum), [phtr] "+r" (ptr): [chonfig] "I" (config))
+
+#define InitNNRF(ptr, config) asm volatile(        \
+    "pv.smlsdotsp.t x0, %[phtr], %[chonfig];"       \
+    : [phtr] "+r" (ptr) : [chonfig] "I" (config))
+
+#define ThresholdCompress(res, val, thrs) asm volatile(                                                \
+    "pv.thrc %[rhes], %[vhal], %[thhrs];" : [rhes] "+r" (res) : [vhal] "r" (val), [thhrs] "r" (thrs))
+
+#define GetConfig(a_update, b_update, a_reg, b_reg) a_update << 4 | b_update << 3 | a_reg << 1 | b_reg
+
+/* Functions for threshold&compress */
+#define check_store(res, pOut)            \
+  if ((res & 0xe0000000) == 0x00000000) { \
+    *pOut = res & 0xff;                   \
+    pOut++;                               \
+    incr_val=ch_out_r; }
+
+#define check_store_4x1(res, pOut)        \
+  if ((res & 0xe0000000) == 0x00000000) { \
+    *pOut = res & 0xff;                   \
+    pOut++; }
+
+#define reset_currThr()                           \
+  if ((uint32_t *) currThr == (uint32_t *) (pThr + ch_out)) { \
+    currThr = (v2s *) pThr;                                    \
+  }
+
+#define MacLoads20(a_update, b_update, a_reg, b_reg, ptr, sum)  __builtin_pulp_mlsdotsp20_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+
+#define MacLoad20(a_update, b_update, a_reg, b_reg, ptr, sum)  __builtin_pulp_mlsdotsup20_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+
+/* Functions for compressed min/max */
+#define CompressedMax(res, in1, in2) asm volatile( \
+                                                   "pv.max.t %[res_], %[in1_], %[in2_];" : [res_] "=r" (res) : [in1_] "r" (in1), [in2_] "r" (in2))
+
+#define CompressedMin(res, in1, in2) asm volatile(                      \
+                                                   "pv.min.t %[res_], %[in1_], %[in2_];" : [res_] "=r" (res) : [in1_] "r" (in1), [in2_] "r" (in2))
+
+#define Max16(res, in1, in2) asm volatile(                      \
+                                                   "pv.max.c %[res_], %[in1_], %[in2_];" : [res_] "=r" (res) : [in1_] "r" (in1), [in2_] "r" (in2))
+
+#define Min16(res, in1, in2) asm volatile(                      \
+                                                                        "pv.min.c %[res_], %[in1_], %[in2_];" : [res_] "=r" (res) : [in1_] "r" (in1), [in2_] "r" (in2))
+
+#define thr_cmp(state, val, threshs)  __builtin_pulp_thresh_compr(state, val, threshs)
+
+typedef unsigned char  v4u __attribute__((vector_size (4)));
+typedef   signed char  v4s __attribute__((vector_size (4)));
+
+#define bitext(x,size,off)                                      __builtin_pulp_bextract(x,size,off)
+#define bitextu(x,size,off)                                     __builtin_pulp_bextractu(x,size,off)
+#ifdef __clang__
+#define bitins(dst,not_mask_imm,src,mask_imm,off)               __builtin_binsert(dst,not_mask_imm,src,mask_imm,off)
+#else
+#define bitins(dst,not_mask_imm,src,mask_imm,off)               __builtin_pulp_binsert(dst,not_mask_imm,src,mask_imm,off)
+#endif
+#define pack(x,y,z,t)                                           __builtin_pulp_pack4(x,y,z,t)
+#define max4(a,b)                                               __builtin_pulp_maxu4(a,b)
+#define maxs4(a, b)                                             __builtin_pulp_max4(a, b)
+#define max8(a, b)                                              __builtin_pulp_maxu8(a, b)
+#define maxs8(a, b)                                             __builtin_pulp_max8(a, b)
+#define max16(a, b)                                             __builtin_pulp_maxu16(a, b)
+#define maxs16(a, b)                                            __builtin_pulp_max16(a, b)
+#define maxs20(a, b)                                            __builtin_pulp_max20(a, b)
+#define max32(a,b)                                              __builtin_pulp_maxusi(a,b)
+#define maxs32(a,b)                                             __builtin_pulp_maxsi(a,b)
+#define min32(a,b)                                              __builtin_pulp_minusi(a,b)
+#define mins32(a,b)                                             __builtin_pulp_minsi(a,b)
+#define min4(a, b)                                              __builtin_pulp_minu4(a, b)
+#define mins4(a, b)                                             __builtin_pulp_min4(a, b)
+#define min8(a, b)                                              __builtin_pulp_minu8(a, b)
+#define mins8(a, b)                                             __builtin_pulp_min8(a, b)
+#define min16(a, b)                                             __builtin_pulp_minu16(a, b)
+#define mins16(a, b)                                            __builtin_pulp_min16(a, b)
+#define mins20(a, b)                                            __builtin_pulp_min20(a, b)
+#define avg4(a,b)                                               __builtin_pulp_avgu4(a,b)
+#define avg8(a,b)                                               __builtin_pulp_avgu8(a,b)
+#define avg16(a,b)                                              __builtin_pulp_avgu16(a,b)
+#define log2(x)                                                 __builtin_pulp_fl1(x)
+#define min(a,b)                                                ((a)<(b)?(a):(b))
+#define SumDotp4(a, b, c)                                       __builtin_pulp_sdotusp4(a, b, c)
+#define SumDotp8(a, b, c)                                       __builtin_pulp_sdotusp8(a, b, c)
+#define SumDotp16(a, b, c)                                      __builtin_pulp_sdotusp16(a, b, c)
+#define SumDotps4(a, b, c)                                      __builtin_pulp_sdotsp4(a, b, c)
+#define SumDotps8(a, b, c)                                      __builtin_pulp_sdotsp8(a, b, c)
+#define SumDotps16(a, b, c)                                     __builtin_pulp_sdotsp16(a, b, c)
+#define clip4(x)                                                __builtin_pulp_clipu_r(x, 15)
+#define clip2(x)                                                __builtin_pulp_clipu_r(x, 3)
+#define clip8(x)                                                __builtin_pulp_clipu_r(x, 255)
+
+#define clips4(x)                                               __builtin_pulp_clip_r(x, 7)
+#define clips2(x)                                               __builtin_pulp_clip_r(x, 1)
+#define clips8(x)                                               __builtin_pulp_clip_r(x, 127)
+#define MacLoadInit(a_update, b_update, a_reg, b_reg, ptr)      __builtin_pulp_mlinitspr_v3(a_update, b_update, a_reg, b_reg, ptr)
+#define MacLoadUpdate(ptr)                                      __builtin_pulp_mlupdatespr_v3(ptr)
+#define MacLoadAssign(ptr)                                      __builtin_pulp_mlassignspr_v3(ptr)
+#define MacLoad4(a_update, b_update, a_reg, b_reg, ptr, sum)    __builtin_pulp_mlsdotsup4_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+#define MacLoad8(a_update, b_update, a_reg, b_reg, ptr, sum)    __builtin_pulp_mlsdotsup8_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+#define MacLoad16(a_update, b_update, a_reg, b_reg, ptr, sum)   __builtin_pulp_mlsdotsup16_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+#define MacLoads4(a_update, b_update, a_reg, b_reg, ptr, sum)   __builtin_pulp_mlsdotsp4_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+#define MacLoads8(a_update, b_update, a_reg, b_reg, ptr, sum)   __builtin_pulp_mlsdotsp8_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+#define MacLoads16(a_update, b_update, a_reg, b_reg, ptr, sum)  __builtin_pulp_mlsdotsp16_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+
+#define PACK_INT8_SIZE(x)                                    (x)
+#define PACK_INT4_SIZE(x)                                    ((x) >> 1)
+#define PACK_INT2_SIZE(x)                                    ((x) >> 2)
+
+#define MemoryFence()                                        asm volatile("":::"memory")
+
+#define LEGACY_MODE(x)                                       asm volatile ("csrwi 0x010," x)
+#define IVEC_FMT(x)                                          asm volatile ("csrwi 0x00D," x)
+#define MIXED_SKIP(x)                                        asm volatile ("csrwi 0x00F," x)
+#define A_ADDRESS(x)                                         asm volatile ("csrw 0x100, %0" :: "r" (x))
+#define W_ADDRESS(x)                                         asm volatile ("csrw 0x101, %0" :: "r" (x))
+#define A_STRIDE(x)                                          asm volatile ("csrw 0x102, %0":: "r" (x))
+#define W_STRIDE(x)                                          asm volatile ("csrw 0x103, %0":: "r" (x))
+#define A_ROLLBACK(x)                                        asm volatile ("csrw 0x104, %0":: "r" (x))
+#define W_ROLLBACK(x)                                        asm volatile ("csrw 0x105, %0":: "r" (x))
+#define A_SKIP(x)                                            asm volatile ("csrwi 0x106," x)
+#define W_SKIP(x)                                            asm volatile ("csrwi 0x107," x)
+
+static uint8_t __attribute__((noinline)) pulp_nn_quant_u2 (
+  int32_t phi,
+  int16_t m,
+  int8_t d
+  ) {
+  int32_t x = (m * phi) >> d;
+  uint8_t res = clip2(x);
+  return res;
+}
+static uint8_t __attribute__((noinline)) pulp_nn_bn_quant_u2 (
+  int32_t phi,
+  int32_t k,
+  int32_t lambda,
+  int8_t d
+  ) {
+  int32_t integer_image_phi = (k * phi) + lambda;
+  int32_t x = (integer_image_phi) >> d;
+  uint8_t res = clip2(x);
+  return res;
+}
+static uint8_t __attribute__((noinline)) pulp_nn_add_quant_u2 (
+  uint8_t pix1,
+  uint8_t pix2,
+  int16_t m1,
+  int16_t m2,
+  int8_t d
+  ) {
+  uint32_t integer_image = pix1*m1 + pix2*m2;
+  uint32_t x = (integer_image) >> d;
+  uint8_t res = clip2(x);
+  return res;
+}
+
+static int8_t __attribute__((noinline)) pulp_nn_quant_i2 (
+  int32_t phi,
+  int16_t m,
+  int8_t d
+  ) {
+  int32_t x = (m * phi) >> d;
+  int8_t res = clips2(x);
+  return res;
+}
+static int8_t __attribute__((noinline)) pulp_nn_bn_quant_i2 (
+  int32_t phi,
+  int32_t k,
+  int32_t lambda,
+  int8_t d
+  ) {
+  int32_t integer_image_phi = (k * phi) + lambda;
+  int32_t x = (integer_image_phi) >> d;
+  int8_t res = clips2(x);
+  return res;
+}
+static int8_t __attribute__((noinline)) pulp_nn_add_quant_i2 (
+  int8_t pix1,
+  int8_t pix2,
+  int16_t m1,
+  int16_t m2,
+  int8_t d
+  ) {
+  int32_t integer_image = pix1*m1 + pix2*m2;
+  int32_t x = (integer_image) >> d;
+  int8_t res = clips2(x);
+  return res;
+}
+
+static uint8_t __attribute__((noinline)) pulp_nn_quant_u4 (
+  int32_t phi,
+  int16_t m,
+  int8_t d
+  ) {
+  int32_t x = (m * phi) >> d;
+  uint8_t res = clip4(x);
+  return res;
+}
+static uint8_t __attribute__((noinline)) pulp_nn_bn_quant_u4 (
+  int32_t phi,
+  int32_t k,
+  int32_t lambda,
+  int8_t d
+  ) {
+  int32_t integer_image_phi = (k * phi) + lambda;
+  int32_t x = (integer_image_phi) >> d;
+  uint8_t res = clip4(x);
+  return res;
+}
+static uint8_t __attribute__((noinline)) pulp_nn_add_quant_u4 (
+  uint8_t pix1,
+  uint8_t pix2,
+  int16_t m1,
+  int16_t m2,
+  int8_t d
+  ) {
+  uint32_t integer_image = pix1*m1 + pix2*m2;
+  uint32_t x = (integer_image) >> d;
+  uint8_t res = clip4(x);
+  return res;
+}
+
+static int8_t __attribute__((noinline)) pulp_nn_quant_i4 (
+  int32_t phi,
+  int16_t m,
+  int8_t d
+  ) {
+  int32_t x = (m * phi) >> d;
+  int8_t res = clips4(x);
+  return res;
+}
+static int8_t __attribute__((noinline)) pulp_nn_bn_quant_i4 (
+  int32_t phi,
+  int32_t k,
+  int32_t lambda,
+  int8_t d
+  ) {
+  int32_t integer_image_phi = (k * phi) + lambda;
+  int32_t x = (integer_image_phi) >> d;
+  int8_t res = clips4(x);
+  return res;
+}
+static int8_t __attribute__((noinline)) pulp_nn_add_quant_i4 (
+  int8_t pix1,
+  int8_t pix2,
+  int16_t m1,
+  int16_t m2,
+  int8_t d
+  ) {
+  int32_t integer_image = pix1*m1 + pix2*m2;
+  int32_t x = (integer_image) >> d;
+  int8_t res = clips4(x);
+  return res;
+}
+
+static uint8_t __attribute__((noinline)) pulp_nn_quant_u8 (
+  int32_t phi,
+  int16_t m,
+  int8_t d
+  ) {
+  int32_t x = (m * phi) >> d;
+  uint8_t res = clip8(x);
+  return res;
+}
+static uint8_t __attribute__((noinline)) pulp_nn_bn_quant_u8 (
+  int32_t phi,
+  int32_t k,
+  int32_t lambda,
+  int8_t d
+  ) {
+  int32_t integer_image_phi = (k * phi) + lambda;
+  int32_t x = (integer_image_phi) >> d;
+  uint8_t res = clip8(x);
+  return res;
+}
+static uint8_t __attribute__((noinline)) pulp_nn_add_quant_u8 (
+  uint8_t pix1,
+  uint8_t pix2,
+  int16_t m1,
+  int16_t m2,
+  int8_t d
+  ) {
+  uint32_t integer_image = pix1*m1 + pix2*m2;
+  uint32_t x = (integer_image) >> d;
+  uint8_t res = clip8(x);
+  return res;
+}
+
+static int8_t __attribute__((noinline)) pulp_nn_quant_i8 (
+  int32_t phi,
+  int16_t m,
+  int8_t d
+  ) {
+  int32_t x = (m * phi) >> d;
+  int8_t res = clips8(x);
+  return res;
+}
+static int8_t __attribute__((noinline)) pulp_nn_bn_quant_i8 (
+  int32_t phi,
+  int32_t k,
+  int32_t lambda,
+  int8_t d
+  ) {
+  int32_t integer_image_phi = (k * phi) + lambda;
+  int32_t x = (integer_image_phi) >> d;
+  int8_t res = clips8(x);
+  return res;
+}
+static int8_t __attribute__((noinline)) pulp_nn_add_quant_i8 (
+  int8_t pix1,
+  int8_t pix2,
+  int16_t m1,
+  int16_t m2,
+  int8_t d
+  ) {
+  int32_t integer_image = pix1*m1 + pix2*m2;
+  int32_t x = (integer_image) >> d;
+  int8_t res = clips8(x);
+  return res;
+}
+
+
+static uint8_t __attribute__((noinline)) pulp_nn_u4_quant(int input, int16_t * pThr)
+{
+  if(input <= pThr[7] )
+  {
+    if( input <= pThr[3])
+    {
+      if( input <= pThr[1])
+      {
+        if( input <= pThr[0])
+          return 0;
+        else
+          return 1;
+      }
+      else
+      {
+        if( input <= pThr[2])
+          return 2;
+        else
+          return 3;
+      }
+    }
+    else
+    {
+      if( input <= pThr[5])
+      {
+        if( input <= pThr[4])
+          return 4;
+        else
+          return 5;
+      }
+      else
+      {
+        if( input <= pThr[6])
+          return 6;
+        else
+          return 7;
+      }
+    }
+  }
+  else
+  {
+    if( input <= pThr[11])
+    {
+      if( input <= pThr[9])
+      {
+        if( input <= pThr[8])
+          return 8;
+        else
+          return 9;
+      }
+      else
+      {
+        if( input <= pThr[10])
+          return 10;
+        else
+          return 11;
+      }
+    }
+    else
+    {
+      if( input <= pThr[13])
+      {
+        if( input <= pThr[12])
+          return 12;
+        else
+          return 13;
+      }
+      else
+      {
+        if( input <= pThr[14])
+          return 14;
+        else
+          return 15;
+      }
+    }
+  }
+}
+
+static uint8_t __attribute__((noinline)) pulp_nn_u2_quant(int input, int16_t * pThr)
+{
+  if( input <= pThr[1])
+  {
+    if( input <= pThr[0])
+        {
+      return 0;
+    }
+        else
+        {
+      return 1;
+    }
+  }
+    else
+    {
+    if( input <= pThr[2])
+        {
+      return 2;
+    }
+        else
+        {
+      return 3;
+    }
+  }
+}
+
+/*
+ * Common
+ */
+
+
+static v4s __attribute__((noinline)) pulp_nn_i4_to_i8_r( int8_t *pSrc)
+{
+  v4s Src = *((v4s*) pSrc);
+  int8_t bext1, bext2, bext3, bext4;
+
+  bext1 = (int8_t) bitext((int) Src, 4, 0);
+  bext2 = (int8_t) bitext((int) Src, 4, 4);
+  bext3 = (int8_t) bitext((int) Src, 4, 8);
+  bext4 = (int8_t) bitext((int) Src, 4, 12);
+  v4s res = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+
+  return res;
+}
+
+static v4u __attribute__((noinline)) pulp_nn_u4_to_u8_r(uint8_t *pSrc)
+{
+  v4u Src = *((v4u*) pSrc);
+  uint8_t bext1, bext2, bext3, bext4;
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 4, 0);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 4, 4);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 4, 8);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 4, 12);
+  v4u res = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+
+  return res;
+}
+
+static v4s __attribute__((noinline)) pulp_nn_i2_to_i8_r( int8_t *pSrc)
+{
+  v4s Src = *((v4s*) pSrc);
+  int8_t bext1, bext2, bext3, bext4;
+
+  bext1 = (int8_t) bitext((int) Src, 2, 0);
+  bext2 = (int8_t) bitext((int) Src, 2, 2);
+  bext3 = (int8_t) bitext((int) Src, 2, 4);
+  bext4 = (int8_t) bitext((int) Src, 2, 6);
+  v4s res = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+
+  return res;
+}
+
+static v4u __attribute__((noinline)) pulp_nn_u2_to_u8_r(uint8_t *pSrc)
+{
+  v4u Src = *((v4u*) pSrc);
+  uint8_t bext1, bext2, bext3, bext4;
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 0);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 2);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 4);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 6);
+  v4u res = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+
+  return res;
+}
+
+static v4s __attribute__((noinline)) pulp_nn_i2_to_i4_r(int8_t *pSrc)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  v4s Src = *((v4s*) pSrc);
+  int8_t bext1, bext2, bext3, bext4;
+  int8_t out1, out2, out3, out4;
+
+  bext1 = (int8_t) bitextu((int) Src, 2, 0);
+  bext2 = (int8_t) bitextu((int) Src, 2, 2);
+  bext3 = (int8_t) bitextu((int) Src, 2, 4);
+  bext4 = (int8_t) bitextu((int) Src, 2, 6);
+
+  out1 = bitins(bext1, n_mask, bext2, mask, off);
+  out2 = bitins(bext3, n_mask, bext4, mask, off);
+
+  bext1 = (int8_t) bitextu((int) Src, 2, 8);
+  bext2 = (int8_t) bitextu((int) Src, 2, 10);
+  bext3 = (int8_t) bitextu((int) Src, 2, 12);
+  bext4 = (int8_t) bitextu((int) Src, 2, 14);
+
+  out3 = bitins(bext1, n_mask, bext2, mask, off);
+  out4 = bitins(bext3, n_mask, bext4, mask, off);
+
+  v4s res = pack((int8_t) out1, (int8_t) out2, (int8_t) out3, (int8_t) out4);
+
+  return res;
+}
+
+static v4u __attribute__((noinline)) pulp_nn_u2_to_u4_r(uint8_t *pSrc)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  v4u Src = *((v4u*) pSrc);
+  uint8_t bext1, bext2, bext3, bext4;
+  uint8_t out1, out2, out3, out4;
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 0);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 2);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 4);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 6);
+
+  out1 = bitins(bext1, n_mask, bext2, mask, off);
+  out2 = bitins(bext3, n_mask, bext4, mask, off);
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 8);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 10);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 12);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 14);
+
+  out3 = bitins(bext1, n_mask, bext2, mask, off);
+  out4 = bitins(bext3, n_mask, bext4, mask, off);
+
+  v4u res = pack((uint8_t) out1, (uint8_t) out2, (uint8_t) out3, (uint8_t) out4);
+
+  return res;
+}
+
+static int8_t *__attribute__((always_inline)) pulp_nn_i4_to_i8( int8_t *pSrc, int8_t *pDst)
+{
+  v4s Src = *((v4s*) pSrc);
+  int8_t bext1, bext2, bext3, bext4;
+
+  pSrc+=4;
+
+  bext1 = (int8_t) bitext((int) Src, 4, 0);
+  bext2 = (int8_t) bitext((int) Src, 4, 4);
+  bext3 = (int8_t) bitext((int) Src, 4, 8);
+  bext4 = (int8_t) bitext((int) Src, 4, 12);
+  *((v4s*)pDst) = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+   MemoryFence();
+  pDst+=4;
+  bext1 = (int8_t) bitext((int) Src, 4, 16);
+  bext2 = (int8_t) bitext((int) Src, 4, 20);
+  bext3 = (int8_t) bitext((int) Src, 4, 24);
+  bext4 = (int8_t) bitext((int) Src, 4, 28);
+  *((v4s*)pDst) = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+
+  return pSrc;
+}
+
+static uint8_t *__attribute__((always_inline)) pulp_nn_u4_to_u8(uint8_t *pSrc, uint8_t *pDst)
+{
+  v4u Src = *((v4u*) pSrc);
+  uint8_t bext1, bext2, bext3, bext4;
+
+  pSrc+=4;
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 4, 0);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 4, 4);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 4, 8);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 4, 12);
+  *((v4u*)pDst) = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 4, 16);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 4, 20);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 4, 24);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 4, 28);
+  *((v4u*)pDst) = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+
+  return pSrc;
+}
+
+static int8_t *__attribute__((always_inline)) pulp_nn_i2_to_i8( int8_t * pSrc, int8_t * pDst)
+{
+  v4s Src = *((v4s*) pSrc);
+  int8_t bext1, bext2, bext3, bext4;
+
+  pSrc+=4;
+
+  bext1 = (int8_t) bitext((int) Src, 2, 0);
+  bext2 = (int8_t) bitext((int) Src, 2, 2);
+  bext3 = (int8_t) bitext((int) Src, 2, 4);
+  bext4 = (int8_t) bitext((int) Src, 2, 6);
+  *((v4s*)pDst) = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (int8_t) bitext((int) Src, 2, 8);
+  bext2 = (int8_t) bitext((int) Src, 2, 10);
+  bext3 = (int8_t) bitext((int) Src, 2, 12);
+  bext4 = (int8_t) bitext((int) Src, 2, 14);
+  *((v4s*)pDst) = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (int8_t) bitext((int) Src, 2, 16);
+  bext2 = (int8_t) bitext((int) Src, 2, 18);
+  bext3 = (int8_t) bitext((int) Src, 2, 20);
+  bext4 = (int8_t) bitext((int) Src, 2, 22);
+  *((v4s*)pDst) = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (int8_t) bitext((int) Src, 2, 24);
+  bext2 = (int8_t) bitext((int) Src, 2, 26);
+  bext3 = (int8_t) bitext((int) Src, 2, 28);
+  bext4 = (int8_t) bitext((int) Src, 2, 30);
+  *((v4s*)pDst) = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+
+  return pSrc;
+}
+
+static uint8_t *__attribute__((always_inline)) pulp_nn_u2_to_u8(uint8_t * pSrc, uint8_t * pDst)
+{
+  v4u Src = *((v4u*) pSrc);
+  uint8_t bext1, bext2, bext3, bext4;
+
+  pSrc+=4;
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 0);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 2);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 4);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 6);
+  *((v4u*)pDst) = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 8);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 10);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 12);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 14);
+  *((v4u*)pDst) = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 16);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 18);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 20);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 22);
+  *((v4u*)pDst) = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 24);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 26);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 28);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 30);
+  *((v4u*)pDst) = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+
+  return pSrc;
+}
+
+static int8_t *__attribute__((always_inline)) pulp_nn_i2_to_i4( int8_t * pSrc, int8_t * pDst)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  v4s Src = *((v4s*) pSrc);
+  int8_t bext1, bext2, bext3, bext4;
+  int8_t out1, out2, out3, out4;
+
+  pSrc+=4;
+
+  bext1 = (int8_t) bitext((int) Src, 2, 0);
+  bext2 = (int8_t) bitext((int) Src, 2, 2);
+  bext3 = (int8_t) bitext((int) Src, 2, 4);
+  bext4 = (int8_t) bitext((int) Src, 2, 6);
+
+  out1 = bitins(bext1, n_mask, bext2, mask, off);
+  out2 = bitins(bext3, n_mask, bext4, mask, off);
+  MemoryFence();
+
+  bext1 = (int8_t) bitext((int) Src, 2, 8);
+  bext2 = (int8_t) bitext((int) Src, 2, 10);
+  bext3 = (int8_t) bitext((int) Src, 2, 12);
+  bext4 = (int8_t) bitext((int) Src, 2, 14);
+
+  out3 = bitins(bext1, n_mask, bext2, mask, off);
+  out4 = bitins(bext3, n_mask, bext4, mask, off);
+
+  *((v4s*)pDst) = pack((int8_t) out1, (int8_t) out2, (int8_t) out3, (int8_t) out4);
+  MemoryFence();
+
+  pDst+=4;
+  bext1 = (int8_t) bitext((int) Src, 2, 16);
+  bext2 = (int8_t) bitext((int) Src, 2, 18);
+  bext3 = (int8_t) bitext((int) Src, 2, 20);
+  bext4 = (int8_t) bitext((int) Src, 2, 22);
+
+  out1 = bitins(bext1, n_mask, bext2, mask, off);
+  out2 = bitins(bext3, n_mask, bext4, mask, off);
+  MemoryFence();
+
+  bext1 = (int8_t) bitext((int) Src, 2, 24);
+  bext2 = (int8_t) bitext((int) Src, 2, 26);
+  bext3 = (int8_t) bitext((int) Src, 2, 28);
+  bext4 = (int8_t) bitext((int) Src, 2, 30);
+
+  out3 = bitins(bext1, n_mask, bext2, mask, off);
+  out4 = bitins(bext3, n_mask, bext4, mask, off);
+
+  *((v4s*)pDst) = pack((int8_t) out1, (int8_t) out2, (int8_t) out3, (int8_t) out4);
+
+  return pSrc;
+}
+
+static uint8_t *__attribute__((always_inline)) pulp_nn_u2_to_u4( uint8_t * pSrc, uint8_t * pDst)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  v4u Src = *((v4u*) pSrc);
+  uint8_t bext1, bext2, bext3, bext4;
+  uint8_t out1, out2, out3, out4;
+
+  pSrc+=4;
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 0);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 2);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 4);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 6);
+
+  out1 = bitins(bext1, n_mask, bext2, mask, off);
+  out2 = bitins(bext3, n_mask, bext4, mask, off);
+  MemoryFence();
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 8);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 10);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 12);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 14);
+
+  out3 = bitins(bext1, n_mask, bext2, mask, off);
+  out4 = bitins(bext3, n_mask, bext4, mask, off);
+
+  *((v4u*)pDst) = pack((uint8_t) out1, (uint8_t) out2, (uint8_t) out3, (uint8_t) out4);
+  MemoryFence();
+
+  pDst+=4;
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 16);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 18);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 20);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 22);
+
+  out1 = bitins(bext1, n_mask, bext2, mask, off);
+  out2 = bitins(bext3, n_mask, bext4, mask, off);
+  MemoryFence();
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 24);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 26);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 28);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 30);
+
+  out3 = bitins(bext1, n_mask, bext2, mask, off);
+  out4 = bitins(bext3, n_mask, bext4, mask, off);
+  
+  *((v4u*)pDst) = pack((uint8_t) out1, (uint8_t) out2, (uint8_t) out3, (uint8_t) out4);
+
+  return pSrc;
+}
+
+/*
+ * XpulpV2
+ */
+
+static void __attribute__((noinline)) pulp_zero_mem(uint8_t * pBuffer, unsigned int size)
+{
+  int lfover = size &0x3;
+  for (int i=0; i<(size>>2); i++)
+  {
+    *((v4u *)pBuffer) = (v4u){0,0,0,0};
+    MemoryFence();
+    pBuffer+=4;
+  }
+  while(lfover)
+  {
+    *pBuffer++=0;
+    lfover--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_im2col_u2_to_u8(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_u2_to_u8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 16;
+  }
+  while (lfover)
+  {
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 0);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 2);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 4);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 6);
+    pOutput++;
+    pInput++;
+    lfover -= 4;
+  }
+}
+static void __attribute__((noinline)) pulp_nn_im2col_i2_to_i8(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_i2_to_i8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 16;
+  }
+  while (lfover)
+  {
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 0);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 2);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 4);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 6);
+    pOutput++;
+    pInput++;
+    lfover -= 4;
+  }
+}
+static void __attribute__((noinline)) pulp_nn_im2col_u4_to_u8(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 3u;
+  int lfover = blockSize & 0x07;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_u4_to_u8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 8;
+  }
+  while (lfover)
+  {
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 4, 0);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 4, 4);
+    pOutput++;
+    pInput++;
+    lfover -= 2;
+  }
+}
+static void __attribute__((noinline)) pulp_nn_im2col_i4_to_i8(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 3u;
+  int lfover = blockSize & 0x07;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_i4_to_i8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 8;
+  }
+  while (lfover)
+  {
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 4, 0);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 4, 4);
+    pOutput++;
+    pInput++;
+    lfover -= 2;
+  }
+}
+static void __attribute__((noinline)) pulp_nn_im2col_u8_to_u8(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 2u;
+  int lfover = blockSize & 0x03;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4u*)pOutput) = *((v4u*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    *((uint8_t*)pOutput) = *((uint8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover--;
+  }
+}
+static void __attribute__((noinline)) pulp_nn_im2col_i8_to_i8(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 2u;
+  int lfover = blockSize & 0x03;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4s*)pOutput) = *((v4s*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    *((int8_t*)pOutput) = *((int8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover--;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_im2col_u2_to_u2(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4u*)pOutput) = *((v4u*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    uint8_t extr;
+    *((uint8_t*)pOutput) = *((uint8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover -= 4;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_i2_to_i2(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4s*)pOutput) = *((v4s*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    int8_t extr;
+    *((int8_t*)pOutput) = *((int8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover -= 4;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_u2_to_u4(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_u2_to_u4(pInput, pOutput);
+    MemoryFence();
+    pOutput += 8;
+  }
+  while (lfover)
+  {
+    uint8_t extr;
+    // TODO: this is potentially dangerous/wrong if the number of channels
+    // is not a multiple of 8!
+    *((v4u *)pOutput) = pulp_nn_u2_to_u4_r(pInput);
+    pInput += 2;
+    pOutput += 4;
+    lfover -= 8;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_i2_to_i4(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_i2_to_i4(pInput, pOutput);
+    MemoryFence();
+    pOutput += 8;
+  }
+  while (lfover)
+  {
+    int8_t extr;
+    // TODO: this is potentially dangerous/wrong if the number of channels
+    // is not a multiple of 8!
+    *((v4s *)pOutput) = pulp_nn_i2_to_i4_r(pInput);
+    pInput += 2;
+    pOutput += 4;
+    lfover -= 8;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_u2_to_u8(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_u2_to_u8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 16;
+  }
+  while (lfover)
+  {
+    uint8_t extr;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 0);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 2);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 4);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 6);
+    pOutput++;
+    pInput++;
+    lfover -= 4;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_i2_to_i8(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_i2_to_i8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 16;
+  }
+  while (lfover)
+  {
+    int8_t extr;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 0);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 2);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 4);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 6);
+    pOutput++;
+    pInput++;
+    lfover -= 4;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_u4_to_u4(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 3u;
+  int lfover = blockSize & 0x07;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4u*)pOutput) = *((v4u*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    uint8_t extr;
+    *((uint8_t*)pOutput) = *((uint8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover -= 2;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_i4_to_i4(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 3u;
+  int lfover = blockSize & 0x07;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4s*)pOutput) = *((v4s*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    int8_t extr;
+    *((int8_t*)pOutput) = *((int8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover -= 2;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_u4_to_u8(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 3u;
+  int lfover = blockSize & 0x07;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_u4_to_u8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 8;
+  }
+  while (lfover)
+  {
+    uint8_t extr;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 4, 0);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 4, 4);
+    pOutput++;
+    pInput++;
+    lfover -= 2;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_i4_to_i8(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 3u;
+  int lfover = blockSize & 0x07;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_i4_to_i8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 8;
+  }
+  while (lfover)
+  {
+    int8_t extr;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 4, 0);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 4, 4);
+    pOutput++;
+    pInput++;
+    lfover -= 2;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_u8_to_u8(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 2u;
+  int lfover = blockSize & 0x03;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4u*)pOutput) = *((v4u*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    uint8_t extr;
+    *((uint8_t*)pOutput) = *((uint8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover -= 1;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_i8_to_i8(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 2u;
+  int lfover = blockSize & 0x03;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4s*)pOutput) = *((v4s*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    int8_t extr;
+    *((int8_t*)pOutput) = *((int8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover -= 1;
+  }
+}
+
+
+
+static void __attribute__((noinline)) pulp_nn_compare_and_replace_if_larger_u8(uint8_t * base,
+                                                uint8_t * target,
+                                                uint16_t length)
+{
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+  v4u inp;
+  v4u com;
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    inp = *((v4u*)pIn);
+    com = *((v4u*)pCom);
+
+    *((v4u*)pIn) = max4(inp, com);
+
+    pCom+=4;
+    pIn+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    if(*pIn<*pCom)
+      *pIn=*pCom;
+    
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_compare_and_replace_if_larger_i8(
+    int8_t * base, int8_t * target, uint16_t length) {
+  int8_t *pIn = base;
+  int8_t *pCom = target;
+  v4s inp;
+  v4s com;
+  int cnt = length >> 2;
+
+  while (cnt > 0u) {
+    inp = *((v4s *)pIn);
+    com = *((v4s *)pCom);
+
+    *((v4s *)pIn) = maxs4(inp, com);
+
+    pCom += 4;
+    pIn += 4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left > 0u) {
+    if (*pIn < *pCom)
+      *pIn = *pCom;
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_avg_and_replace_u8(uint8_t * base,
+                                  uint8_t * target,
+                                  uint16_t length)
+{
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+
+  while (length>0u)
+  {
+    *pIn = ((*pIn + *pCom) >> 1);
+
+    pIn++;
+    pCom++;
+    length--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_compare_and_replace_if_larger_u4(uint8_t * base,
+                                                uint8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+  v4u inp[2];
+  v4u com[2];
+  uint8_t *out;
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    pulp_nn_u4_to_u8(pIn, (uint8_t *)inp);
+    pulp_nn_u4_to_u8(pCom, (uint8_t *)com);
+
+    *((v4u *)out) = max4(inp[0], com[0]);
+
+    *((uint8_t*)pIn) = bitins(*out, n_mask, *(out + 1), mask, off);
+    pIn++;
+    *((uint8_t*)pIn) = bitins(*(out + 2), n_mask, *(out + 3), mask, off);
+    pIn++;
+
+    *((v4u *)out) = max4(inp[1], com[1]);
+
+    *((uint8_t*)pIn) = bitins(*out, n_mask, *(out + 1), mask, off);
+    pIn++;
+    *((uint8_t*)pIn) = bitins(*(out + 2), n_mask, *(out + 3), mask, off);
+    pIn++;
+
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 4, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 4, 4);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 4, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 4, 4);
+
+    if(inA0<inB0)
+      inA0=inB0;
+
+    if(inA1<inB1)
+      inA1=inB1;
+
+    *((uint8_t*)pIn) = bitins(inA0, n_mask, inA1, mask, off);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_compare_and_replace_if_larger_i4(int8_t * base,
+                                                int8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  int8_t *pIn = base;
+  int8_t *pCom = target;
+  v4s inp[2];
+  v4s com[2];
+  int8_t *out;
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    pulp_nn_i4_to_i8(pIn, (int8_t *)inp);
+    pulp_nn_i4_to_i8(pCom, (int8_t *)com);
+
+    *((v4s *)out) = maxs4(inp[0], com[0]);
+
+    *((int8_t*)pIn) = bitins(*out, n_mask, *(out + 1), mask, off);
+    pIn++;
+    *((int8_t*)pIn) = bitins(*(out + 2), n_mask, *(out + 3), mask, off);
+    pIn++;
+
+    *((v4s *)out) = maxs4(inp[1], com[1]);
+
+    *((int8_t*)pIn) = bitins(*out, n_mask, *(out + 1), mask, off);
+    pIn++;
+    *((int8_t*)pIn) = bitins(*(out + 2), n_mask, *(out + 3), mask, off);
+    pIn++;
+
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    int8_t inA0 = (int8_t) bitext((int) *pIn, 4, 0);
+    int8_t inA1 = (int8_t) bitext((int) *pIn, 4, 4);
+    int8_t inB0 = (int8_t) bitext((int) *pCom, 4, 0);
+    int8_t inB1 = (int8_t) bitext((int) *pCom, 4, 4);
+
+    if(inA0<inB0)
+      inA0=inB0;
+
+    if(inA1<inB1)
+      inA1=inB1;
+
+    *((int8_t*)pIn) = bitins(inA0, n_mask, inA1, mask, off);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_avg_and_replace_u4(uint8_t * base,
+                                  uint8_t * target,
+                                  uint16_t length)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+
+  while (length>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 4, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 4, 4);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 4, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 4, 4);
+
+    inA0 = ((inA0 + inB0) >> 1);
+    inA1 = ((inA1 + inB1) >> 1);
+
+    *((uint8_t*)pIn) = bitins(inA0, n_mask, inA1, mask, off);
+
+    pIn++;
+    pCom++;
+    length--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_compare_and_replace_if_larger_u2(uint8_t * base,
+                                                uint8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask2 = 0x0c;
+  int8_t n_mask2 = ~ mask2;
+  int8_t mask4 = 0x30;
+  int8_t n_mask4 = ~ mask4;
+  int8_t mask6 = 0xc0;
+  int8_t n_mask6 = ~ mask6;
+  int8_t off2 = 2;
+  int8_t off4 = 4;
+  int8_t off6 = 6;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+  v4u inp[4];
+  v4u com[4];
+  uint8_t *out;
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    pulp_nn_u2_to_u8(pIn, inp);
+    pulp_nn_u2_to_u8(pCom, com);
+
+    *((v4u*)out) = max4(inp[0], com[0]);
+
+    uint8_t inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    *((v4u*)out) = max4(inp[1], com[1]);
+
+    inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    *((v4u*)out) = max4(inp[2], com[2]);
+
+    inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    *((v4u*)out) = max4(inp[3], com[3]);
+
+    inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 2, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 2, 2);
+    uint8_t inA2 = (uint8_t) bitextu((unsigned int) *pIn, 2, 4);
+    uint8_t inA3 = (uint8_t) bitextu((unsigned int) *pIn, 2, 6);
+    v4u inA4 = pack((uint8_t) inA0, (uint8_t) inA1, (uint8_t) inA2, (uint8_t) inA3);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 2, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 2, 2);
+    uint8_t inB2 = (uint8_t) bitextu((unsigned int) *pCom, 2, 4);
+    uint8_t inB3 = (uint8_t) bitextu((unsigned int) *pCom, 2, 6);
+    v4u inB4 = pack((uint8_t) inB0, (uint8_t) inB1, (uint8_t) inB2, (uint8_t) inB3);
+
+    *((v4u*)out) = max4(inA4, inB4);
+
+    uint8_t inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+
+static void __attribute__((noinline)) pulp_nn_compare_and_replace_if_larger_i2(int8_t * base,
+                                                int8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask2 = 0x0c;
+  int8_t n_mask2 = ~ mask2;
+  int8_t mask4 = 0x30;
+  int8_t n_mask4 = ~ mask4;
+  int8_t mask6 = 0xc0;
+  int8_t n_mask6 = ~ mask6;
+  int8_t off2 = 2;
+  int8_t off4 = 4;
+  int8_t off6 = 6;
+
+  int8_t *pIn = base;
+  int8_t *pCom = target;
+  v4s inp[4];
+  v4s com[4];
+  int8_t *out;
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    pulp_nn_i2_to_i8(pIn, inp);
+    pulp_nn_i2_to_i8(pCom, com);
+
+    *((v4s*)out) = maxs4(inp[0], com[0]);
+
+    int8_t inA = (int8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((int8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    *((v4s*)out) = maxs4(inp[1], com[1]);
+
+    inA = (int8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((int8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    *((v4s*)out) = maxs4(inp[2], com[2]);
+
+    inA = (int8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((int8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    *((v4s*)out) = maxs4(inp[3], com[3]);
+
+    inA = (int8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((int8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    int8_t inA0 = (int8_t) bitext((unsigned int) *pIn, 2, 0);
+    int8_t inA1 = (int8_t) bitext((unsigned int) *pIn, 2, 2);
+    int8_t inA2 = (int8_t) bitext((unsigned int) *pIn, 2, 4);
+    int8_t inA3 = (int8_t) bitext((unsigned int) *pIn, 2, 6);
+    v4s inA4 = pack((int8_t) inA0, (int8_t) inA1, (int8_t) inA2, (int8_t) inA3);
+    int8_t inB0 = (int8_t) bitext((unsigned int) *pCom, 2, 0);
+    int8_t inB1 = (int8_t) bitext((unsigned int) *pCom, 2, 2);
+    int8_t inB2 = (int8_t) bitext((unsigned int) *pCom, 2, 4);
+    int8_t inB3 = (int8_t) bitext((unsigned int) *pCom, 2, 6);
+    v4s inB4 = pack((int8_t) inB0, (int8_t) inB1, (int8_t) inB2, (int8_t) inB3);
+
+    *((v4s*)out) = maxs4(inA4, inB4);
+
+    int8_t inA = (int8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((int8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_avg_and_replace_u2(uint8_t * base,
+                                  uint8_t * target,
+                                  uint16_t length)
+{
+  int8_t mask2 = 0x0c;
+  int8_t n_mask2 = ~ mask2;
+  int8_t mask4 = 0x30;
+  int8_t n_mask4 = ~ mask4;
+  int8_t mask6 = 0xc0;
+  int8_t n_mask6 = ~ mask6;
+  int8_t off2 = 2;
+  int8_t off4 = 4;
+  int8_t off6 = 6;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+
+  while (length>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 2, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 2, 2);
+    uint8_t inA2 = (uint8_t) bitextu((unsigned int) *pIn, 2, 4);
+    uint8_t inA3 = (uint8_t) bitextu((unsigned int) *pIn, 2, 6);
+
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 2, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 2, 2);
+    uint8_t inB2 = (uint8_t) bitextu((unsigned int) *pCom, 2, 4);
+    uint8_t inB3 = (uint8_t) bitextu((unsigned int) *pCom, 2, 6);
+
+    inA0 = ((inA0 + inB0) >> 1);
+    inA1 = ((inA1 + inB1) >> 1);
+    inA2 = ((inA2 + inB2) >> 1);
+    inA3 = ((inA3 + inB3) >> 1);
+
+    uint8_t inA = (uint8_t) bitins(inA0, n_mask2, inA1, mask2, off2);
+    inA = bitins(inA, n_mask4, inA2, mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, inA3, mask6, off6);
+
+    pIn++;
+    pCom++;
+    length--;
+  }
+}
+
+/*
+ * XpulpNN
+ */
+
+static void __attribute__((noinline)) xpulp_nn_zero_mem_u8(uint8_t * pBuffer, unsigned int size)
+{
+  int lfover = size &0x3;
+  for (int i=0; i<(size>>2); i++)
+  {
+    *((v4u *)pBuffer) = (v4u){0,0,0,0};
+    MemoryFence();
+    pBuffer+=4;
+  }
+  while(lfover)
+  {
+    *pBuffer++=0;
+    lfover--;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_zero_mem_u4(uint8_t * pBuffer, unsigned int size)
+{
+  int lfover = size &0x7;
+  for (int i=0; i<(size>>3); i++)
+  {
+    *((v4u *)pBuffer) = (v4u){0,0,0,0};
+    MemoryFence();
+    pBuffer+=4;
+  }
+  while(lfover)
+  {
+    *pBuffer++=0;
+    lfover-=2;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_zero_mem_u2(uint8_t * pBuffer, unsigned int size)
+{
+  int lfover = size &0xf;
+  for (int i=0; i<(size>>4); i++)
+  {
+    *((v4u *)pBuffer) = (v4u){0,0,0,0};
+    MemoryFence();
+    pBuffer+=4;
+  }
+  while(lfover)
+  {
+    *pBuffer++=0;
+    lfover-=4;
+  }
+}
+
+
+static void __attribute__((noinline)) xpulp_tnn_zero_mem_ternary(uint8_t * pBuffer, unsigned int size, unsigned int uns)
+{
+  uint8_t pad_val = 0xd9;
+  uint32_t pad_vec = 0xd9d9d9d9;
+  if (uns) {
+    // if we are using an unsigned kernel, we need to pad with -1 because the hardware will add a +1 to ALL values!
+    pad_val = 0xff;
+    pad_vec = 0xffffffff;
+  }
+  int lfover = size &0xf;
+  for (int i=0; i<(size>>4); i++)
+    {
+      *((v4u *)pBuffer) = (v4u)pad_vec;
+      MemoryFence();
+      pBuffer+=4;
+    }
+  while(lfover)
+    {
+      *pBuffer++=pad_val;
+      lfover-=4;
+    }
+}
+
+static void __attribute__((noinline)) xpulp_nn_compare_and_replace_if_larger_u4(uint8_t * base,
+                                                uint8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    *((uint32_t *)pIn) = max8(*((uint32_t *)pIn), *((uint32_t *)pCom));
+
+    pIn+=4;
+    pCom+=4;
+
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 4, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 4, 4);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 4, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 4, 4);
+
+    if(inA0<inB0)
+      inA0=inB0;
+
+    if(inA1<inB1)
+      inA1=inB1;
+
+    *((uint8_t*)pIn) = bitins(inA0, n_mask, inA1, mask, off);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_compare_and_replace_if_larger_i4(
+    int8_t * base, int8_t * target, uint16_t length) {
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~mask;
+  int8_t off = 0x04;
+
+  int8_t *pIn = base;
+  int8_t *pCom = target;
+
+  int cnt = length >> 2;
+
+  while (cnt > 0u) {
+    *((int32_t *)pIn) = maxs8(*((int32_t *)pIn), *((int32_t *)pCom));
+
+    pIn += 4;
+    pCom += 4;
+
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left > 0u) {
+    int8_t inA0 = (int8_t)bitext((int)*pIn, 4, 0);
+    int8_t inA1 = (int8_t)bitext((int)*pIn, 4, 4);
+    int8_t inB0 = (int8_t)bitext((int)*pCom, 4, 0);
+    int8_t inB1 = (int8_t)bitext((int)*pCom, 4, 4);
+
+    if (inA0 < inB0)
+      inA0 = inB0;
+
+    if (inA1 < inB1)
+      inA1 = inB1;
+
+    *((int8_t *)pIn) = bitins(inA0, n_mask, inA1, mask, off);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_avg_and_replace_u4(uint8_t * base,
+                                  uint8_t * target,
+                                  uint16_t length)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+
+  int cnt = length >> 2;
+
+  while (cnt > 0u)
+  {
+    *((uint32_t *)pIn) = avg8(*((uint32_t *)pIn), *((uint32_t *)pCom));
+
+    pIn+=4;
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+
+  while (left>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 4, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 4, 4);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 4, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 4, 4);
+
+    inA0 = ((inA0 + inB0) >> 1);
+    inA1 = ((inA1 + inB1) >> 1);
+
+    *((uint8_t*)pIn) = bitins(inA0, n_mask, inA1, mask, off);
+
+    pIn++;
+    pCom++;
+    length--;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_compare_and_replace_if_larger_u2(uint8_t * base,
+                                                uint8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask2 = 0x0c;
+  int8_t n_mask2 = ~ mask2;
+  int8_t mask4 = 0x30;
+  int8_t n_mask4 = ~ mask4;
+  int8_t mask6 = 0xc0;
+  int8_t n_mask6 = ~ mask6;
+  int8_t off2 = 2;
+  int8_t off4 = 4;
+  int8_t off6 = 6;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+  uint8_t *out;
+
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    *((uint32_t *)pIn) = max16(*((uint32_t *)pIn), *((uint32_t *)pCom));
+
+    pIn+=4;
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 2, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 2, 2);
+    uint8_t inA2 = (uint8_t) bitextu((unsigned int) *pIn, 2, 4);
+    uint8_t inA3 = (uint8_t) bitextu((unsigned int) *pIn, 2, 6);
+    v4u inA4 = pack((uint8_t) inA0, (uint8_t) inA1, (uint8_t) inA2, (uint8_t) inA3);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 2, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 2, 2);
+    uint8_t inB2 = (uint8_t) bitextu((unsigned int) *pCom, 2, 4);
+    uint8_t inB3 = (uint8_t) bitextu((unsigned int) *pCom, 2, 6);
+    v4u inB4 = pack((uint8_t) inB0, (uint8_t) inB1, (uint8_t) inB2, (uint8_t) inB3);
+
+    *((v4u*)out) = max4(inA4, inB4);
+
+    uint8_t inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_compare_and_replace_if_larger_i2(int8_t * base,
+                                                int8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask2 = 0x0c;
+  int8_t n_mask2 = ~ mask2;
+  int8_t mask4 = 0x30;
+  int8_t n_mask4 = ~ mask4;
+  int8_t mask6 = 0xc0;
+  int8_t n_mask6 = ~ mask6;
+  int8_t off2 = 2;
+  int8_t off4 = 4;
+  int8_t off6 = 6;
+
+  int8_t *pIn = base;
+  int8_t *pCom = target;
+  int8_t *out;
+
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    *((int32_t *)pIn) = maxs16(*((int32_t *)pIn), *((int32_t *)pCom));
+
+    pIn+=4;
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    int8_t inA0 = (int8_t) bitext((int) *pIn, 2, 0);
+    int8_t inA1 = (int8_t) bitext((int) *pIn, 2, 2);
+    int8_t inA2 = (int8_t) bitext((int) *pIn, 2, 4);
+    int8_t inA3 = (int8_t) bitext((int) *pIn, 2, 6);
+    v4s inA4 = pack((int8_t) inA0, (int8_t) inA1, (int8_t) inA2, (int8_t) inA3);
+    int8_t inB0 = (int8_t) bitext((int) *pCom, 2, 0);
+    int8_t inB1 = (int8_t) bitext((int) *pCom, 2, 2);
+    int8_t inB2 = (int8_t) bitext((int) *pCom, 2, 4);
+    int8_t inB3 = (int8_t) bitext((int) *pCom, 2, 6);
+    v4s inB4 = pack((int8_t) inB0, (int8_t) inB1, (int8_t) inB2, (int8_t) inB3);
+
+    *((v4s*)out) = maxs4(inA4, inB4);
+
+    int8_t inA = (int8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((int8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_tnn_compare_and_replace_if_larger_ternary(int8_t * base,
+                                                int8_t * target,
+                                                uint16_t length)
+{
+  uint8_t mask2 = 0x0c;
+  uint8_t n_mask2 = ~ mask2;
+  uint8_t mask4 = 0x30;
+  uint8_t n_mask4 = ~ mask4;
+  uint8_t mask6 = 0xc0;
+  uint8_t n_mask6 = ~ mask6;
+  uint8_t off2 = 2;
+  uint8_t off4 = 4;
+  uint8_t off6 = 6;
+
+  uint8_t *pIn = (uint8_t *) base;
+  uint8_t *pCom = (uint8_t *) target;
+  uint8_t *out;
+
+  int cnt = length >> 2;
+  uint32_t result;
+
+  while(cnt > 0u)
+  {
+    uint32_t in1 = *((uint32_t *)pIn);
+    uint32_t in2 = *((int32_t *)pCom);
+    result = maxs20(in1, in2);
+    *((uint32_t *)pIn) = result;
+
+    pIn+=4;
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  if (left>0u)
+  {
+    // do the vector max on the whole word - we won't use the leftover bytes
+    uint32_t in1 = *((uint32_t *)pIn);
+    uint32_t in2 = *((int32_t *)pCom);
+    result = maxs20(in1, in2);
+
+    // ...and copy back the relevant bytes of the result to pIn
+    for (int i=0; i<left; i++)
+      *((uint8_t*)(pIn + i)) = (uint8_t) (result >> (8*i));
+
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_avg_and_replace_u2(uint8_t * base,
+                                  uint8_t * target,
+                                  uint16_t length)
+{
+  int8_t mask2 = 0x0c;
+  int8_t n_mask2 = ~ mask2;
+  int8_t mask4 = 0x30;
+  int8_t n_mask4 = ~ mask4;
+  int8_t mask6 = 0xc0;
+  int8_t n_mask6 = ~ mask6;
+  int8_t off2 = 2;
+  int8_t off4 = 4;
+  int8_t off6 = 6;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+  uint8_t *out;
+
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    *((uint32_t *)pIn) = avg16(*((uint32_t *)pIn), *((uint32_t *)pCom));
+
+    pIn+=4;
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 2, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 2, 2);
+    uint8_t inA2 = (uint8_t) bitextu((unsigned int) *pIn, 2, 4);
+    uint8_t inA3 = (uint8_t) bitextu((unsigned int) *pIn, 2, 6);
+    v4u inA4 = pack((uint8_t) inA0, (uint8_t) inA1, (uint8_t) inA2, (uint8_t) inA3);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 2, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 2, 2);
+    uint8_t inB2 = (uint8_t) bitextu((unsigned int) *pCom, 2, 4);
+    uint8_t inB3 = (uint8_t) bitextu((unsigned int) *pCom, 2, 6);
+    v4u inB4 = pack((uint8_t) inB0, (uint8_t) inB1, (uint8_t) inB2, (uint8_t) inB3);
+
+    *((v4u*)out) = avg4(inA4, inB4);
+
+    uint8_t inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+#endif
diff --git a/rt_nn_tests/xptnn_maxpool/pulp_nn_utils_xpnn.h b/rt_nn_tests/xptnn_maxpool/pulp_nn_utils_xpnn.h
new file mode 100644
index 0000000..0c783ae
--- /dev/null
+++ b/rt_nn_tests/xptnn_maxpool/pulp_nn_utils_xpnn.h
@@ -0,0 +1,1937 @@
+/*
+ * pulp_nn_utils.h
+ * Nazareno   Bruschi  <nazareno.bruschi@unibo.it>
+ * Alessandro Nadalini <alessandro.nadalini3@unibo.it>
+ * Georg Rutishauser   <georgr@iis.ee.ethz.ch>
+ *
+ * Copyright (C) 2019-2020 ETH Zurich & University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef __PULPNN_UTILS__
+#define __PULPNN_UTILS__
+
+#include "pmsis.h"
+#ifdef GAP_SDK
+#include "pulp.h"
+#endif
+
+#define bitext(x,size,off)                                      __builtin_pulp_bextract(x,size,off)
+#define bitextu(x,size,off)                                     __builtin_pulp_bextractu(x,size,off)
+#define bitins(dst,not_mask_imm,src,mask_imm,off)               __builtin_pulp_binsert(dst,not_mask_imm,src,mask_imm,off)
+#define pack(x,y,z,t)                                           __builtin_pulp_pack4(x,y,z,t)
+#define max4(a,b)                                               __builtin_pulp_maxu4(a,b)
+#define maxs4(a, b)                                             __builtin_pulp_max4(a, b)
+#define max8(a, b)                                              __builtin_pulp_maxu8(a, b)
+#define maxs8(a, b)                                             __builtin_pulp_max8(a, b)
+#define max16(a, b)                                             __builtin_pulp_maxu16(a, b)
+#define maxs16(a, b)                                            __builtin_pulp_max16(a, b)
+#define max32(a,b)                                              __builtin_pulp_maxusi(a,b)
+#define maxs32(a,b)                                             __builtin_pulp_maxsi(a,b)
+#define min32(a,b)                                              __builtin_pulp_minusi(a,b)
+#define mins32(a,b)                                             __builtin_pulp_minsi(a,b)
+#define min4(a, b)                                              __builtin_pulp_minu4(a, b)
+#define mins4(a, b)                                             __builtin_pulp_min4(a, b)
+#define min8(a, b)                                              __builtin_pulp_minu8(a, b)
+#define mins8(a, b)                                             __builtin_pulp_min8(a, b)
+#define min16(a, b)                                             __builtin_pulp_minu16(a, b)
+#define mins16(a, b)                                            __builtin_pulp_min16(a, b)
+#define avg4(a,b)                                               __builtin_pulp_avgu4(a,b)
+#define avg8(a,b)                                               __builtin_pulp_avgu8(a,b)
+#define avg16(a,b)                                              __builtin_pulp_avgu16(a,b)
+#define log2(x)                                                 __builtin_pulp_fl1(x)
+#define min(a,b)                                                ((a)<(b)?(a):(b))
+#define SumDotp4(a, b, c)                                       __builtin_pulp_sdotusp4(a, b, c)
+#define SumDotp8(a, b, c)                                       __builtin_pulp_sdotusp8(a, b, c)
+#define SumDotp16(a, b, c)                                      __builtin_pulp_sdotusp16(a, b, c)
+#define SumDotps4(a, b, c)                                      __builtin_pulp_sdotsp4(a, b, c)
+#define SumDotps8(a, b, c)                                      __builtin_pulp_sdotsp8(a, b, c)
+#define SumDotps16(a, b, c)                                     __builtin_pulp_sdotsp16(a, b, c)
+#define clip4(x)                                                __builtin_pulp_clipu_r(x, 15)
+#define clip2(x)                                                __builtin_pulp_clipu_r(x, 3)
+#define clip8(x)                                                __builtin_pulp_clipu_r(x, 255)
+
+#define clips4(x)                                               __builtin_pulp_clip_r(x, 7)
+#define clips2(x)                                               __builtin_pulp_clip_r(x, 1)
+#define clips8(x)                                               __builtin_pulp_clip_r(x, 127)
+#define MacLoadInit(a_update, b_update, a_reg, b_reg, ptr)      __builtin_pulp_mlinitspr_v3(a_update, b_update, a_reg, b_reg, ptr)
+#define MacLoadUpdate(ptr)                                      __builtin_pulp_mlupdatespr_v3(ptr)
+#define MacLoadAssign(ptr)                                      __builtin_pulp_mlassignspr_v3(ptr)
+#define MacLoad4(a_update, b_update, a_reg, b_reg, ptr, sum)    __builtin_pulp_mlsdotsup4_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+#define MacLoad8(a_update, b_update, a_reg, b_reg, ptr, sum)    __builtin_pulp_mlsdotsup8_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+#define MacLoad16(a_update, b_update, a_reg, b_reg, ptr, sum)   __builtin_pulp_mlsdotsup16_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+#define MacLoads4(a_update, b_update, a_reg, b_reg, ptr, sum)   __builtin_pulp_mlsdotsp4_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+#define MacLoads8(a_update, b_update, a_reg, b_reg, ptr, sum)   __builtin_pulp_mlsdotsp8_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+#define MacLoads16(a_update, b_update, a_reg, b_reg, ptr, sum)  __builtin_pulp_mlsdotsp16_v3(a_update, b_update, a_reg, b_reg, ptr, sum)
+
+#define PACK_INT8_SIZE(x)                                    (x)
+#define PACK_INT4_SIZE(x)                                    ((x) >> 1)
+#define PACK_INT2_SIZE(x)                                    ((x) >> 2)
+
+#define MemoryFence()                                        asm volatile("":::"memory")
+
+#define LEGACY_MODE(x)                                       asm volatile ("csrwi 0x010," x)
+#define IVEC_FMT(x)                                          asm volatile ("csrwi 0x00D," x)
+#define MIXED_SKIP(x)                                        asm volatile ("csrwi 0x00F," x)
+#define A_ADDRESS(x)                                         asm volatile ("csrw 0x100, %0" :: "r" (x))
+#define W_ADDRESS(x)                                         asm volatile ("csrw 0x101, %0" :: "r" (x))
+#define A_STRIDE(x)                                          asm volatile ("csrw 0x102, %0":: "r" (x))
+#define W_STRIDE(x)                                          asm volatile ("csrw 0x103, %0":: "r" (x))
+#define A_ROLLBACK(x)                                        asm volatile ("csrw 0x104, %0":: "r" (x))
+#define W_ROLLBACK(x)                                        asm volatile ("csrw 0x105, %0":: "r" (x))
+#define A_SKIP(x)                                            asm volatile ("csrwi 0x106," x)
+#define W_SKIP(x)                                            asm volatile ("csrwi 0x107," x)
+
+static uint8_t __attribute__((noinline)) pulp_nn_quant_u2 (
+  int32_t phi,
+  int16_t m,
+  int8_t d
+  ) {
+  int32_t x = (m * phi) >> d;
+  uint8_t res = clip2(x);
+  return res;
+}
+static uint8_t __attribute__((noinline)) pulp_nn_bn_quant_u2 (
+  int32_t phi,
+  int32_t k,
+  int32_t lambda,
+  int8_t d
+  ) {
+  int32_t integer_image_phi = (k * phi) + lambda;
+  int32_t x = (integer_image_phi) >> d;
+  uint8_t res = clip2(x);
+  return res;
+}
+static uint8_t __attribute__((noinline)) pulp_nn_add_quant_u2 (
+  uint8_t pix1,
+  uint8_t pix2,
+  int16_t m1,
+  int16_t m2,
+  int8_t d
+  ) {
+  uint32_t integer_image = pix1*m1 + pix2*m2;
+  uint32_t x = (integer_image) >> d;
+  uint8_t res = clip2(x);
+  return res;
+}
+
+static int8_t __attribute__((noinline)) pulp_nn_quant_i2 (
+  int32_t phi,
+  int16_t m,
+  int8_t d
+  ) {
+  int32_t x = (m * phi) >> d;
+  int8_t res = clips2(x);
+  return res;
+}
+static int8_t __attribute__((noinline)) pulp_nn_bn_quant_i2 (
+  int32_t phi,
+  int32_t k,
+  int32_t lambda,
+  int8_t d
+  ) {
+  int32_t integer_image_phi = (k * phi) + lambda;
+  int32_t x = (integer_image_phi) >> d;
+  int8_t res = clips2(x);
+  return res;
+}
+static int8_t __attribute__((noinline)) pulp_nn_add_quant_i2 (
+  int8_t pix1,
+  int8_t pix2,
+  int16_t m1,
+  int16_t m2,
+  int8_t d
+  ) {
+  int32_t integer_image = pix1*m1 + pix2*m2;
+  int32_t x = (integer_image) >> d;
+  int8_t res = clips2(x);
+  return res;
+}
+
+static uint8_t __attribute__((noinline)) pulp_nn_quant_u4 (
+  int32_t phi,
+  int16_t m,
+  int8_t d
+  ) {
+  int32_t x = (m * phi) >> d;
+  uint8_t res = clip4(x);
+  return res;
+}
+static uint8_t __attribute__((noinline)) pulp_nn_bn_quant_u4 (
+  int32_t phi,
+  int32_t k,
+  int32_t lambda,
+  int8_t d
+  ) {
+  int32_t integer_image_phi = (k * phi) + lambda;
+  int32_t x = (integer_image_phi) >> d;
+  uint8_t res = clip4(x);
+  return res;
+}
+static uint8_t __attribute__((noinline)) pulp_nn_add_quant_u4 (
+  uint8_t pix1,
+  uint8_t pix2,
+  int16_t m1,
+  int16_t m2,
+  int8_t d
+  ) {
+  uint32_t integer_image = pix1*m1 + pix2*m2;
+  uint32_t x = (integer_image) >> d;
+  uint8_t res = clip4(x);
+  return res;
+}
+
+static int8_t __attribute__((noinline)) pulp_nn_quant_i4 (
+  int32_t phi,
+  int16_t m,
+  int8_t d
+  ) {
+  int32_t x = (m * phi) >> d;
+  int8_t res = clips4(x);
+  return res;
+}
+static int8_t __attribute__((noinline)) pulp_nn_bn_quant_i4 (
+  int32_t phi,
+  int32_t k,
+  int32_t lambda,
+  int8_t d
+  ) {
+  int32_t integer_image_phi = (k * phi) + lambda;
+  int32_t x = (integer_image_phi) >> d;
+  int8_t res = clips4(x);
+  return res;
+}
+static int8_t __attribute__((noinline)) pulp_nn_add_quant_i4 (
+  int8_t pix1,
+  int8_t pix2,
+  int16_t m1,
+  int16_t m2,
+  int8_t d
+  ) {
+  int32_t integer_image = pix1*m1 + pix2*m2;
+  int32_t x = (integer_image) >> d;
+  int8_t res = clips4(x);
+  return res;
+}
+
+static uint8_t __attribute__((noinline)) pulp_nn_quant_u8 (
+  int32_t phi,
+  int16_t m,
+  int8_t d
+  ) {
+  int32_t x = (m * phi) >> d;
+  uint8_t res = clip8(x);
+  return res;
+}
+static uint8_t __attribute__((noinline)) pulp_nn_bn_quant_u8 (
+  int32_t phi,
+  int32_t k,
+  int32_t lambda,
+  int8_t d
+  ) {
+  int32_t integer_image_phi = (k * phi) + lambda;
+  int32_t x = (integer_image_phi) >> d;
+  uint8_t res = clip8(x);
+  return res;
+}
+static uint8_t __attribute__((noinline)) pulp_nn_add_quant_u8 (
+  uint8_t pix1,
+  uint8_t pix2,
+  int16_t m1,
+  int16_t m2,
+  int8_t d
+  ) {
+  uint32_t integer_image = pix1*m1 + pix2*m2;
+  uint32_t x = (integer_image) >> d;
+  uint8_t res = clip8(x);
+  return res;
+}
+
+static int8_t __attribute__((noinline)) pulp_nn_quant_i8 (
+  int32_t phi,
+  int16_t m,
+  int8_t d
+  ) {
+  int32_t x = (m * phi) >> d;
+  int8_t res = clips8(x);
+  return res;
+}
+static int8_t __attribute__((noinline)) pulp_nn_bn_quant_i8 (
+  int32_t phi,
+  int32_t k,
+  int32_t lambda,
+  int8_t d
+  ) {
+  int32_t integer_image_phi = (k * phi) + lambda;
+  int32_t x = (integer_image_phi) >> d;
+  int8_t res = clips8(x);
+  return res;
+}
+static int8_t __attribute__((noinline)) pulp_nn_add_quant_i8 (
+  int8_t pix1,
+  int8_t pix2,
+  int16_t m1,
+  int16_t m2,
+  int8_t d
+  ) {
+  int32_t integer_image = pix1*m1 + pix2*m2;
+  int32_t x = (integer_image) >> d;
+  int8_t res = clips8(x);
+  return res;
+}
+
+
+static uint8_t __attribute__((noinline)) pulp_nn_u4_quant(int input, int16_t * pThr)
+{
+  if(input <= pThr[7] )
+  {
+    if( input <= pThr[3])
+    {
+      if( input <= pThr[1])
+      {
+        if( input <= pThr[0])
+          return 0;
+        else
+          return 1;
+      }
+      else
+      {
+        if( input <= pThr[2])
+          return 2;
+        else
+          return 3;
+      }
+    }
+    else
+    {
+      if( input <= pThr[5])
+      {
+        if( input <= pThr[4])
+          return 4;
+        else
+          return 5;
+      }
+      else
+      {
+        if( input <= pThr[6])
+          return 6;
+        else
+          return 7;
+      }
+    }
+  }
+  else
+  {
+    if( input <= pThr[11])
+    {
+      if( input <= pThr[9])
+      {
+        if( input <= pThr[8])
+          return 8;
+        else
+          return 9;
+      }
+      else
+      {
+        if( input <= pThr[10])
+          return 10;
+        else
+          return 11;
+      }
+    }
+    else
+    {
+      if( input <= pThr[13])
+      {
+        if( input <= pThr[12])
+          return 12;
+        else
+          return 13;
+      }
+      else
+      {
+        if( input <= pThr[14])
+          return 14;
+        else
+          return 15;
+      }
+    }
+  }
+}
+
+static uint8_t __attribute__((noinline)) pulp_nn_u2_quant(int input, int16_t * pThr)
+{
+  if( input <= pThr[1])
+  {
+    if( input <= pThr[0])
+        {
+      return 0;
+    }
+        else
+        {
+      return 1;
+    }
+  }
+    else
+    {
+    if( input <= pThr[2])
+        {
+      return 2;
+    }
+        else
+        {
+      return 3;
+    }
+  }
+}
+
+/*
+ * Common
+ */
+
+
+static v4s __attribute__((noinline)) pulp_nn_i4_to_i8_r( int8_t *pSrc)
+{
+  v4s Src = *((v4s*) pSrc);
+  int8_t bext1, bext2, bext3, bext4;
+
+  bext1 = (int8_t) bitext((int) Src, 4, 0);
+  bext2 = (int8_t) bitext((int) Src, 4, 4);
+  bext3 = (int8_t) bitext((int) Src, 4, 8);
+  bext4 = (int8_t) bitext((int) Src, 4, 12);
+  v4s res = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+
+  return res;
+}
+
+static v4u __attribute__((noinline)) pulp_nn_u4_to_u8_r(uint8_t *pSrc)
+{
+  v4u Src = *((v4u*) pSrc);
+  uint8_t bext1, bext2, bext3, bext4;
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 4, 0);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 4, 4);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 4, 8);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 4, 12);
+  v4u res = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+
+  return res;
+}
+
+static v4s __attribute__((noinline)) pulp_nn_i2_to_i8_r( int8_t *pSrc)
+{
+  v4s Src = *((v4s*) pSrc);
+  int8_t bext1, bext2, bext3, bext4;
+
+  bext1 = (int8_t) bitext((int) Src, 2, 0);
+  bext2 = (int8_t) bitext((int) Src, 2, 2);
+  bext3 = (int8_t) bitext((int) Src, 2, 4);
+  bext4 = (int8_t) bitext((int) Src, 2, 6);
+  v4s res = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+
+  return res;
+}
+
+static v4u __attribute__((noinline)) pulp_nn_u2_to_u8_r(uint8_t *pSrc)
+{
+  v4u Src = *((v4u*) pSrc);
+  uint8_t bext1, bext2, bext3, bext4;
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 0);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 2);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 4);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 6);
+  v4u res = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+
+  return res;
+}
+
+static v4s __attribute__((noinline)) pulp_nn_i2_to_i4_r(int8_t *pSrc)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  v4s Src = *((v4s*) pSrc);
+  int8_t bext1, bext2, bext3, bext4;
+  int8_t out1, out2, out3, out4;
+
+  bext1 = (int8_t) bitextu((int) Src, 2, 0);
+  bext2 = (int8_t) bitextu((int) Src, 2, 2);
+  bext3 = (int8_t) bitextu((int) Src, 2, 4);
+  bext4 = (int8_t) bitextu((int) Src, 2, 6);
+
+  out1 = bitins(bext1, n_mask, bext2, mask, off);
+  out2 = bitins(bext3, n_mask, bext4, mask, off);
+
+  bext1 = (int8_t) bitextu((int) Src, 2, 8);
+  bext2 = (int8_t) bitextu((int) Src, 2, 10);
+  bext3 = (int8_t) bitextu((int) Src, 2, 12);
+  bext4 = (int8_t) bitextu((int) Src, 2, 14);
+
+  out3 = bitins(bext1, n_mask, bext2, mask, off);
+  out4 = bitins(bext3, n_mask, bext4, mask, off);
+
+  v4s res = pack((int8_t) out1, (int8_t) out2, (int8_t) out3, (int8_t) out4);
+
+  return res;
+}
+
+static v4u __attribute__((noinline)) pulp_nn_u2_to_u4_r(uint8_t *pSrc)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  v4u Src = *((v4u*) pSrc);
+  uint8_t bext1, bext2, bext3, bext4;
+  uint8_t out1, out2, out3, out4;
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 0);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 2);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 4);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 6);
+
+  out1 = bitins(bext1, n_mask, bext2, mask, off);
+  out2 = bitins(bext3, n_mask, bext4, mask, off);
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 8);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 10);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 12);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 14);
+
+  out3 = bitins(bext1, n_mask, bext2, mask, off);
+  out4 = bitins(bext3, n_mask, bext4, mask, off);
+
+  v4u res = pack((uint8_t) out1, (uint8_t) out2, (uint8_t) out3, (uint8_t) out4);
+
+  return res;
+}
+
+static int8_t *__attribute__((always_inline)) pulp_nn_i4_to_i8( int8_t *pSrc, int8_t *pDst)
+{
+  v4s Src = *((v4s*) pSrc);
+  int8_t bext1, bext2, bext3, bext4;
+
+  pSrc+=4;
+
+  bext1 = (int8_t) bitext((int) Src, 4, 0);
+  bext2 = (int8_t) bitext((int) Src, 4, 4);
+  bext3 = (int8_t) bitext((int) Src, 4, 8);
+  bext4 = (int8_t) bitext((int) Src, 4, 12);
+  *((v4s*)pDst) = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+   MemoryFence();
+  pDst+=4;
+  bext1 = (int8_t) bitext((int) Src, 4, 16);
+  bext2 = (int8_t) bitext((int) Src, 4, 20);
+  bext3 = (int8_t) bitext((int) Src, 4, 24);
+  bext4 = (int8_t) bitext((int) Src, 4, 28);
+  *((v4s*)pDst) = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+
+  return pSrc;
+}
+
+static uint8_t *__attribute__((always_inline)) pulp_nn_u4_to_u8(uint8_t *pSrc, uint8_t *pDst)
+{
+  v4u Src = *((v4u*) pSrc);
+  uint8_t bext1, bext2, bext3, bext4;
+
+  pSrc+=4;
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 4, 0);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 4, 4);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 4, 8);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 4, 12);
+  *((v4u*)pDst) = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 4, 16);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 4, 20);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 4, 24);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 4, 28);
+  *((v4u*)pDst) = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+
+  return pSrc;
+}
+
+static int8_t *__attribute__((always_inline)) pulp_nn_i2_to_i8( int8_t * pSrc, int8_t * pDst)
+{
+  v4s Src = *((v4s*) pSrc);
+  int8_t bext1, bext2, bext3, bext4;
+
+  pSrc+=4;
+
+  bext1 = (int8_t) bitext((int) Src, 2, 0);
+  bext2 = (int8_t) bitext((int) Src, 2, 2);
+  bext3 = (int8_t) bitext((int) Src, 2, 4);
+  bext4 = (int8_t) bitext((int) Src, 2, 6);
+  *((v4s*)pDst) = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (int8_t) bitext((int) Src, 2, 8);
+  bext2 = (int8_t) bitext((int) Src, 2, 10);
+  bext3 = (int8_t) bitext((int) Src, 2, 12);
+  bext4 = (int8_t) bitext((int) Src, 2, 14);
+  *((v4s*)pDst) = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (int8_t) bitext((int) Src, 2, 16);
+  bext2 = (int8_t) bitext((int) Src, 2, 18);
+  bext3 = (int8_t) bitext((int) Src, 2, 20);
+  bext4 = (int8_t) bitext((int) Src, 2, 22);
+  *((v4s*)pDst) = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (int8_t) bitext((int) Src, 2, 24);
+  bext2 = (int8_t) bitext((int) Src, 2, 26);
+  bext3 = (int8_t) bitext((int) Src, 2, 28);
+  bext4 = (int8_t) bitext((int) Src, 2, 30);
+  *((v4s*)pDst) = pack((int8_t) bext1, (int8_t) bext2, (int8_t) bext3, (int8_t) bext4);
+
+  return pSrc;
+}
+
+static uint8_t *__attribute__((always_inline)) pulp_nn_u2_to_u8(uint8_t * pSrc, uint8_t * pDst)
+{
+  v4u Src = *((v4u*) pSrc);
+  uint8_t bext1, bext2, bext3, bext4;
+
+  pSrc+=4;
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 0);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 2);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 4);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 6);
+  *((v4u*)pDst) = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 8);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 10);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 12);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 14);
+  *((v4u*)pDst) = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 16);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 18);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 20);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 22);
+  *((v4u*)pDst) = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+  MemoryFence();
+  pDst+=4;
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 24);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 26);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 28);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 30);
+  *((v4u*)pDst) = pack((uint8_t) bext1, (uint8_t) bext2, (uint8_t) bext3, (uint8_t) bext4);
+
+  return pSrc;
+}
+
+static int8_t *__attribute__((always_inline)) pulp_nn_i2_to_i4( int8_t * pSrc, int8_t * pDst)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  v4s Src = *((v4s*) pSrc);
+  int8_t bext1, bext2, bext3, bext4;
+  int8_t out1, out2, out3, out4;
+
+  pSrc+=4;
+
+  bext1 = (int8_t) bitext((int) Src, 2, 0);
+  bext2 = (int8_t) bitext((int) Src, 2, 2);
+  bext3 = (int8_t) bitext((int) Src, 2, 4);
+  bext4 = (int8_t) bitext((int) Src, 2, 6);
+
+  out1 = bitins(bext1, n_mask, bext2, mask, off);
+  out2 = bitins(bext3, n_mask, bext4, mask, off);
+  MemoryFence();
+
+  bext1 = (int8_t) bitext((int) Src, 2, 8);
+  bext2 = (int8_t) bitext((int) Src, 2, 10);
+  bext3 = (int8_t) bitext((int) Src, 2, 12);
+  bext4 = (int8_t) bitext((int) Src, 2, 14);
+
+  out3 = bitins(bext1, n_mask, bext2, mask, off);
+  out4 = bitins(bext3, n_mask, bext4, mask, off);
+
+  *((v4s*)pDst) = pack((int8_t) out1, (int8_t) out2, (int8_t) out3, (int8_t) out4);
+  MemoryFence();
+
+  pDst+=4;
+  bext1 = (int8_t) bitext((int) Src, 2, 16);
+  bext2 = (int8_t) bitext((int) Src, 2, 18);
+  bext3 = (int8_t) bitext((int) Src, 2, 20);
+  bext4 = (int8_t) bitext((int) Src, 2, 22);
+
+  out1 = bitins(bext1, n_mask, bext2, mask, off);
+  out2 = bitins(bext3, n_mask, bext4, mask, off);
+  MemoryFence();
+
+  bext1 = (int8_t) bitext((int) Src, 2, 24);
+  bext2 = (int8_t) bitext((int) Src, 2, 26);
+  bext3 = (int8_t) bitext((int) Src, 2, 28);
+  bext4 = (int8_t) bitext((int) Src, 2, 30);
+
+  out3 = bitins(bext1, n_mask, bext2, mask, off);
+  out4 = bitins(bext3, n_mask, bext4, mask, off);
+
+  *((v4s*)pDst) = pack((int8_t) out1, (int8_t) out2, (int8_t) out3, (int8_t) out4);
+
+  return pSrc;
+}
+
+static uint8_t *__attribute__((always_inline)) pulp_nn_u2_to_u4( uint8_t * pSrc, uint8_t * pDst)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  v4u Src = *((v4u*) pSrc);
+  uint8_t bext1, bext2, bext3, bext4;
+  uint8_t out1, out2, out3, out4;
+
+  pSrc+=4;
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 0);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 2);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 4);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 6);
+
+  out1 = bitins(bext1, n_mask, bext2, mask, off);
+  out2 = bitins(bext3, n_mask, bext4, mask, off);
+  MemoryFence();
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 8);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 10);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 12);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 14);
+
+  out3 = bitins(bext1, n_mask, bext2, mask, off);
+  out4 = bitins(bext3, n_mask, bext4, mask, off);
+
+  *((v4u*)pDst) = pack((uint8_t) out1, (uint8_t) out2, (uint8_t) out3, (uint8_t) out4);
+  MemoryFence();
+
+  pDst+=4;
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 16);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 18);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 20);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 22);
+
+  out1 = bitins(bext1, n_mask, bext2, mask, off);
+  out2 = bitins(bext3, n_mask, bext4, mask, off);
+  MemoryFence();
+
+  bext1 = (uint8_t) bitextu((unsigned int) Src, 2, 24);
+  bext2 = (uint8_t) bitextu((unsigned int) Src, 2, 26);
+  bext3 = (uint8_t) bitextu((unsigned int) Src, 2, 28);
+  bext4 = (uint8_t) bitextu((unsigned int) Src, 2, 30);
+
+  out3 = bitins(bext1, n_mask, bext2, mask, off);
+  out4 = bitins(bext3, n_mask, bext4, mask, off);
+  
+  *((v4u*)pDst) = pack((uint8_t) out1, (uint8_t) out2, (uint8_t) out3, (uint8_t) out4);
+
+  return pSrc;
+}
+
+/*
+ * XpulpV2
+ */
+
+static void __attribute__((noinline)) pulp_zero_mem(uint8_t * pBuffer, unsigned int size)
+{
+  int lfover = size &0x3;
+  for (int i=0; i<(size>>2); i++)
+  {
+    *((v4u *)pBuffer) = (v4u){0,0,0,0};
+    MemoryFence();
+    pBuffer+=4;
+  }
+  while(lfover)
+  {
+    *pBuffer++=0;
+    lfover--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_im2col_u2_to_u8(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_u2_to_u8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 16;
+  }
+  while (lfover)
+  {
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 0);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 2);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 4);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 6);
+    pOutput++;
+    pInput++;
+    lfover -= 4;
+  }
+}
+static void __attribute__((noinline)) pulp_nn_im2col_i2_to_i8(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_i2_to_i8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 16;
+  }
+  while (lfover)
+  {
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 0);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 2);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 4);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 6);
+    pOutput++;
+    pInput++;
+    lfover -= 4;
+  }
+}
+static void __attribute__((noinline)) pulp_nn_im2col_u4_to_u8(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 3u;
+  int lfover = blockSize & 0x07;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_u4_to_u8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 8;
+  }
+  while (lfover)
+  {
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 4, 0);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 4, 4);
+    pOutput++;
+    pInput++;
+    lfover -= 2;
+  }
+}
+static void __attribute__((noinline)) pulp_nn_im2col_i4_to_i8(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 3u;
+  int lfover = blockSize & 0x07;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_i4_to_i8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 8;
+  }
+  while (lfover)
+  {
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 4, 0);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 4, 4);
+    pOutput++;
+    pInput++;
+    lfover -= 2;
+  }
+}
+static void __attribute__((noinline)) pulp_nn_im2col_u8_to_u8(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 2u;
+  int lfover = blockSize & 0x03;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4u*)pOutput) = *((v4u*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    *((uint8_t*)pOutput) = *((uint8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover--;
+  }
+}
+static void __attribute__((noinline)) pulp_nn_im2col_i8_to_i8(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 2u;
+  int lfover = blockSize & 0x03;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4s*)pOutput) = *((v4s*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    *((int8_t*)pOutput) = *((int8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover--;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_im2col_u2_to_u2(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4u*)pOutput) = *((v4u*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    uint8_t extr;
+    *((uint8_t*)pOutput) = *((uint8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover -= 4;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_i2_to_i2(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4s*)pOutput) = *((v4s*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    int8_t extr;
+    *((int8_t*)pOutput) = *((int8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover -= 4;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_u2_to_u4(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_u2_to_u4(pInput, pOutput);
+    MemoryFence();
+    pOutput += 8;
+  }
+  while (lfover)
+  {
+    uint8_t extr;
+    // TODO: this is potentially dangerous/wrong if the number of channels
+    // is not a multiple of 8!
+    *((v4u *)pOutput) = pulp_nn_u2_to_u4_r(pInput);
+    pInput += 2;
+    pOutput += 4;
+    lfover -= 8;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_i2_to_i4(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_i2_to_i4(pInput, pOutput);
+    MemoryFence();
+    pOutput += 8;
+  }
+  while (lfover)
+  {
+    int8_t extr;
+    // TODO: this is potentially dangerous/wrong if the number of channels
+    // is not a multiple of 8!
+    *((v4s *)pOutput) = pulp_nn_i2_to_i4_r(pInput);
+    pInput += 2;
+    pOutput += 4;
+    lfover -= 8;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_u2_to_u8(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_u2_to_u8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 16;
+  }
+  while (lfover)
+  {
+    uint8_t extr;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 0);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 2);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 4);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 2, 6);
+    pOutput++;
+    pInput++;
+    lfover -= 4;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_i2_to_i8(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 4u;
+  int lfover = blockSize & 0x0f;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_i2_to_i8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 16;
+  }
+  while (lfover)
+  {
+    int8_t extr;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 0);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 2);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 4);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 2, 6);
+    pOutput++;
+    pInput++;
+    lfover -= 4;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_u4_to_u4(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 3u;
+  int lfover = blockSize & 0x07;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4u*)pOutput) = *((v4u*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    uint8_t extr;
+    *((uint8_t*)pOutput) = *((uint8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover -= 2;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_i4_to_i4(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 3u;
+  int lfover = blockSize & 0x07;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4s*)pOutput) = *((v4s*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    int8_t extr;
+    *((int8_t*)pOutput) = *((int8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover -= 2;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_u4_to_u8(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 3u;
+  int lfover = blockSize & 0x07;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_u4_to_u8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 8;
+  }
+  while (lfover)
+  {
+    uint8_t extr;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 4, 0);
+    pOutput++;
+    *((uint8_t*)pOutput) = (uint8_t) bitextu((unsigned int) *pInput, 4, 4);
+    pOutput++;
+    pInput++;
+    lfover -= 2;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_i4_to_i8(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 3u;
+  int lfover = blockSize & 0x07;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    pInput = pulp_nn_i4_to_i8(pInput, pOutput);
+    MemoryFence();
+    pOutput += 8;
+  }
+  while (lfover)
+  {
+    int8_t extr;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 4, 0);
+    pOutput++;
+    *((int8_t*)pOutput) = (int8_t) bitext((int) *pInput, 4, 4);
+    pOutput++;
+    pInput++;
+    lfover -= 2;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_u8_to_u8(uint8_t * pInput, uint8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 2u;
+  int lfover = blockSize & 0x03;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4u*)pOutput) = *((v4u*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    uint8_t extr;
+    *((uint8_t*)pOutput) = *((uint8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover -= 1;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_im2col_i8_to_i8(int8_t * pInput, int8_t * pOutput, unsigned int blockSize)
+{
+  unsigned int blkCnt = blockSize >> 2u;
+  int lfover = blockSize & 0x03;
+  for (int i = 0; i<blkCnt; i++)
+  {
+    *((v4s*)pOutput) = *((v4s*)pInput);
+    pInput+=4;
+    pOutput+=4;
+  }
+  while (lfover)
+  {
+    int8_t extr;
+    *((int8_t*)pOutput) = *((int8_t*)pInput);
+    pOutput++;
+    pInput++;
+    lfover -= 1;
+  }
+}
+
+
+
+static void __attribute__((noinline)) pulp_nn_compare_and_replace_if_larger_u8(uint8_t * base,
+                                                uint8_t * target,
+                                                uint16_t length)
+{
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+  v4u inp;
+  v4u com;
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    inp = *((v4u*)pIn);
+    com = *((v4u*)pCom);
+
+    *((v4u*)pIn) = max4(inp, com);
+
+    pCom+=4;
+    pIn+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    if(*pIn<*pCom)
+      *pIn=*pCom;
+    
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_compare_and_replace_if_larger_i8(
+    int8_t * base, int8_t * target, uint16_t length) {
+  int8_t *pIn = base;
+  int8_t *pCom = target;
+  v4s inp;
+  v4s com;
+  int cnt = length >> 2;
+
+  while (cnt > 0u) {
+    inp = *((v4s *)pIn);
+    com = *((v4s *)pCom);
+
+    *((v4s *)pIn) = maxs4(inp, com);
+
+    pCom += 4;
+    pIn += 4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left > 0u) {
+    if (*pIn < *pCom)
+      *pIn = *pCom;
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_avg_and_replace_u8(uint8_t * base,
+                                  uint8_t * target,
+                                  uint16_t length)
+{
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+
+  while (length>0u)
+  {
+    *pIn = ((*pIn + *pCom) >> 1);
+
+    pIn++;
+    pCom++;
+    length--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_compare_and_replace_if_larger_u4(uint8_t * base,
+                                                uint8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+  v4u inp[2];
+  v4u com[2];
+  uint8_t *out;
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    pulp_nn_u4_to_u8(pIn, (uint8_t *)inp);
+    pulp_nn_u4_to_u8(pCom, (uint8_t *)com);
+
+    *((v4u *)out) = max4(inp[0], com[0]);
+
+    *((uint8_t*)pIn) = bitins(*out, n_mask, *(out + 1), mask, off);
+    pIn++;
+    *((uint8_t*)pIn) = bitins(*(out + 2), n_mask, *(out + 3), mask, off);
+    pIn++;
+
+    *((v4u *)out) = max4(inp[1], com[1]);
+
+    *((uint8_t*)pIn) = bitins(*out, n_mask, *(out + 1), mask, off);
+    pIn++;
+    *((uint8_t*)pIn) = bitins(*(out + 2), n_mask, *(out + 3), mask, off);
+    pIn++;
+
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 4, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 4, 4);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 4, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 4, 4);
+
+    if(inA0<inB0)
+      inA0=inB0;
+
+    if(inA1<inB1)
+      inA1=inB1;
+
+    *((uint8_t*)pIn) = bitins(inA0, n_mask, inA1, mask, off);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_compare_and_replace_if_larger_i4(int8_t * base,
+                                                int8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  int8_t *pIn = base;
+  int8_t *pCom = target;
+  v4s inp[2];
+  v4s com[2];
+  int8_t *out;
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    pulp_nn_i4_to_i8(pIn, (int8_t *)inp);
+    pulp_nn_i4_to_i8(pCom, (int8_t *)com);
+
+    *((v4s *)out) = maxs4(inp[0], com[0]);
+
+    *((int8_t*)pIn) = bitins(*out, n_mask, *(out + 1), mask, off);
+    pIn++;
+    *((int8_t*)pIn) = bitins(*(out + 2), n_mask, *(out + 3), mask, off);
+    pIn++;
+
+    *((v4s *)out) = maxs4(inp[1], com[1]);
+
+    *((int8_t*)pIn) = bitins(*out, n_mask, *(out + 1), mask, off);
+    pIn++;
+    *((int8_t*)pIn) = bitins(*(out + 2), n_mask, *(out + 3), mask, off);
+    pIn++;
+
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    int8_t inA0 = (int8_t) bitext((int) *pIn, 4, 0);
+    int8_t inA1 = (int8_t) bitext((int) *pIn, 4, 4);
+    int8_t inB0 = (int8_t) bitext((int) *pCom, 4, 0);
+    int8_t inB1 = (int8_t) bitext((int) *pCom, 4, 4);
+
+    if(inA0<inB0)
+      inA0=inB0;
+
+    if(inA1<inB1)
+      inA1=inB1;
+
+    *((int8_t*)pIn) = bitins(inA0, n_mask, inA1, mask, off);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_avg_and_replace_u4(uint8_t * base,
+                                  uint8_t * target,
+                                  uint16_t length)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+
+  while (length>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 4, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 4, 4);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 4, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 4, 4);
+
+    inA0 = ((inA0 + inB0) >> 1);
+    inA1 = ((inA1 + inB1) >> 1);
+
+    *((uint8_t*)pIn) = bitins(inA0, n_mask, inA1, mask, off);
+
+    pIn++;
+    pCom++;
+    length--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_compare_and_replace_if_larger_u2(uint8_t * base,
+                                                uint8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask2 = 0x0c;
+  int8_t n_mask2 = ~ mask2;
+  int8_t mask4 = 0x30;
+  int8_t n_mask4 = ~ mask4;
+  int8_t mask6 = 0xc0;
+  int8_t n_mask6 = ~ mask6;
+  int8_t off2 = 2;
+  int8_t off4 = 4;
+  int8_t off6 = 6;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+  v4u inp[4];
+  v4u com[4];
+  uint8_t *out;
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    pulp_nn_u2_to_u8(pIn, inp);
+    pulp_nn_u2_to_u8(pCom, com);
+
+    *((v4u*)out) = max4(inp[0], com[0]);
+
+    uint8_t inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    *((v4u*)out) = max4(inp[1], com[1]);
+
+    inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    *((v4u*)out) = max4(inp[2], com[2]);
+
+    inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    *((v4u*)out) = max4(inp[3], com[3]);
+
+    inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 2, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 2, 2);
+    uint8_t inA2 = (uint8_t) bitextu((unsigned int) *pIn, 2, 4);
+    uint8_t inA3 = (uint8_t) bitextu((unsigned int) *pIn, 2, 6);
+    v4u inA4 = pack((uint8_t) inA0, (uint8_t) inA1, (uint8_t) inA2, (uint8_t) inA3);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 2, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 2, 2);
+    uint8_t inB2 = (uint8_t) bitextu((unsigned int) *pCom, 2, 4);
+    uint8_t inB3 = (uint8_t) bitextu((unsigned int) *pCom, 2, 6);
+    v4u inB4 = pack((uint8_t) inB0, (uint8_t) inB1, (uint8_t) inB2, (uint8_t) inB3);
+
+    *((v4u*)out) = max4(inA4, inB4);
+
+    uint8_t inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+
+static void __attribute__((noinline)) pulp_nn_compare_and_replace_if_larger_i2(int8_t * base,
+                                                int8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask2 = 0x0c;
+  int8_t n_mask2 = ~ mask2;
+  int8_t mask4 = 0x30;
+  int8_t n_mask4 = ~ mask4;
+  int8_t mask6 = 0xc0;
+  int8_t n_mask6 = ~ mask6;
+  int8_t off2 = 2;
+  int8_t off4 = 4;
+  int8_t off6 = 6;
+
+  int8_t *pIn = base;
+  int8_t *pCom = target;
+  v4s inp[4];
+  v4s com[4];
+  int8_t *out;
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    pulp_nn_i2_to_i8(pIn, inp);
+    pulp_nn_i2_to_i8(pCom, com);
+
+    *((v4s*)out) = maxs4(inp[0], com[0]);
+
+    int8_t inA = (int8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((int8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    *((v4s*)out) = maxs4(inp[1], com[1]);
+
+    inA = (int8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((int8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    *((v4s*)out) = maxs4(inp[2], com[2]);
+
+    inA = (int8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((int8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    *((v4s*)out) = maxs4(inp[3], com[3]);
+
+    inA = (int8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((int8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+    pIn++;
+
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    int8_t inA0 = (int8_t) bitext((unsigned int) *pIn, 2, 0);
+    int8_t inA1 = (int8_t) bitext((unsigned int) *pIn, 2, 2);
+    int8_t inA2 = (int8_t) bitext((unsigned int) *pIn, 2, 4);
+    int8_t inA3 = (int8_t) bitext((unsigned int) *pIn, 2, 6);
+    v4s inA4 = pack((int8_t) inA0, (int8_t) inA1, (int8_t) inA2, (int8_t) inA3);
+    int8_t inB0 = (int8_t) bitext((unsigned int) *pCom, 2, 0);
+    int8_t inB1 = (int8_t) bitext((unsigned int) *pCom, 2, 2);
+    int8_t inB2 = (int8_t) bitext((unsigned int) *pCom, 2, 4);
+    int8_t inB3 = (int8_t) bitext((unsigned int) *pCom, 2, 6);
+    v4s inB4 = pack((int8_t) inB0, (int8_t) inB1, (int8_t) inB2, (int8_t) inB3);
+
+    *((v4s*)out) = maxs4(inA4, inB4);
+
+    int8_t inA = (int8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((int8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) pulp_nn_avg_and_replace_u2(uint8_t * base,
+                                  uint8_t * target,
+                                  uint16_t length)
+{
+  int8_t mask2 = 0x0c;
+  int8_t n_mask2 = ~ mask2;
+  int8_t mask4 = 0x30;
+  int8_t n_mask4 = ~ mask4;
+  int8_t mask6 = 0xc0;
+  int8_t n_mask6 = ~ mask6;
+  int8_t off2 = 2;
+  int8_t off4 = 4;
+  int8_t off6 = 6;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+
+  while (length>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 2, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 2, 2);
+    uint8_t inA2 = (uint8_t) bitextu((unsigned int) *pIn, 2, 4);
+    uint8_t inA3 = (uint8_t) bitextu((unsigned int) *pIn, 2, 6);
+
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 2, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 2, 2);
+    uint8_t inB2 = (uint8_t) bitextu((unsigned int) *pCom, 2, 4);
+    uint8_t inB3 = (uint8_t) bitextu((unsigned int) *pCom, 2, 6);
+
+    inA0 = ((inA0 + inB0) >> 1);
+    inA1 = ((inA1 + inB1) >> 1);
+    inA2 = ((inA2 + inB2) >> 1);
+    inA3 = ((inA3 + inB3) >> 1);
+
+    uint8_t inA = (uint8_t) bitins(inA0, n_mask2, inA1, mask2, off2);
+    inA = bitins(inA, n_mask4, inA2, mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, inA3, mask6, off6);
+
+    pIn++;
+    pCom++;
+    length--;
+  }
+}
+
+/*
+ * XpulpNN
+ */
+
+static void __attribute__((noinline)) xpulp_nn_zero_mem_u8(uint8_t * pBuffer, unsigned int size)
+{
+  int lfover = size &0x3;
+  for (int i=0; i<(size>>2); i++)
+  {
+    *((v4u *)pBuffer) = (v4u){0,0,0,0};
+    MemoryFence();
+    pBuffer+=4;
+  }
+  while(lfover)
+  {
+    *pBuffer++=0;
+    lfover--;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_zero_mem_u4(uint8_t * pBuffer, unsigned int size)
+{
+  int lfover = size &0x7;
+  for (int i=0; i<(size>>3); i++)
+  {
+    *((v4u *)pBuffer) = (v4u){0,0,0,0};
+    MemoryFence();
+    pBuffer+=4;
+  }
+  while(lfover)
+  {
+    *pBuffer++=0;
+    lfover-=2;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_zero_mem_u2(uint8_t * pBuffer, unsigned int size)
+{
+  int lfover = size &0xf;
+  for (int i=0; i<(size>>4); i++)
+  {
+    *((v4u *)pBuffer) = (v4u){0,0,0,0};
+    MemoryFence();
+    pBuffer+=4;
+  }
+  while(lfover)
+  {
+    *pBuffer++=0;
+    lfover-=4;
+  }
+}
+
+
+static void __attribute__((noinline)) xpulp_nn_compare_and_replace_if_larger_u4(uint8_t * base,
+                                                uint8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    *((uint32_t *)pIn) = max8(*((uint32_t *)pIn), *((uint32_t *)pCom));
+
+    pIn+=4;
+    pCom+=4;
+
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 4, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 4, 4);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 4, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 4, 4);
+
+    if(inA0<inB0)
+      inA0=inB0;
+
+    if(inA1<inB1)
+      inA1=inB1;
+
+    *((uint8_t*)pIn) = bitins(inA0, n_mask, inA1, mask, off);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_compare_and_replace_if_larger_i4(
+    int8_t * base, int8_t * target, uint16_t length) {
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~mask;
+  int8_t off = 0x04;
+
+  int8_t *pIn = base;
+  int8_t *pCom = target;
+
+  int cnt = length >> 2;
+
+  while (cnt > 0u) {
+    *((int32_t *)pIn) = maxs8(*((int32_t *)pIn), *((int32_t *)pCom));
+
+    pIn += 4;
+    pCom += 4;
+
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left > 0u) {
+    int8_t inA0 = (int8_t)bitext((int)*pIn, 4, 0);
+    int8_t inA1 = (int8_t)bitext((int)*pIn, 4, 4);
+    int8_t inB0 = (int8_t)bitext((int)*pCom, 4, 0);
+    int8_t inB1 = (int8_t)bitext((int)*pCom, 4, 4);
+
+    if (inA0 < inB0)
+      inA0 = inB0;
+
+    if (inA1 < inB1)
+      inA1 = inB1;
+
+    *((int8_t *)pIn) = bitins(inA0, n_mask, inA1, mask, off);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_avg_and_replace_u4(uint8_t * base,
+                                  uint8_t * target,
+                                  uint16_t length)
+{
+  int8_t mask = 0xf0;
+  int8_t n_mask = ~ mask;
+  int8_t off = 0x04;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+
+  int cnt = length >> 2;
+
+  while (cnt > 0u)
+  {
+    *((uint32_t *)pIn) = avg8(*((uint32_t *)pIn), *((uint32_t *)pCom));
+
+    pIn+=4;
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+
+  while (left>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 4, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 4, 4);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 4, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 4, 4);
+
+    inA0 = ((inA0 + inB0) >> 1);
+    inA1 = ((inA1 + inB1) >> 1);
+
+    *((uint8_t*)pIn) = bitins(inA0, n_mask, inA1, mask, off);
+
+    pIn++;
+    pCom++;
+    length--;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_compare_and_replace_if_larger_u2(uint8_t * base,
+                                                uint8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask2 = 0x0c;
+  int8_t n_mask2 = ~ mask2;
+  int8_t mask4 = 0x30;
+  int8_t n_mask4 = ~ mask4;
+  int8_t mask6 = 0xc0;
+  int8_t n_mask6 = ~ mask6;
+  int8_t off2 = 2;
+  int8_t off4 = 4;
+  int8_t off6 = 6;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+  uint8_t *out;
+
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    *((uint32_t *)pIn) = max16(*((uint32_t *)pIn), *((uint32_t *)pCom));
+
+    pIn+=4;
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 2, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 2, 2);
+    uint8_t inA2 = (uint8_t) bitextu((unsigned int) *pIn, 2, 4);
+    uint8_t inA3 = (uint8_t) bitextu((unsigned int) *pIn, 2, 6);
+    v4u inA4 = pack((uint8_t) inA0, (uint8_t) inA1, (uint8_t) inA2, (uint8_t) inA3);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 2, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 2, 2);
+    uint8_t inB2 = (uint8_t) bitextu((unsigned int) *pCom, 2, 4);
+    uint8_t inB3 = (uint8_t) bitextu((unsigned int) *pCom, 2, 6);
+    v4u inB4 = pack((uint8_t) inB0, (uint8_t) inB1, (uint8_t) inB2, (uint8_t) inB3);
+
+    *((v4u*)out) = max4(inA4, inB4);
+
+    uint8_t inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+static void __attribute__((noinline)) xpulp_nn_compare_and_replace_if_larger_i2(int8_t * base,
+                                                int8_t * target,
+                                                uint16_t length)
+{
+  int8_t mask2 = 0x0c;
+  int8_t n_mask2 = ~ mask2;
+  int8_t mask4 = 0x30;
+  int8_t n_mask4 = ~ mask4;
+  int8_t mask6 = 0xc0;
+  int8_t n_mask6 = ~ mask6;
+  int8_t off2 = 2;
+  int8_t off4 = 4;
+  int8_t off6 = 6;
+
+  int8_t *pIn = base;
+  int8_t *pCom = target;
+  int8_t *out;
+
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    *((int32_t *)pIn) = maxs16(*((int32_t *)pIn), *((int32_t *)pCom));
+
+    pIn+=4;
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    int8_t inA0 = (int8_t) bitext((int) *pIn, 2, 0);
+    int8_t inA1 = (int8_t) bitext((int) *pIn, 2, 2);
+    int8_t inA2 = (int8_t) bitext((int) *pIn, 2, 4);
+    int8_t inA3 = (int8_t) bitext((int) *pIn, 2, 6);
+    v4s inA4 = pack((int8_t) inA0, (int8_t) inA1, (int8_t) inA2, (int8_t) inA3);
+    int8_t inB0 = (int8_t) bitext((int) *pCom, 2, 0);
+    int8_t inB1 = (int8_t) bitext((int) *pCom, 2, 2);
+    int8_t inB2 = (int8_t) bitext((int) *pCom, 2, 4);
+    int8_t inB3 = (int8_t) bitext((int) *pCom, 2, 6);
+    v4s inB4 = pack((int8_t) inB0, (int8_t) inB1, (int8_t) inB2, (int8_t) inB3);
+
+    *((v4s*)out) = maxs4(inA4, inB4);
+
+    int8_t inA = (int8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((int8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+static void __attribute__((noinline)) xpulp_nn_avg_and_replace_u2(uint8_t * base,
+                                  uint8_t * target,
+                                  uint16_t length)
+{
+  int8_t mask2 = 0x0c;
+  int8_t n_mask2 = ~ mask2;
+  int8_t mask4 = 0x30;
+  int8_t n_mask4 = ~ mask4;
+  int8_t mask6 = 0xc0;
+  int8_t n_mask6 = ~ mask6;
+  int8_t off2 = 2;
+  int8_t off4 = 4;
+  int8_t off6 = 6;
+
+  uint8_t *pIn = base;
+  uint8_t *pCom = target;
+  uint8_t *out;
+
+  int cnt = length >> 2;
+
+  while(cnt > 0u)
+  {
+    *((uint32_t *)pIn) = avg16(*((uint32_t *)pIn), *((uint32_t *)pCom));
+
+    pIn+=4;
+    pCom+=4;
+    cnt--;
+  }
+
+  int left = length & 0x3;
+  while (left>0u)
+  {
+    uint8_t inA0 = (uint8_t) bitextu((unsigned int) *pIn, 2, 0);
+    uint8_t inA1 = (uint8_t) bitextu((unsigned int) *pIn, 2, 2);
+    uint8_t inA2 = (uint8_t) bitextu((unsigned int) *pIn, 2, 4);
+    uint8_t inA3 = (uint8_t) bitextu((unsigned int) *pIn, 2, 6);
+    v4u inA4 = pack((uint8_t) inA0, (uint8_t) inA1, (uint8_t) inA2, (uint8_t) inA3);
+    uint8_t inB0 = (uint8_t) bitextu((unsigned int) *pCom, 2, 0);
+    uint8_t inB1 = (uint8_t) bitextu((unsigned int) *pCom, 2, 2);
+    uint8_t inB2 = (uint8_t) bitextu((unsigned int) *pCom, 2, 4);
+    uint8_t inB3 = (uint8_t) bitextu((unsigned int) *pCom, 2, 6);
+    v4u inB4 = pack((uint8_t) inB0, (uint8_t) inB1, (uint8_t) inB2, (uint8_t) inB3);
+
+    *((v4u*)out) = avg4(inA4, inB4);
+
+    uint8_t inA = (uint8_t) bitins(*out, n_mask2, *(out + 1), mask2, off2);
+    inA = bitins(inA, n_mask4, *(out + 2), mask4, off4);
+    *((uint8_t*)pIn) = bitins(inA, n_mask6, *(out + 3), mask6, off6);
+
+    pIn++;
+    pCom++;
+    left--;
+  }
+}
+
+
+#endif
diff --git a/rt_nn_tests/xptnn_maxpool/test.c b/rt_nn_tests/xptnn_maxpool/test.c
new file mode 100644
index 0000000..d9186c0
--- /dev/null
+++ b/rt_nn_tests/xptnn_maxpool/test.c
@@ -0,0 +1,171 @@
+#include <stdio.h>
+#include <stdint.h>
+#include <stdlib.h>
+
+
+#include "xpulp_tnn_matmul_ternary.h"
+#include "xpulp_tnn_matmul_ternary_4x1.h"
+#include "xpulp_tnn_conv_ternary.h"
+#include "xpulp_tnn_conv_ternary_signed.h"
+#ifndef PROFILE // HACKYTIME
+#include "xpulp_tnn_conv1d_ternary.h"
+#include "xpulp_tnn_conv1d_ternary_signed.h"
+#endif
+#include "xpulp_tnn_maxpool_ternary.h"
+#include "xpulp_tnn_linear_ternary_i32_signed.h"
+#include "xpulp_tnn_linear_ternary_i32.h"
+#include "pmsis.h"
+
+#include "data_statstest.h"
+//#include "pulp_nn_kernels.h"
+#include "pulp_nn_mix_kernels.h"
+
+
+#define start_cycle_counter()                                   asm volatile("csrw 0xCC0, 0x01;")
+#define stop_cycle_counter()                                    asm volatile("csrw 0xCC0, 0x00;")
+#define read_cycle_counter(x)                                   asm volatile("csrr %0, 0x780;" : "=r" (x))
+#define reset_cycle_counter()                                   asm volatile("csrw 0x780, 0x0;")
+
+uint8_t im2col[IM2COL_DIM] = {0};
+uint8_t outputs[OUTPUT_DIM] = {0};
+
+int32_t outputs_fp[OUTPUT_DIM_FP] = {0};
+
+#ifndef PROFILE
+int num_cycles;
+#else
+int im2col_cycles;
+int hotloop_prep_cycles;
+int hotloop_cycles;
+int threshold_cycles;
+int requant_cycles;
+int hotloop_leftover_cycles;
+int matmul4x2_leftover_cycles;
+#endif
+
+void call_krnl_0();
+void test_0();
+
+int main(int argc, char *argv[])
+{
+#if KRAKEN_PTEST == 1
+  kraken_padframe_aon_pad_gpioa_cfg_rxe_set(24, 0);
+  kraken_padframe_aon_pad_gpioa_cfg_trie_set(24, 0);
+  kraken_padframe_aon_pad_gpioa_cfg_chip2pad_set(24, 0);
+#endif
+  int n_mismatches = 0;
+  int mismatches_tot = 0;
+
+    
+
+  #ifndef PROFILE
+  num_cycles = 0;
+  #endif
+
+  if (get_core_id() == 0) {
+  printf("===> TEST 0: Running xpulp_tnn_maxpool_ternary...\n");
+  printf("  dims_in     = [32, 32]\n");
+  printf("  dims_kernel = [2, 2]\n");
+  printf("  ch_in/out   = [40, 40]\n");
+  //printf("  padding_y_top    = [%d]\n", padding_y_top);
+  //printf("  padding_y_bottom = [%d]\n", padding_y_bottom);
+  //printf("  padding_x_left   = [%d]\n", padding_x_left);
+  //printf("  padding_x_right  = [%d]\n", padding_x_right);
+  //printf("  stride_x         = [%d]\n", stride_x);
+  //printf("  stride_y         = [%d]\n", stride_y);
+  }
+  test_0();
+  #ifndef PROFILE
+  stop_cycle_counter();
+  read_cycle_counter(num_cycles);
+  if (get_core_id() == 0) {
+  printf("===> TEST 0: Finished running xpulp_tnn_maxpool_ternary\n");
+  printf("num_cycles = %d\n", num_cycles);
+  printf("MACs = 1638400\n");
+  printf("MACs/cycle = %.4f\n", 1638400/num_cycles);
+  }
+  #endif
+  if (get_core_id() == 0) {
+  printf("Checking for mismatches..\n");
+  n_mismatches = 0;
+
+  for(int i=0; i < 2048; i++) {
+    if (outputs[i] != exp_outp_0[i]){
+        printf("***Mismatch in test 0 at iteration %d: Expected: %x, got: %x\n", i, exp_outp_0[i], outputs[i]);
+        n_mismatches++;
+    }
+  }
+  }
+  mismatches_tot += n_mismatches;
+
+  if (get_core_id() == 0) {
+  printf("Got %d mismatches in %d tests\n", mismatches_tot, 1);
+  }
+  return mismatches_tot;
+}
+
+
+void call_krnl_0(void) {
+  uint8_t * pInp;
+  uint8_t * pIm2ColBuffer;
+  int8_t * pBias = NULL;
+  uint8_t * pOut;
+  int8_t * pWeight;
+  uint32_t * pThr;
+  int32_t * pKappa, pLambda;
+  pInp = inp_l1;
+  
+  pOut = outp_l1;
+#if KRAKEN_PTEST == 1
+  if (pi_core_id() == 0) {
+    kraken_padframe_aon_pad_gpioa_cfg_chip2pad_set(24, 1);
+  } 
+#endif
+  xpulp_tnn_maxpool_ternary(
+    pInp,
+    pOut,
+    32,
+    32,
+    40,
+    16,
+    16,
+    2,
+    2,
+    0,
+    0,
+    0,
+    0,
+    2,
+    2);
+
+#if KRAKEN_PTEST == 1
+  if (pi_core_id() == 0) {
+    kraken_padframe_aon_pad_gpioa_cfg_chip2pad_set(24, 0);
+  }
+#endif
+}
+
+void test_0(void) {
+  // DMA transfer inputs from L2 to L1
+  if (pi_core_id() == 0) {
+     plp_dma_memcpy(pIn_0, inp_l1, 8192, 1);
+     plp_dma_barrier();
+  }
+  pi_cl_team_barrier(0);
+  call_krnl_0();
+  // get outputs back with DMA
+  if (pi_core_id() == 0) {
+     plp_dma_memcpy(outputs, outp_l1, 2048, 0);
+     plp_dma_barrier();
+  }
+}
+
+
+
+
+
+
+
+
+
+
diff --git a/rt_nn_tests/xptnn_maxpool/xpulp_tnn_conv1d_ternary.h b/rt_nn_tests/xptnn_maxpool/xpulp_tnn_conv1d_ternary.h
new file mode 100644
index 0000000..f1a08a6
--- /dev/null
+++ b/rt_nn_tests/xptnn_maxpool/xpulp_tnn_conv1d_ternary.h
@@ -0,0 +1,35 @@
+#ifndef __XPULP_NN_CONV1D_TERNARY_H
+#define __XPULP_NN_CONV1D_TERNARY_H
+#include <stdio.h>
+#include "pulp_nn_utils.h"
+
+
+
+// TODO: review argument order
+void __attribute__((noinline)) xpulp_tnn_conv1d_ternary(
+                                                        uint8_t  *pIn,
+                                                        uint8_t  *pIm2ColBuffer,
+                                                        int8_t   *pBias,
+                                                        uint8_t  *pOut,
+                                                        uint8_t  *pWeight,
+                                                        uint32_t *pThr,
+                                                        uint16_t  dim_in_x,
+                                                        uint16_t  ch_in,
+                                                        uint16_t  dim_out_x,
+                                                        uint16_t  ch_out,
+                                                        uint16_t  dim_kernel_x,
+                                                        uint16_t  padding_x_left,
+                                                        uint16_t  padding_x_right,
+                                                        uint16_t  stride_x,
+#ifndef PROFILE                                         
+                                                        uint16_t  dilation_x);
+#else
+                                                        uint16_t  dilation_x,
+                                                        uint32_t *im2col_cycles,
+                                                        uint32_t *hotloop_prep_cycles,
+                                                        uint32_t *hotloop_cycles,
+                                                        uint32_t *threshold_cycles,
+                                                        uint32_t *hotloop_leftover_cycles,
+                                                        uint32_t *matmul4x2_leftover_cycles);
+#endif
+#endif
diff --git a/rt_nn_tests/xptnn_maxpool/xpulp_tnn_conv1d_ternary_signed.h b/rt_nn_tests/xptnn_maxpool/xpulp_tnn_conv1d_ternary_signed.h
new file mode 100644
index 0000000..0f8ba08
--- /dev/null
+++ b/rt_nn_tests/xptnn_maxpool/xpulp_tnn_conv1d_ternary_signed.h
@@ -0,0 +1,35 @@
+#ifndef __XPULP_NN_CONV1D_TERNARY_SIGNED_H
+#define __XPULP_NN_CONV1D_TERNARY_SIGNED_H
+#include <stdio.h>
+#include "pulp_nn_utils.h"
+
+
+
+// TODO: review argument order
+void __attribute__((noinline)) xpulp_tnn_conv1d_ternary_signed(
+                                                        uint8_t  *pIn,
+                                                        uint8_t  *pIm2ColBuffer,
+                                                        int8_t   *pBias,
+                                                        uint8_t  *pOut,
+                                                        uint8_t  *pWeight,
+                                                        uint32_t *pThr,
+                                                        uint16_t  dim_in_x,
+                                                        uint16_t  ch_in,
+                                                        uint16_t  dim_out_x,
+                                                        uint16_t  ch_out,
+                                                        uint16_t  dim_kernel_x,
+                                                        uint16_t  padding_x_left,
+                                                        uint16_t  padding_x_right,
+                                                        uint16_t  stride_x,
+#ifndef PROFILE                                         
+                                                        uint16_t  dilation_x);
+#else
+                                                        uint16_t  dilation_x,
+                                                        uint32_t *im2col_cycles,
+                                                        uint32_t *hotloop_prep_cycles,
+                                                        uint32_t *hotloop_cycles,
+                                                        uint32_t *threshold_cycles,
+                                                        uint32_t *hotloop_leftover_cycles,
+                                                        uint32_t *matmul4x2_leftover_cycles);
+#endif
+#endif
diff --git a/rt_nn_tests/xptnn_maxpool/xpulp_tnn_conv_ternary.h b/rt_nn_tests/xptnn_maxpool/xpulp_tnn_conv_ternary.h
new file mode 100644
index 0000000..f4a8759
--- /dev/null
+++ b/rt_nn_tests/xptnn_maxpool/xpulp_tnn_conv_ternary.h
@@ -0,0 +1,40 @@
+#ifndef __XPULP_NN_CONV_TERNARY_H
+#define __XPULP_NN_CONV_TERNARY_H
+#include <stdio.h>
+#include "pulp_nn_utils.h"
+
+
+
+// TODO: review argument order
+void __attribute__((noinline)) xpulp_tnn_conv_ternary(
+                                                     uint8_t  *pIn,
+                                                     uint8_t  *pIm2ColBuffer,
+                                                     int8_t   *pBias,
+                                                     uint8_t  *pOut,
+                                                     uint8_t  *pWeight,
+                                                     uint32_t *pThr,
+                                                     uint16_t  dim_in_x,
+                                                     uint16_t  dim_in_y,
+                                                     uint16_t  ch_in,
+                                                     uint16_t  dim_out_x,
+                                                     uint16_t  dim_out_y,
+                                                     uint16_t  ch_out,
+                                                     uint16_t  dim_kernel_x,
+                                                     uint16_t  dim_kernel_y,
+                                                     uint16_t  padding_y_top,
+                                                     uint16_t  padding_y_bottom,
+                                                     uint16_t  padding_x_left,
+                                                     uint16_t  padding_x_right,
+                                                     uint16_t  stride_x,
+#ifndef PROFILE
+                                                     uint16_t  stride_y);
+#else
+                                                     uint16_t  stride_y,
+                                                     uint32_t *im2col_cycles,
+                                                     uint32_t *hotloop_prep_cycles,
+                                                     uint32_t *hotloop_cycles,
+                                                     uint32_t *threshold_cycles,
+                                                     uint32_t *hotloop_leftover_cycles,
+                                                     uint32_t *matmul4x2_leftover_cycles);
+#endif
+#endif
diff --git a/rt_nn_tests/xptnn_maxpool/xpulp_tnn_conv_ternary_signed.h b/rt_nn_tests/xptnn_maxpool/xpulp_tnn_conv_ternary_signed.h
new file mode 100644
index 0000000..d05c2ab
--- /dev/null
+++ b/rt_nn_tests/xptnn_maxpool/xpulp_tnn_conv_ternary_signed.h
@@ -0,0 +1,40 @@
+#ifndef __XPULP_NN_CONV_TERNARY_SIGNED_H
+#define __XPULP_NN_CONV_TERNARY_SIGNED_H
+#include <stdio.h>
+#include "pulp_nn_utils.h"
+
+
+
+// TODO: review argument order
+void __attribute__((noinline)) xpulp_tnn_conv_ternary_signed(
+                                                     uint8_t  *pIn,
+                                                     uint8_t  *pIm2ColBuffer,
+                                                     int8_t   *pBias,
+                                                     uint8_t  *pOut,
+                                                     uint8_t  *pWeight,
+                                                     uint32_t *pThr,
+                                                     uint16_t  dim_in_x,
+                                                     uint16_t  dim_in_y,
+                                                     uint16_t  ch_in,
+                                                     uint16_t  dim_out_x,
+                                                     uint16_t  dim_out_y,
+                                                     uint16_t  ch_out,
+                                                     uint16_t  dim_kernel_x,
+                                                     uint16_t  dim_kernel_y,
+                                                     uint16_t  padding_y_top,
+                                                     uint16_t  padding_y_bottom,
+                                                     uint16_t  padding_x_left,
+                                                     uint16_t  padding_x_right,
+                                                     uint16_t  stride_x,
+#ifndef PROFILE
+                                                     uint16_t  stride_y);
+#else
+                                                     uint16_t  stride_y,
+                                                     uint32_t *im2col_cycles,
+                                                     uint32_t *hotloop_prep_cycles,
+                                                     uint32_t *hotloop_cycles,
+                                                     uint32_t *threshold_cycles,
+                                                     uint32_t *hotloop_leftover_cycles,
+                                                     uint32_t *matmul4x2_leftover_cycles);
+#endif
+#endif
diff --git a/rt_nn_tests/xptnn_maxpool/xpulp_tnn_linear_ternary_i32.h b/rt_nn_tests/xptnn_maxpool/xpulp_tnn_linear_ternary_i32.h
new file mode 100644
index 0000000..5c353d3
--- /dev/null
+++ b/rt_nn_tests/xptnn_maxpool/xpulp_tnn_linear_ternary_i32.h
@@ -0,0 +1,12 @@
+#ifndef __XPULP_TNN_LINEAR_TERNARY_I32_H
+#define __XPULP_TNN_LINEAR_TERNARY_I32_H
+
+void __attribute__((noinline)) xpulp_tnn_linear_ternary_i32(
+                   uint8_t *pIn,
+                   int8_t *pBias,
+                   int8_t *pOut,
+                   int8_t *pWeight,
+                   uint16_t dim_vec,
+                   uint16_t num_o_neurons);
+
+#endif
\ No newline at end of file
diff --git a/rt_nn_tests/xptnn_maxpool/xpulp_tnn_linear_ternary_i32_signed.h b/rt_nn_tests/xptnn_maxpool/xpulp_tnn_linear_ternary_i32_signed.h
new file mode 100644
index 0000000..218c5c4
--- /dev/null
+++ b/rt_nn_tests/xptnn_maxpool/xpulp_tnn_linear_ternary_i32_signed.h
@@ -0,0 +1,12 @@
+#ifndef __XPULP_TNN_LINEAR_TERNARY_I32_SIGNED_H
+#define __XPULP_TNN_LINEAR_TERNARY_I32_SIGNED_H
+
+void __attribute__((noinline)) xpulp_tnn_linear_ternary_i32_signed(
+                   int8_t *pIn,
+                   int8_t *pBias,
+                   int8_t *pOut,
+                   int8_t *pWeight,
+                   uint16_t dim_vec,
+                   uint16_t num_o_neurons);
+
+#endif
\ No newline at end of file
diff --git a/rt_nn_tests/xptnn_maxpool/xpulp_tnn_matmul_ternary.h b/rt_nn_tests/xptnn_maxpool/xpulp_tnn_matmul_ternary.h
new file mode 100644
index 0000000..914daee
--- /dev/null
+++ b/rt_nn_tests/xptnn_maxpool/xpulp_tnn_matmul_ternary.h
@@ -0,0 +1,113 @@
+#ifndef __XPULP_TNN_MATMUL_TERNARY_H
+#define __XPULP_TNN_MATMUL_TERNARY_H
+
+#include "pulp_nn_utils.h"
+
+uint8_t * __attribute__((noinline)) xpulp_tnn_matmul_ternary(
+                        uint8_t  *pIn,
+                        int8_t   *pBias,
+                        uint32_t *pThr,
+                        uint8_t  *pOut,
+                        uint8_t  *pOut2,
+                        uint8_t  *pWeight,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+#ifndef PROFILE
+                        uint32_t *thrc_res1,
+                        uint32_t *thrc_res2);
+#else
+                        uint32_t *thrc_res1,
+                        uint32_t *thrc_res2,
+                        uint32_t *hotloop_prep_cycles,
+                        uint32_t *hotloop_cycles,
+                        uint32_t *threshold_cycles,
+                        uint32_t *hotloop_leftover_cycles);
+#endif
+
+inline void __attribute__((aligned(4))) hotloop_4x2(
+  int *sum, int *sum2, int *sum3, int *sum4, int *sum5, int *sum6, int *sum7, int *sum8,
+  uint32_t *ptrA, uint32_t *ptrA2, uint32_t *ptrA3, uint32_t *ptrA4,
+  uint32_t *ptrB, uint32_t *ptrB2, int num_col_im2col_words
+){
+  for (int j=0; j<num_col_im2col_words; j++)
+  {
+    *sum = MacLoad20(0, 0, 0, 0, ptrA, *sum);
+    *sum2 = MacLoad20(0, 0, 1, 0, ptrA2, *sum2);
+    *sum3 = MacLoad20(0, 0, 2, 0, ptrA3, *sum3);
+    *sum4 = MacLoad20(0, 1, 3, 0, ptrB, *sum4);
+    ptrB = MacLoadUpdate(ptrB);
+
+    *sum5 = MacLoad20(1, 0, 0, 1, ptrA, *sum5);
+    ptrA = MacLoadUpdate(ptrA);
+    *sum6 = MacLoad20(1, 0, 1, 1, ptrA2, *sum6);
+    ptrA2 = MacLoadUpdate(ptrA2);
+    *sum7 = MacLoad20(1, 0, 2, 1, ptrA3, *sum7);
+    ptrA3 = MacLoadUpdate(ptrA3);
+    *sum8 = MacLoad20(1, 0, 3, 1, ptrA4, *sum8);
+    ptrA4 = MacLoadUpdate(ptrA4);
+
+    ptrB2 = MacLoadInit(0, 1, 0, 1, ptrB2);
+  }
+}
+
+
+inline void __attribute__((aligned(4))) hotloop_3x2(
+  int *sum, int *sum2, int *sum3, int *sum4, int *sum5, int *sum6,
+  uint32_t *ptrA, uint32_t *ptrA2, uint32_t *ptrA3, uint32_t *ptrB, uint32_t *ptrB2, int num_col_im2col_words
+){
+  for (int j=0; j<num_col_im2col_words; j++)
+  {
+    *sum = MacLoad20(0, 0, 0, 0, ptrA, *sum);
+    *sum2 = MacLoad20(0, 0, 1, 0, ptrA2, *sum2);
+    *sum3 = MacLoad20(0, 1, 2, 0, ptrB, *sum3);
+    ptrB = MacLoadUpdate(ptrB);
+
+    *sum4 = MacLoad20(1, 0, 0, 1, ptrA, *sum4);
+    ptrA = MacLoadUpdate(ptrA);
+    *sum5 = MacLoad20(1, 0, 1, 1, ptrA2, *sum5);
+    ptrA2 = MacLoadUpdate(ptrA2);
+    *sum6 = MacLoad20(1, 0, 2, 1, ptrA3, *sum6);
+    ptrA3 = MacLoadUpdate(ptrA3);
+
+    ptrB2 = MacLoadInit(0, 1, 0, 1, ptrB2);
+  }
+}
+
+
+inline void __attribute__((aligned(4))) hotloop_2x2(
+  int *sum, int *sum2, int *sum3, int *sum4,
+  uint32_t *ptrA, uint32_t *ptrA2, uint32_t *ptrB, uint32_t *ptrB2, int num_col_im2col_words
+){
+  for (int j=0; j<num_col_im2col_words; j++)
+  {
+    *sum = MacLoad20(0, 0, 0, 0, ptrA, *sum);
+    *sum2 = MacLoad20(0, 1, 1, 0, ptrB, *sum2);
+    ptrB = MacLoadUpdate(ptrB);
+
+    *sum3 = MacLoad20(1, 0, 0, 1, ptrA, *sum3);
+    ptrA = MacLoadUpdate(ptrA);
+    *sum4 = MacLoad20(1, 0, 1, 1, ptrA2, *sum4);
+    ptrA2 = MacLoadUpdate(ptrA2);
+
+    ptrB2 = MacLoadInit(0, 1, 0, 1, ptrB2);
+  }
+}
+
+
+inline void __attribute__((aligned(4))) hotloop_1x2(
+  int *sum, int *sum2, uint32_t *ptrA, uint32_t *ptrB, uint32_t *ptrB2, int num_col_im2col_words
+){
+  for (int j=0; j<num_col_im2col_words; j++)
+  {
+    *sum = MacLoad20(0, 1, 0, 0, ptrB, *sum);
+    ptrB = MacLoadUpdate(ptrB);
+    *sum2 = MacLoad20(1, 0, 0, 1, ptrA, *sum2);
+    ptrA = MacLoadUpdate(ptrA);
+
+    ptrB2 = MacLoadInit(0, 1, 0, 1, ptrB2);
+  }
+}
+
+
+
+#endif
\ No newline at end of file
diff --git a/rt_nn_tests/xptnn_maxpool/xpulp_tnn_matmul_ternary_4x1.h b/rt_nn_tests/xptnn_maxpool/xpulp_tnn_matmul_ternary_4x1.h
new file mode 100644
index 0000000..fd06e93
--- /dev/null
+++ b/rt_nn_tests/xptnn_maxpool/xpulp_tnn_matmul_ternary_4x1.h
@@ -0,0 +1,96 @@
+#ifndef __XPULP_TNN_MATMUL_TERNARY_4X1_H
+#define __XPULP_TNN_MATMUL_TERNARY_4X1_H
+
+#include "pulp_nn_utils.h"
+
+uint8_t * __attribute__((noinline)) xpulp_tnn_matmul_ternary_4x1(
+                        uint8_t  *pIn,
+                        int8_t   *pBias,
+                        uint32_t *pThr,
+                        uint8_t  *pOut,
+                        uint8_t  *pWeight,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+#ifndef PROFILE
+                        uint32_t *thrc_res);
+#else
+                        uint32_t *thrc_res,
+                        uint32_t *hotloop_prep_cycles,
+                        uint32_t *hotloop_cycles,
+                        uint32_t *threshold_cycles,
+                        uint32_t *hotloop_leftover_cycles);
+#endif
+
+inline void __attribute__((aligned(4))) hotloop_4x1(
+  int *sum, int *sum2, int *sum3, int *sum4,
+  uint32_t *ptrA, uint32_t *ptrA2, uint32_t *ptrA3, uint32_t *ptrA4,
+  uint32_t *ptrB, int num_col_im2col_words
+){
+  for (int j=0; j<num_col_im2col_words; j++)
+  {
+    *sum = MacLoad20(1, 0, 0, 0, ptrA, *sum);
+    ptrA = MacLoadUpdate(ptrA);
+    *sum2 = MacLoad20(1, 0, 1, 0, ptrA2, *sum2);
+    ptrA2 = MacLoadUpdate(ptrA2);
+    *sum3 = MacLoad20(1, 0, 2, 0, ptrA3, *sum3);
+    ptrA3 = MacLoadUpdate(ptrA3);
+    *sum4 = MacLoad20(1, 0, 3, 0, ptrA4, *sum4);
+    ptrA4 = MacLoadUpdate(ptrA4);
+
+    ptrB = MacLoadInit(0, 1, 0, 0, ptrB);
+    asm volatile ("nop;");
+  }
+}
+
+
+inline void __attribute__((aligned(4))) hotloop_3x1(
+  int *sum, int *sum2, int *sum3,
+  uint32_t *ptrA, uint32_t *ptrA2, uint32_t *ptrA3, uint32_t *ptrB, int num_col_im2col_words
+){
+  for (int j=0; j<num_col_im2col_words; j++)
+  {
+    *sum = MacLoad20(1, 0, 0, 0, ptrA, *sum);
+    ptrA = MacLoadUpdate(ptrA);
+    *sum2 = MacLoad20(1, 0, 1, 0, ptrA2, *sum2);
+    ptrA2 = MacLoadUpdate(ptrA2);
+    *sum3 = MacLoad20(1, 0, 2, 0, ptrA3, *sum3);
+    ptrA3 = MacLoadUpdate(ptrA3);
+
+    ptrB = MacLoadInit(0, 1, 0, 0, ptrB);
+    asm volatile ("nop;");
+  }
+}
+
+
+inline void __attribute__((aligned(4))) hotloop_2x1(
+  int *sum, int *sum2, uint32_t *ptrA, uint32_t *ptrA2, uint32_t *ptrB, int num_col_im2col_words
+){
+  for (int j=0; j<num_col_im2col_words; j++)
+  {
+    *sum = MacLoad20(1, 0, 0, 0, ptrA, *sum);
+    ptrA = MacLoadUpdate(ptrA);
+    *sum2 = MacLoad20(1, 0, 1, 0, ptrA2, *sum2);
+    ptrA2 = MacLoadUpdate(ptrA2);
+
+    ptrB = MacLoadInit(0, 1, 0, 0, ptrB);
+    asm volatile ("nop;");
+  }
+}
+
+
+inline void __attribute__((aligned(4))) hotloop_1x1(
+  int *sum, uint32_t *ptrA, uint32_t *ptrB, int num_col_im2col_words
+){
+  for (int j=0; j<num_col_im2col_words; j++)
+  {
+    *sum = MacLoad20(1, 0, 0, 0, ptrA, *sum);
+    ptrA = MacLoadUpdate(ptrA);
+
+    ptrB = MacLoadInit(0, 1, 0, 0, ptrB);
+    asm volatile ("nop;");
+  }
+}
+
+
+
+#endif
\ No newline at end of file
diff --git a/rt_nn_tests/xptnn_maxpool/xpulp_tnn_matmul_ternary_signed.h b/rt_nn_tests/xptnn_maxpool/xpulp_tnn_matmul_ternary_signed.h
new file mode 100644
index 0000000..1493b5d
--- /dev/null
+++ b/rt_nn_tests/xptnn_maxpool/xpulp_tnn_matmul_ternary_signed.h
@@ -0,0 +1,113 @@
+#ifndef __XPULP_TNN_MATMUL_TERNARY_SIGNED_H
+#define __XPULP_TNN_MATMUL_TERNARY_SIGNED_H
+
+#include "pulp_nn_utils.h"
+
+uint8_t * __attribute__((noinline)) xpulp_tnn_matmul_ternary_signed(
+                        uint8_t  *pIn,
+                        int8_t   *pBias,
+                        uint32_t *pThr,
+                        uint8_t  *pOut,
+                        uint8_t  *pOut2,
+                        uint8_t  *pWeight,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+#ifndef PROFILE
+                        uint32_t *thrc_res1,
+                        uint32_t *thrc_res2);
+#else
+                        uint32_t *thrc_res1,
+                        uint32_t *thrc_res2,
+                        uint32_t *hotloop_prep_cycles,
+                        uint32_t *hotloop_cycles,
+                        uint32_t *threshold_cycles,
+                        uint32_t *hotloop_leftover_cycles);
+#endif
+
+inline void __attribute__((aligned(4))) hotloop_4x2(
+  int *sum, int *sum2, int *sum3, int *sum4, int *sum5, int *sum6, int *sum7, int *sum8,
+  uint32_t *ptrA, uint32_t *ptrA2, uint32_t *ptrA3, uint32_t *ptrA4,
+  uint32_t *ptrB, uint32_t *ptrB2, int num_col_im2col_words
+){
+  for (int j=0; j<num_col_im2col_words; j++)
+  {
+    *sum = MacLoads20(0, 0, 0, 0, ptrA, *sum);
+    *sum2 = MacLoads20(0, 0, 1, 0, ptrA2, *sum2);
+    *sum3 = MacLoads20(0, 0, 2, 0, ptrA3, *sum3);
+    *sum4 = MacLoads20(0, 1, 3, 0, ptrB, *sum4);
+    ptrB = MacLoadUpdate(ptrB);
+
+    *sum5 = MacLoads20(1, 0, 0, 1, ptrA, *sum5);
+    ptrA = MacLoadUpdate(ptrA);
+    *sum6 = MacLoads20(1, 0, 1, 1, ptrA2, *sum6);
+    ptrA2 = MacLoadUpdate(ptrA2);
+    *sum7 = MacLoads20(1, 0, 2, 1, ptrA3, *sum7);
+    ptrA3 = MacLoadUpdate(ptrA3);
+    *sum8 = MacLoads20(1, 0, 3, 1, ptrA4, *sum8);
+    ptrA4 = MacLoadUpdate(ptrA4);
+
+    ptrB2 = MacLoadInit(0, 1, 0, 1, ptrB2);
+  }
+}
+
+
+inline void __attribute__((aligned(4))) hotloop_3x2(
+  int *sum, int *sum2, int *sum3, int *sum4, int *sum5, int *sum6,
+  uint32_t *ptrA, uint32_t *ptrA2, uint32_t *ptrA3, uint32_t *ptrB, uint32_t *ptrB2, int num_col_im2col_words
+){
+  for (int j=0; j<num_col_im2col_words; j++)
+  {
+    *sum = MacLoads20(0, 0, 0, 0, ptrA, *sum);
+    *sum2 = MacLoads20(0, 0, 1, 0, ptrA2, *sum2);
+    *sum3 = MacLoads20(0, 1, 2, 0, ptrB, *sum3);
+    ptrB = MacLoadUpdate(ptrB);
+
+    *sum4 = MacLoads20(1, 0, 0, 1, ptrA, *sum4);
+    ptrA = MacLoadUpdate(ptrA);
+    *sum5 = MacLoads20(1, 0, 1, 1, ptrA2, *sum5);
+    ptrA2 = MacLoadUpdate(ptrA2);
+    *sum6 = MacLoads20(1, 0, 2, 1, ptrA3, *sum6);
+    ptrA3 = MacLoadUpdate(ptrA3);
+
+    ptrB2 = MacLoadInit(0, 1, 0, 1, ptrB2);
+  }
+}
+
+
+inline void __attribute__((aligned(4))) hotloop_2x2(
+  int *sum, int *sum2, int *sum3, int *sum4,
+  uint32_t *ptrA, uint32_t *ptrA2, uint32_t *ptrB, uint32_t *ptrB2, int num_col_im2col_words
+){
+  for (int j=0; j<num_col_im2col_words; j++)
+  {
+    *sum = MacLoads20(0, 0, 0, 0, ptrA, *sum);
+    *sum2 = MacLoads20(0, 1, 1, 0, ptrB, *sum2);
+    ptrB = MacLoadUpdate(ptrB);
+
+    *sum3 = MacLoads20(1, 0, 0, 1, ptrA, *sum3);
+    ptrA = MacLoadUpdate(ptrA);
+    *sum4 = MacLoads20(1, 0, 1, 1, ptrA2, *sum4);
+    ptrA2 = MacLoadUpdate(ptrA2);
+
+    ptrB2 = MacLoadInit(0, 1, 0, 1, ptrB2);
+  }
+}
+
+
+inline void __attribute__((aligned(4))) hotloop_1x2(
+  int *sum, int *sum2, uint32_t *ptrA, uint32_t *ptrB, uint32_t *ptrB2, int num_col_im2col_words
+){
+  for (int j=0; j<num_col_im2col_words; j++)
+  {
+    *sum = MacLoads20(0, 1, 0, 0, ptrB, *sum);
+    ptrB = MacLoadUpdate(ptrB);
+    *sum2 = MacLoads20(1, 0, 0, 1, ptrA, *sum2);
+    ptrA = MacLoadUpdate(ptrA);
+
+    ptrB2 = MacLoadInit(0, 1, 0, 1, ptrB2);
+  }
+}
+
+
+
+#endif
\ No newline at end of file
diff --git a/rt_nn_tests/xptnn_maxpool/xpulp_tnn_matmul_ternary_signed_4x1.h b/rt_nn_tests/xptnn_maxpool/xpulp_tnn_matmul_ternary_signed_4x1.h
new file mode 100644
index 0000000..a847bfb
--- /dev/null
+++ b/rt_nn_tests/xptnn_maxpool/xpulp_tnn_matmul_ternary_signed_4x1.h
@@ -0,0 +1,96 @@
+#ifndef __XPULP_TNN_MATMUL_TERNARY_SIGNED_4X1_H
+#define __XPULP_TNN_MATMUL_TERNARY_SIGNED_4X1_H
+
+#include "pulp_nn_utils.h"
+
+uint8_t * __attribute__((noinline)) xpulp_tnn_matmul_ternary_signed_4x1(
+                        uint8_t  *pIn,
+                        int8_t   *pBias,
+                        uint32_t *pThr,
+                        uint8_t  *pOut,
+                        uint8_t  *pWeight,
+                        uint16_t num_col_im2col,
+                        uint16_t ch_out,
+#ifndef PROFILE
+                        uint32_t *thrc_res);
+#else
+                        uint32_t *thrc_res,
+                        uint32_t *hotloop_prep_cycles,
+                        uint32_t *hotloop_cycles,
+                        uint32_t *threshold_cycles,
+                        uint32_t *hotloop_leftover_cycles);
+#endif
+
+inline void __attribute__((aligned(4))) hotloop_4x1(
+  int *sum, int *sum2, int *sum3, int *sum4,
+  uint32_t *ptrA, uint32_t *ptrA2, uint32_t *ptrA3, uint32_t *ptrA4,
+  uint32_t *ptrB, int num_col_im2col_words
+){
+  for (int j=0; j<num_col_im2col_words; j++)
+  {
+    *sum = MacLoads20(1, 0, 0, 0, ptrA, *sum);
+    ptrA = MacLoadUpdate(ptrA);
+    *sum2 = MacLoads20(1, 0, 1, 0, ptrA2, *sum2);
+    ptrA2 = MacLoadUpdate(ptrA2);
+    *sum3 = MacLoads20(1, 0, 2, 0, ptrA3, *sum3);
+    ptrA3 = MacLoadUpdate(ptrA3);
+    *sum4 = MacLoads20(1, 0, 3, 0, ptrA4, *sum4);
+    ptrA4 = MacLoadUpdate(ptrA4);
+
+    ptrB = MacLoadInit(0, 1, 0, 0, ptrB);
+    asm volatile ("nop;");
+  }
+}
+
+
+inline void __attribute__((aligned(4))) hotloop_3x1(
+  int *sum, int *sum2, int *sum3,
+  uint32_t *ptrA, uint32_t *ptrA2, uint32_t *ptrA3, uint32_t *ptrB, int num_col_im2col_words
+){
+  for (int j=0; j<num_col_im2col_words; j++)
+  {
+    *sum = MacLoads20(1, 0, 0, 0, ptrA, *sum);
+    ptrA = MacLoadUpdate(ptrA);
+    *sum2 = MacLoads20(1, 0, 1, 0, ptrA2, *sum2);
+    ptrA2 = MacLoadUpdate(ptrA2);
+    *sum3 = MacLoads20(1, 0, 2, 0, ptrA3, *sum3);
+    ptrA3 = MacLoadUpdate(ptrA3);
+
+    ptrB = MacLoadInit(0, 1, 0, 0, ptrB);
+    asm volatile ("nop;");
+  }
+}
+
+
+inline void __attribute__((aligned(4))) hotloop_2x1(
+  int *sum, int *sum2, uint32_t *ptrA, uint32_t *ptrA2, uint32_t *ptrB, int num_col_im2col_words
+){
+  for (int j=0; j<num_col_im2col_words; j++)
+  {
+    *sum = MacLoads20(1, 0, 0, 0, ptrA, *sum);
+    ptrA = MacLoadUpdate(ptrA);
+    *sum2 = MacLoads20(1, 0, 1, 0, ptrA2, *sum2);
+    ptrA2 = MacLoadUpdate(ptrA2);
+
+    ptrB = MacLoadInit(0, 1, 0, 0, ptrB);
+    asm volatile ("nop;");
+  }
+}
+
+
+inline void __attribute__((aligned(4))) hotloop_1x1(
+  int *sum, uint32_t *ptrA, uint32_t *ptrB, int num_col_im2col_words
+){
+  for (int j=0; j<num_col_im2col_words; j++)
+  {
+    *sum = MacLoads20(1, 0, 0, 0, ptrA, *sum);
+    ptrA = MacLoadUpdate(ptrA);
+
+    ptrB = MacLoadInit(0, 1, 0, 0, ptrB);
+    asm volatile ("nop;");
+  }
+}
+
+
+
+#endif
\ No newline at end of file
diff --git a/rt_nn_tests/xptnn_maxpool/xpulp_tnn_maxpool_ternary.c b/rt_nn_tests/xptnn_maxpool/xpulp_tnn_maxpool_ternary.c
new file mode 100644
index 0000000..ed9f87d
--- /dev/null
+++ b/rt_nn_tests/xptnn_maxpool/xpulp_tnn_maxpool_ternary.c
@@ -0,0 +1,159 @@
+/*
+ * xpulp_tnn_maxpool_ternary.c
+ * Georg Rutishauser <georgr@iis.ee.ethz.ch>
+ * Nazareno Bruschi <nazareno.bruschi@unibo.it>
+ * Angelo Garofalo <angelo.garofalo@unibo.it>
+ *
+ * Copyright (C) 2018-2020 University of Bologna
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include "pmsis.h"
+#include "pulp_nn_utils.h"
+
+
+void __attribute__ ((noinline)) xpulp_tnn_maxpool_ternary(
+                                                          int8_t * pIn,
+                                                          int8_t * pOut,
+                                                          uint16_t  dim_im_in_x,
+                                                          uint16_t  dim_im_in_y,
+                                                          // ch_im_in: UNCOMPRESSED number of input/output channels
+                                                          uint16_t  ch_im_in,
+                                                          uint16_t  dim_im_out_x,
+                                                          uint16_t  dim_im_out_y,
+                                                          uint16_t  dim_kernel_x,
+                                                          uint16_t  dim_kernel_y,
+                                                          uint16_t  padding_t,
+                                                          uint16_t  padding_b,
+                                                          uint16_t  padding_l,
+                                                          uint16_t  padding_r,
+                                                          uint16_t  stride_x,
+                                                          uint16_t  stride_y)
+
+{
+  #ifdef FC_TEST
+  int n_cores = 1;
+  int core_id = 0;
+  # else
+  int n_cores = NUM_CORES;
+  int core_id = pi_core_id();
+  if (dim_im_in_y < NUM_CORES)
+  {
+    n_cores = dim_im_in_y;
+  }
+  #endif
+  int  Log2Core = log2(n_cores);
+  // number of bytes:  5 els per byte
+  int ch_im_in_r = ch_im_in/5;
+
+
+  int chunck = (dim_im_in_y >> Log2Core) + ((dim_im_in_y & (NUM_CORES-1))!=0);
+
+  int start = min(chunck * core_id, dim_im_in_y);
+  int stop = min(start + chunck, dim_im_in_y);
+  int   i_x, i_y;
+
+  // logic is the same as for i2, except we use the ternary version of xpulp_nn_compare_and_replace...
+  for (i_y = start; i_y < stop; i_y++)
+  {
+    for (i_x = 0; i_x < dim_im_out_x; i_x++)
+    {
+      /* for each output pixel */
+      int8_t     *target = pIn + (i_y * dim_im_in_x + i_x) * ch_im_in_r;
+      uint8_t     *win_start;
+      uint8_t     *win_stop;
+      if (i_x * stride_x - padding_l < 0)
+      {
+        win_start = target;
+      }
+      else
+      {
+        win_start = pIn + (i_y * dim_im_in_x + i_x * stride_x - padding_l) * ch_im_in_r;
+      }
+
+      if (i_x * stride_x - padding_l + dim_kernel_x >= dim_im_in_x)
+      {
+        win_stop = pIn + (i_y * dim_im_in_x + dim_im_in_x) * ch_im_in_r;
+      }
+      else
+      {
+        win_stop = pIn + (i_y * dim_im_in_x + i_x * stride_x - padding_l + dim_kernel_x) * ch_im_in_r;
+      }
+
+      /* first step is to copy over initial data */
+      for (int i = 0; i< ch_im_in_r; i++) target[i] = win_start[i];
+
+      /* start the max operation from the second part */
+      win_start += ch_im_in_r;
+      for (; win_start < win_stop; win_start += ch_im_in_r)
+      {
+        xpulp_tnn_compare_and_replace_if_larger_ternary(target, win_start, ch_im_in_r);
+      }
+    }
+  }
+#ifndef FC_TEST
+  pi_cl_team_barrier(0);
+#endif
+  if (dim_im_out_y < NUM_CORES)
+  {
+    n_cores = dim_im_out_y;
+  }
+  Log2Core = log2(n_cores);
+  int chunck2 = (dim_im_out_y >> Log2Core) + ((dim_im_out_y & (NUM_CORES-1))!=0);
+  int start2 = chunck2 * core_id;
+  int stop2 = min(start2 + chunck2, dim_im_out_y);
+
+    /* then does the pooling along y axis */
+  for (i_y = start2; i_y < stop2; i_y++)
+  {
+    /* for each output row */
+    int8_t *target = pOut + i_y * dim_im_out_x * ch_im_in_r;
+    int8_t *row_start;
+    int8_t *row_end;
+    /* setting the starting row */
+    if (i_y * stride_y - padding_t < 0)
+    {
+      row_start = pIn;
+    }
+    else
+    {
+      row_start = pIn + (i_y * stride_y - padding_t) * dim_im_in_x * ch_im_in_r;
+    }
+    /* setting the stopping row */
+    if (i_y * stride_y - padding_t + dim_kernel_y >= dim_im_in_y)
+    {
+      row_end = pIn + dim_im_in_y * dim_im_in_x * ch_im_in_r;
+    }
+    else
+    {
+      row_end = pIn + (i_y * stride_y - padding_t + dim_kernel_y) * dim_im_in_x * ch_im_in_r;
+    }
+
+    /* copy over the first row */
+    for (int i = 0; i< dim_im_out_x * ch_im_in_r; i++)
+    {
+      target[i] = (int8_t) row_start[i];
+    }
+    /* move over to next row */
+    row_start += ch_im_in_r * dim_im_in_x;
+
+    for (; row_start < row_end; row_start += dim_im_in_x * ch_im_in_r)
+    {
+      xpulp_tnn_compare_and_replace_if_larger_ternary(target, row_start, dim_im_out_x * ch_im_in_r);
+    }
+  }
+  #ifndef FC_TEST
+  pi_cl_team_barrier(0);
+  #endif
+}
diff --git a/rt_nn_tests/xptnn_maxpool/xpulp_tnn_maxpool_ternary.h b/rt_nn_tests/xptnn_maxpool/xpulp_tnn_maxpool_ternary.h
new file mode 100644
index 0000000..2b65fca
--- /dev/null
+++ b/rt_nn_tests/xptnn_maxpool/xpulp_tnn_maxpool_ternary.h
@@ -0,0 +1,20 @@
+#ifndef __XPULP_TNN_MAXPOOL_TERNARY_H
+#define __XPULP_TNN_MAXPOOL_TERNARY_H
+void __attribute__ ((noinline)) xpulp_tnn_maxpool_ternary(
+                                                          int8_t * pIn,
+                                                          int8_t * pOut,
+                                                          uint16_t  dim_im_in_x,
+                                                          uint16_t  dim_im_in_y,
+                                                          // ch_im_in: UNCOMPRESSED number of input/output channels
+                                                          uint16_t  ch_im_in,
+                                                          uint16_t  dim_im_out_x,
+                                                          uint16_t  dim_im_out_y,
+                                                          uint16_t  dim_kernel_x,
+                                                          uint16_t  dim_kernel_y,
+                                                          uint16_t  padding_t,
+                                                          uint16_t  padding_b,
+                                                          uint16_t  padding_l,
+                                                          uint16_t  padding_r,
+                                                          uint16_t  stride_x,
+                                                          uint16_t  stride_y);
+#endif